Conversion between markup formats

Version on this page:2.5
LTS Haskell 13.25:2.5
Stackage Nightly 2019-06-12:2.7.2
Latest on Hackage:2.7.3

See all snapshots pandoc appears in

GPL-2.0-only licensed and maintained by John MacFarlane

Module documentation for 2.5


github release hackage release homebrew stackage LTS package CircleCI appveyor build status license pandoc-discuss on google groups

The universal markup converter

Pandoc is a Haskell library for converting from one markup format to another, and a command-line tool that uses this library. It can convert from

It can convert to

Pandoc can also produce PDF output via LaTeX, Groff ms, or HTML.

Pandoc’s enhanced version of Markdown includes syntax for tables, definition lists, metadata blocks, footnotes, citations, math, and much more. See the User’s Manual below under Pandoc’s Markdown.

Pandoc has a modular design: it consists of a set of readers, which parse text in a given format and produce a native representation of the document (an abstract syntax tree or AST), and a set of writers, which convert this native representation into a target format. Thus, adding an input or output format requires only adding a reader or writer. Users can also run custom pandoc filters to modify the intermediate AST (see the documentation for filters and lua filters).

Because pandoc’s intermediate representation of a document is less expressive than many of the formats it converts between, one should not expect perfect conversions between every format and every other. Pandoc attempts to preserve the structural elements of a document, but not formatting details such as margin size. And some document elements, such as complex tables, may not fit into pandoc’s simple document model. While conversions from pandoc’s Markdown to all formats aspire to be perfect, conversions from formats more expressive than pandoc’s Markdown can be expected to be lossy.


Here’s how to install pandoc.


Pandoc’s website contains a full User’s Guide. It is also available here as pandoc-flavored Markdown. The website also contains some examples of the use of pandoc and a limited online demo.


Pull requests, bug reports, and feature requests are welcome. Please make sure to read the contributor guidelines before opening a new issue.


© 2006-2018 John MacFarlane ( Released under the GPL, version 2 or greater. This software carries no warranty of any kind. (See COPYRIGHT for full copyright and warranty notices.)


pandoc (2.7.3)

* Add `jira` (Atlassian's Jira wiki markup) as output format (#2497,
Albert Krewinkel).

* Add `tex_math_dollars` to `multimarkdownExtensions` (#5512).
This form is now supported in multimarkdown,
in addition to `tex_math_double_backslash`.

* Fix `--self-contained` so it works when output format has extensions.
Previously if you used `--self-contained` with `html-smart` or
`html+smart`, it wouldn't work.

* Add template variable `curdir` with working directory
from which pandoc is run (#5464).

* Markdown reader: don't create implicit reference for empty header

* Muse reader: allow images inside link descriptions (Alexander Krotov).

* HTML reader: epub related fixes.

+ With epub extensions, check for `epub:type` in addition to `type`.
+ Fix problem with noteref parsing which caused block-level
content to be eaten with the noteref.
+ Rename `pAnyTag` to `pAny`.
+ Refactor note resolution.
+ Trim definition list terms (Alexander Krotov).

* LaTeX reader:

+ Add braces when resolving `\DeclareMathOperator`
(#5441). These seem to be needed for xelatex but not pdflatex.
+ Allow newlines in `\mintinline`.
+ Pass through unknown listings language as class (#5540).
Previously if the language was not in the list of languages supported
by listings, it would not be added as a class, so highlighting
would not be triggered.
+ `rawLaTeXInline`: Include trailing `{}`s in raw latex commands (#5439).
This change affects the markdown reader and other readers that allow raw
LaTeX. Previously, trailing `{}` would be included for unknown
commands, but not for known commands. However, they are sometimes used
to avoid a trailing space after the command. The chances that a `{}`
after a LaTeX command is not part of the command are very small.

* MediaWiki reader: handle multiple attributes in table row (#5471,

* Docx reader: Add support for `w:rtl` (#5545). Elements with this
property are put into Span inlines with `dir="rtl"`.

* DocBook reader: Issue `IgnoredElement` warnings.

* Org reader (Albert Krewinkel):

+ Fix planning elements in headers level 3 and higher
(#5494). Planning info is now always placed before
the subtree contents. Previously, the planning info was placed after
the content if the header's subtree was converted to a list, which
happens with headers of level 3 and higher per default.
+ Omit, but warn about unknown export options. Unknown export
options are properly ignored and omitted from the output.
+ Prefer plain symbols over math symbols (#5483).
Symbols like `\alpha` are output plain and unemphasized, not as math.
+ Recognize emphasis after TODO/DONE keyword (#5484).

* FB2 reader:

+ Skip unknown elements rather than throwing errors (#5560).
Sometimes custom elements are used (e.g. `id` element
inside `author`); previously the reader would halt with an error.
Now it skips the element and issues an `IgnoredElement` warning.
+ Parse notes (#5493, Alexander Krotov).
+ Internal improvements (Alexander Krotov).

* OpenDocument writer: Roll back automatic figure/table numbering
(#5474). This was added in pandoc 2.7.2, but it makes it impossible
to use pandoc-crossref. So this has been rolled back for now,
until we find a good solution to make this behavior optional
(or a creative way to let pandoc-crossref and this feature
to coexist).

* New module Text.Pandoc.Writers.Jira, exporting `writeJira` [API
change] (Albert Krewinkel).

* EPUB writer:

+ Don't include 'landmarks' if there aren't any.
Previously we could get an empty ol element, which caused
validation errors with epubcheck.
+ Ensure unique ids for styleesheets in content.opf (#5463).
+ Make stylesheet link compatible with kindlegen (#5466,
Eric Schrijver). Pandoc omitted `type="text/css"` from both
`<style>` and `<rel="stylesheet">` elements in all templates, which
is valid according to the spec. However, Amazon’s kindlegen software
relies on this attribute on `<link>` elements when detecting
stylesheets to include.

* HTML writer:

+ Output video and audio elements depending on file
extension of the image path (Mauro Bieg).
+ Emit empty alt tag in figures (#5518, Mauro Bieg).
The same text is already in the <figcaption> and
screen-readers would read it twice, see #4737.
+ Don't add variation selector if it's already there.
This fixes round-trip failures.
+ Prevent gratuitious emojification on iOS (#5469).
iOS chooses to render a number of Unicode entities, including '↩', as
big colorful emoji. This can be defeated by appending Unicode VARIATION
SELECTOR-15'/'VARIATION SELECTOR-16'. So we now append this character
when escaping strings, for both '↩' and '↔'. If other characters prove
problematic, they can simply be added to `needsVariationSelector`.
+ Add `class="heading"` to level 7+ Headers rendered as `<p>` elements

* RST writer: treat Span with no attributes as transparent (#5446).
Previously an Emph inside a Span was being treated as
nested markup and ignored. With this patch, the Span
is just ignored.

* LaTeX writer:

+ Include inline code attributes with `--listings` (#5420).
+ Don't produce columns environment unless beamer (#5485).
+ Fix footnote in image caption. Regression: the fix for #4683 broke
this case.
+ Don't highlight code in headings (#5574). This causes
compilation errors.
+ Use `\mbox` to get proper behavior inside `\sout` (#5529).

* EPUB writer: Fix document section assignments (#5546).
For example, introduction should go in bodymatter, not frontmatter, and
epigraph, conclusion, and afterward should go in bodymatter, not
backmatter. For the full list of assignments, see the manual.

* Markdown writer:

+ Add backslashes to avoid unwanted interpretation of
definition list terms as other kinds of block (#554).
+ Ensure the code fence is long enough (#5519).
Previously too few backticks were used when the code block
contained an indented line of backticks. (Ditto tildes.)
+ Handle labels with integer names (Jesse Rosenthal, #5495).
Previously if labels had integer names, it could produce a conflict
with auto-labeled reference links. Now we test for a conflict and
find the next available integer. This involves adding a new state
variable `stPrevRefs` to keep track of refs used in other document
parts when using `--reference-location=block|section`

* Textile writer: fix closing tag for math output (Albert Krewinkel).
Opening and closing tag for math output match now.

* Org writer: always indent src blocks content by 2 spaces (#5440, Albert
Krewinkel). Emacs always uses two spaces when indenting the content of
src blocks, e.g., when exiting a `C-c '` edit-buffer. Pandoc used to
indent contents by the space-equivalent of one tab, but now always uses
two spaces, too.

* Asciidoc writer:

+ Use `` `+...+` `` form for inline code.
The old `` `a__b__c` `` yields emphasis inside code in asciidoc.
To get a pure literal code span, use `` `+a__b__c+` ``.
+ Use proper smart quotes with asciidoctor (#5487).
Asciidoctor has a different format for smart quotes.
+ Use doubled ## when necessary for spans (#5566).
+ Ensure correct nesting of strong/emph (#5565): strong
must be the outer element.

* JATS writer:

+ Wrap elements with p when needed (#5570). The JATS spec restricts
what elements can go inside `fn` and `list-item`. So we wrap other
elements inside `<p specific-use="wrapper">` when needed.
+ Properly handle footnotes (#5511) according to "best practice."
(Group them at the end in `<fn-group>` and use `<xref>` elements
to link them.)
+ Fix citations with PMID so they validate (#5481). This includes
an update to data/jats.csl.
+ Ensure validity of `<pub-date>` by parsing the date and
extracting year, month, and day, as expected. Also add an
iso-8601-date attribute automatically.
+ Don't use `<break>` element for LineBreak. It is only
allowed in a few special contexts, and not in `<p>` elements.
+ Don't make `<string-name>` a child of `<string>`, which is illegal.

* FB2 writer:

+ Do not wrap note references into `<sup>` and brackets
(Alexander Krotov). Existing FB2 readers, such as FBReader, already
display links with type="note" as a superscript.
+ Use genre metadata field (#5478).

* Muse writer: do not escape empty line after `<br>` (Alexander Krotov).

* Add unicode code point in "Missing character" warning (#5538).
If the character isn't in the console font, the message is pretty useless,
so we show the code point for anything non-ASCII.

* Lua: add Version type to simplify comparisons (Albert Krewinkel).
Version specifiers like `PANDOC_VERSION` and `PANDOC_API_VERSION` are
turned into `Version` objects. The objects simplify version-appropriate
comparisons while maintaining backward-compatibility.
A function `pandoc.types.Version` is added as part of the newly
introduced module `pandoc.types`, allowing users to create version
objects in scripts.

* pandoc lua module (Albert Krewinkel):

- Fix deletion of nonexistent attributes (#5569).
- Better tests for Attr and AttributeList.

* pandoc.mediabag lua module (Albert Krewinkel):

+ Add function `delete` for deleting a single item.
+ Add function `empty` for removing all entries.
+ Add function `items` for iterating over mediabag.

* Text.Pandoc.Class: Fix handling of `file:` URL scheme in
`downloadOrRead` (#5517, Mauro Bieg). Previously `file:/`
URLs were handled wrongly and pandoc attempted to make HTTP
requests, which failed.

* Text.Pandoc.MIME: add `mediaCategory` [API change] (Mauro Bieg).

* Text.Pandoc.Shared:

+ Add `onlySimpleTableCells` [API change]
(Mauro Bieg) and use this to consolidate simple-table detection
(#5524). This fixes an inconsistency in the HTML reader, which did not
treat tables with `<p>` inside cells as simple.
+ `metaToJSON`: treat digits starting with 0 as a string, not a number
(#5479). This fixes a regression in YAML metadata in pandoc 2.7.2.

* Text.Pandoc.Logging: Add `IgnoredElement` constructor for
`LogMessage`. `SkippedConetnt` doesn't work for some XML-based
readers which don't have access to parsec source positions.

* Text.Pandoc.Asciify: Add Turkish undotted-i (#5433, Mauro Bieg).

* Improve output of Lua tests (#5499, Albert Krewinkel).
This makes use of tasty-lua, a package to write tests in Lua
and integrate the results into Tasty output. Test output becomes
more informative: individual tests and test groups become visible
in test output. Failures are reported with helpful error messages.

* Lua: add `pandoc.system` module (#5468, Albert Krewinkel).
The `system` Lua module provides utility functions to interact with the
operating- and file system. E.g.
pandoc.system.with_temporary_directory('tikz', function (dir)
-- write and compile a TikZ file with pdflatex

* LaTeX template: Add pdflang to hypersetup if lang is set (#5443).

* beamer template: Fix using Beamer with geometry (#5505, Daniel Maslowski).
Beamer already loads geometry, so we need to use the `\geometry`
command to set geometry options.

* EPUB2/3 templates: Move inline styles to default epub.css (#5466).
NOTE: Those who use a custom CSS stylesheet with EPUB should
add these lines:

code{ white-space: pre-wrap; }
span.smallcaps{ font-variant: small-caps; }
span.underline{ text-decoration: underline; }
q { quotes: "“" "”" "‘" "’"; }
div.column{ display: inline-block; vertical-align: top; width: 50%; }

* reveal.js template:

+ Updates for revealjs 3.8.0 (#5435, ebiim).
+ Remove reference to head.min.js (#5448, Winnie Hellmann).
NOTE: users will need to update reveal.js to at least 3.8.0
for their presentations generated with this version of pandoc
to work correctly.

* Text.Pandoc.PDF:

+ Replace `</>` with literal `/` (#5451).
We use forward-slash for a directory separator in tmpDir,
even on Windows (because that's what tex likes). So we
should not put a backslash between the tmpDir and the
filename on Windows. This is harmless enough in normal
Windows setups, but it breaks on Cygwin.
Thanks to @cc2x for noticing and diagnosing the problem.
+ Allow use of `-output-directory` in `--pdf-engine-opt` (#5462).
This is currently possible with `mklatex` and `-outdir`, but
was not yet possible with xelatex and `-output-directory`.
+ For PDF via ms/pdfroff, make TOC appear at beginning and in
PDF bookmarks (#5553). Previously the TOC appeared at the end
of the document, and was not bookmarked. To keep it at the end,
add `--pdf-engine-opt=--no-toc-relocation` to your command line.

* Fix broken links in documents (#5473, Shim Myeongseob).
Also, use absolute links to when possible, so that
the links can be followed by people viewing these documents
on GitHub.

* Improved sample lua tikz filter in lua-filters docs (#5445,
Matthew Doty). There are three changes:

+ It only processes elements which begin with `\begin{tikzpicture}`
+ It uses pdf2svg instead of imagemagick to preserve fidelity
+ The images produced have transparent backgrounds

* MANUAL.txt:

+ Add note about `title-meta`, `author-meta`, `date-meta` (#5486).
+ Fix typo (#5489, Christian Krause).

* add test/tables.xwiki to git and pandoc.cabal (#5498, Mauro Bieg).

* Disable missing-home-modules warning in `stack.yaml`.
Otherwise `stack ghci` fails.

pandoc (2.7.2)

* Add XWiki writer (#1800, Derek Chen-Becker).
Add `Text.Pandoc.Writers.XWiki`, exporting `writeXWiki` [API change].

* Dokuwiki Reader: parse single curly brace (#5416, Mauro Bieg).

* Vimwiki reader: improve handling of internal links (#5414).
We no longer append `.html` to link targets, and we add a title
`wikilink`. This mirrors behavior of other wiki readers. Generally
the `.html` extension is not wanted. It may be important for output
to HTML in certain circumstances, but it can always be added using a
filter that matches on links with title `wikilink`.

If your workflow requires the current behavior, here is a lua filter
that will add the `.html` extension:

function Link(el)
if el.title == 'wikilink' then = .. ".html"
return el
* ipynb reader:

+ Use format `ipynb` for raw cell where no format given.
+ Avoid introducing spurious `.0` on integers in metadata.

* Markdown reader: fenced div takes priority over setext header.

* HTML reader: read `data-foo` attribute into `foo` (#5392).
The HTML writer adds the `data-` prefix for HTML5 for nonstandard
attributes. But the attributes are represented in the AST without
the `data-` prefix, so we should strip this when reading HTML.

* LaTeX reader: Improve autolink detection (#5340).

* PowerPoint writer (Jesse Rosenthal):

+ Expand builtin reference doc to model all layouts.
The previous built-in reference doc had only title and content
layouts. Add in a section-header slide and a two-content slide, so
users can more easily modify it to build their own templates.
+ Always open up in slide view. When editing a
template/reference-doc, the user might be in Master view, but when
producing a slide show, it is assumed that slide view will be
+ Remove `handoutsMasterList` from template presentation.xml
+ Fix numerous errors in templating (#5402). Previously, some templates
produced by Office 365 (MacOS) would not render with `--reference-doc`
correctly. We now apply correct shapes for content, and build
shape trees correctly.
+ Make default placeholder type for template lookup.
+ Apply speaker notes to metadata slide if applicable.
+ Test for speaker notes after breaking header.
+ Correctly handle notes after section-title header.
Previously, if notes came after a section-title header (ie, a level-1
header in a slide-level=2 presentation), they would go on the next
slide. This keeps them on the slide with the header.
+ Internal improvements.

* ipynb writer:

+ Use format `ipynb` for raw cell where no format given.
According to nbformat docs, this is supposed to render in every
format. We don't do that, but we at least preserve it as a raw
block in markdown, so you can round-trip.
+ Consolidate adjacent raw blocks. Sometimes pandoc creates two
HTML blocks, e.g. one for the open tag and one for a close tag.
If these aren't consolidated, only one will show up in output cell.
+ Fixed carry-over of nbformat from metadata.
+ Preserve `nbformat_minor` if it's given. This helps with round-tripping.

* LaTeX writer:

+ Avoid inadvertently creating ?` or !` ligatures (#5407).
These are upside down ? and !, resp.
+ Fix footnotes in table caption and cells (#5367). This fixes a
bug wherein footnotes appeared in the wrong order, and with
duplicate numbers, when in table captions and cells.
We now use regular `\footnote` commands, even in the table
caption and the minipages containing cells. Apparently
longtable knows how to handle this.

* HTML writer: Don't add data- prefix to RDFa attributes (#5403).

* JATS writer: Ensure that plain strings go inside `<pub-id>` tag (#5397).

* Markdown writer:

+ Better rendering of numbers (#5398). If the number is integral,
we render it as an integral not a float.
+ Proper rendering of empty map in YAML metadata (#5398). Should
be `{}`, not empty string.
+ Properly escape attributes in Markdown writer (#5369).
+ Be sure implicit figures work in list contexts (#5368).
Previously they would sometimes not work: e.g., when they
occured in final paragraphs in lists that were originally
parsed as Plain and converted later using PlainToPara.

* Docx writer: Use `w:br` without attributes for line breaks (#5377).
We previously added the attribute `type="textWrapping"`, but
this causes problems on Word Online.

* LaTeX template (Andrew Dunning):

+ Ensure correct heading/table order (#5365). Improve workaround
(#1658) for tables following headings. The new solution works whether
or not the `indent` variable is enabled.
+ Remove `subparagraph` variable. The default is now to use run-in
style for level 4 and 5 headings (`\paragraph` and `\subparagraph`).
To get the previous default behavior (where these were formatted as
blocks, like `\subsubsection`), set the `block-headings` variable.
+ Add pandoc to PDF metadata (#5388).
+ Group graphics-related code (#5389).
+ Move `\setstretch` after front matter (#5179). Ensures that
`\maketitle`, `\tableofcontents`, and so forth are not affected by
changes to line spacing.

* Update data/jats.csl to avoid commas between name-part elements

* Add support for golang (`go`) with `--listings` (#5427).

* Text.Pandoc.Shared - improve `metaToJSON` behavior with numbers.
We now do a better job marshalling numbers from MetaString
or MetaInlines into JSON Number.

* Text.Pandoc.Writers.Shared: `metaValueToJSON`: use Number Values for
integers. Pandoc's MetaValue doesn't have a distinguished number type,
so numbers are put in MetaStrings. If the MetaString consists
entirely of digits, we convert it to a Number. We should probably
consider adding a MetaNumber constructor to MetaValue, for better
round-tripping with JSON etc. This change aids round-tripping in
ipynb metadata fields, like `toc_depth`.

* Text.Pandoc.Class: `fetchItem`: don't treat UNC paths as
protocol-relative URLs (#5127). These are paths beginning `//?/UNC/...`.

* Text.Pandoc.ImageSize: Improve `pdfSize` so it handles
a wider range of PDFs (#4322, with help from Richard Davis).

* Text.Pandoc.Pretty: avoid stack overflow by using strict sum (#5401).

* Fix harmless error in file-scope code (#5422).

* MANUAL.txt:

+ Improve 'header' and 'heading' usage (#5423, Andrew
Dunning). The term 'header' was being used where 'heading'
is more appropriate.
+ Add paragraph on options affecting markdown in ipynb.

* stack.yaml - remove -Wmissing-home-modules
This seems to cause problems with stack ghci. Remove RTS options.

* Add ghc-options to cabal.project.

* appveyor.yml - use ghc 8.6.4. Fixes segfault issues on Windows (#5037).

* linux build process: Remove clone of pandoc-citeproc (#5366).
It wasn't being used; cabal.project specifies the version to use.

pandoc (2.7.1)

* Add tectonic as an option for --pdf-engine (#5345, Cormac Relf).
Runs tectonic on STDIN instead of a temporary .tex file, so that it
looks in the working directory for `\include` and `\input` like the rest
of the engines. Allows overriding the output directory
args with `--pdf-engine-opt=--outdir --pdf-engine-opt="$DIR"`.

* Allow `-o/--output` to be used with `--print-default-data-file`,
`--print-highlighting-style`, `--print-default-template`.
Note that `-o` must occur BEFORE the `--print*` command on the command
line (this is documented, #5357).

* LaTeX reader:

+ Support `\underline`, `\ul`, `\uline` (#5359, Paul
Tilley). These are parsed as a Span with class `underline`, as
with other readers.
+ Ensure that `\Footcite` and `\Footcites` get put in a note.

* ipynb reader:

+ Remove sensitivity to `raw_html`, `raw_tex` extensions.
We now include every output format. Pruning is handled by
+ Better handling of cell metadata. We now include even complex cell
metadata in the Div's attributes (as JSON, in complex cases, or
as plain strings in simple cases).

* ipynb writer:

+ Recurse into native divs for output cell data (#5354).
+ Render cell metadata fields from div attributes.

* Docx writer: avoid extra copy of abstractNum and num elements
in numbering.xml. This caused pandoc-produced docx files to
be uneditable using Word Online (#5358).

* Markdown writer: improve handling of raw blocks/inline.
We now emit raw content using `raw_attribute` when no more
direct method is available. Use of `raw_attribute` can be
forced by disabling `raw_html` and `raw_tex`.

* LaTeX writer: Add classes for frontmatter support (#5353, Andrew Dunning)
and remove frontmatter from `scrreprt`.

* LaTeX template:

+ Improve readability (#5363, Andrew Dunning).
+ Robust section numbering removal (#5351, Andrew Dunning).
Ensures that section numbering does not reappear with custom
section levels. See <>.
+ Better handling of front/main/backmatter (#5348).
In pandoc 2.7 we assumed that every class with chapters would accept
`\frontmatter`, `\mainmatter`, and `\backmatter`. This is not so (e.g.
report does not). So pandoc 2.7 breaks on report class by including an
unsupported command. Instead of the `book-class` variable, we use two
variables, `has-chapters` and `has-frontmatter`, and set these
intelligently in the writer.

* Text.Pandoc.Shared: Improve `filterIpynbOutput`. Ensure that
images are prioritized over text. `best` should include everything
for ipynb.

* Tests.Old: specify `--data-dir=../data` to ensure tests can find
data files even if they haven't been installed. Remove old
`pandoc_datadir` environment variable, which hasn't done anything for a
long time.

* MANUAL.txt: Add recommendation to use `raw_attribute` with ipynb (#5354).

* Use cmark-gfm-hs 0.1.8 (note that 0.1.7 is buggy).

* Use latest pandoc-citeproc, texmath.

pandoc (2.7)

* Use XDG data directory for user data directory (#3582). Instead of
`$HOME/.pandoc`, the default user data directory is now
`$XDG_DATA_HOME/pandoc`, where `XDG_DATA_HOME` defaults to
`$HOME/.local/share` but can be overridden by setting the
environment variable. If this directory is missing, then
`$HOME/.pandoc` is searched instead, for backwards compatibility.
However, we recommend moving local pandoc data files from
`$HOME/.pandoc` to `$HOME/.local/share/pandoc`. On Windows the
default user data directory remains the same.

* Slide show formats behavior change: content under headers
less than slide level is no longer ignored, but included in
the title slide (for HTML slide shows) or in a slide after
the title slide (for beamer). This change makes possible
2D reveal.js slideshows with content in the top slide on
each stack (#4317, #5237).

* Add command line option `--ipynb-output=all|none|best` (#5339).
Output cells in ipynb notebooks often contain several different
versions of an output, with different MIME types, e.g. an HTML
table and a plain-text fallback. Specifying `--ipynb-output=best`
(the default) ensures that the best version for the output format
is used. `all` includes all versions, and `none` suppresses them
all, leaving output cells empty.

* `asciidoctor` is now an output format separate from `asciidoc`,
to accommodate some minor implementation-specific differences
(currently just in the treatment of display math).

* Add `latexmk` as an option for `--pdf-engine` (#3195).
Note that you can use `--pdf-engine-opt=-outdir=bar` to specify
a persistent temp directory.

* Markdown reader:

+ Improve tight/loose list handling (#5285). Previously the
algorithm allowed list items with a mix of Para and Plain, which
is never wanted.
+ Add newline when parsing blocks in YAML (#5271). Otherwise last
block gets parsed as a Plain rather than a Para. This is a
regression in pandoc 2.x. This patch restores pandoc 1.19
+ Make `yamlToMeta` respect extensions (#5272, Mauro Bieg).
This adds a `ReaderOptions` parameter to `yamlToMeta` [API change].
+ Fix bug parsing fenced code blocks (#5304). Previously parsing
would break if the code block contained a string of backticks of
sufficient length followed by something other than end of line.

* LaTeX reader: don't let `\egroup` match `{`. `braced` now actually
requires nested braces. Otherwise some legitimate command and
environment definitions can break.

* Docx reader (Jesse Rosenthal):

+ Rename `getDocumentPath` as `getDocumentXmlPath`.
+ Use field notation for setting `ReaderEnv`.
+ Figure out `document.xml` path once at the beginning of parsing, and
add it to the environment, so we can avoid repeated lookups.
+ Dynamically determine main document xml path (#5277).
The desktop Word program places the main document file in
`word/document.xml`, but the online word places it in
`word/document2.xml`. This file path is actually stated in the root
`_rels/.rels` file, in the `Relationship` element with an
`http://../officedocument` type.
+ Fix paths in archive to prevent Windows failure (#5277).
Some paths in archives are absolute (have an opening slash) which, for
reasons unknown, produces a failure in the test suite on MS
Windows. This fixes that by removing the leading slash if it exists.
+ Add comments to aid code readability.
+ Trim space inside the last inline (#5273).
+ Unwrap sdt elements in footnotes and comments (#5302).

* Muse reader (Alexander Krotov):

+ Test that block level markup does not break `<verbatim>`.
+ Add secondary note support.

* ipynb reader: handle images referring to attachments. Previously
we didn't strip off the attachment: prefix, so even though the
attachment was available in the mediabag, pandoc couldn't find it.

* JATS reader:

+ Fix parsing of figures (#5321). This ensures that a figure
containing a single image is parsed as a pandoc "implicit
figure" (i.e., a Para with a single Image whose title attribute
begins with `fig:`). More complex figures will still be parsed
as divs.
+ Support `fig-group` block element (#5317).
+ Handle citations with multiple references (#5310). The `rid`
attribute can have a space-separated list of ids.

* AsciiDoc Writer: Add `writeAsciiDoctor` [API change, Tarik Graba].
Handle display math appropriately for Asciidoctor.

* JATS writer: wrap figure caption in `<p>` to fix validation (#5290,
Mauro Bieg).

* HTML writer:

+ Implement WAI-ARIA roles for (end)notes, citations, and
bibliography (#4213). Note that doc-biblioref is only used when
link-citations produces links, since it belongs on links.
+ Include content (including speaker notes) in title slides
(#4317, #5237).

* ipynb writer:

+ Ensure final newline.
+ Only include metadata under `jupyter` field.
+ Don't create attachments for images with absolute URIs,
including data: URIs (#5303).
+ Keep plain text fallbacks in output even if a richer format is
included (#5293). We don't know what output format will be needed.
See the `--ipynb-output` command line option for a way to control
what formats are included in the output.

* Markdown writer: use `markdown="1"` when appropriate for Divs:
when `native_divs` and `markdown_in_html_blocks` are disabled
but `raw_html` and `markdown_attribute` are enabled.

* LaTeX writer:

+ Use right fold for `escapeString`. This is more elegant than
the explicit recursive code we were using.
+ Avoid `{}` after control sequences when escaping.
`\ldots{}.` doesn't behave as well as `\ldots.` with the latex
ellipsis package. This patch causes pandoc to avoid emitting
the `{}` when it is not necessary. Now `\ldots` and other
control sequences used in escaping will be followed by either
a `{}`, a space, or nothing, depending on context.
+ For beamer, include contents under headers superordinate to
slidelevel (#4317). Currently we keep the fancy title slide, and
add a new slide with the same title and whatever content was
under the header.

* Powerpoint writer (Jesse Rosenthal): support underlines.
Use span with single class "underline" as in docx writer.

* Muse writer: escape secondary notes (Alexander Krotov).

* FB2 writer: add section identifiers support (#5229, John KetzerX).

* Make `--fail-if-warnings` work for PDF output (#5343).

* Lua filters (Albert Krewinkel):

+ Load module `pandoc` before calling `init.lua` (#5287). The file
`init.lua` in pandoc's data directory is run as part of pandoc's
Lua initialization process. Previously, the `pandoc` module was
loaded in `init.lua`, and the structure for marshaling was
set up after. This allowed simple patching of element
marshaling, but made using `init.lua` more difficult. Now, all
required modules are now loaded before calling `init.lua`. The
file can be used entirely for user customization. Patching
marshaling functions, while discouraged, is still possible via
the `debug` module.
+ All Lua modules bundled with pandoc, i.e., `pandoc.List`,
`pandoc.mediabag`, `pandoc.utils`, and `text` are re-exported from the
`pandoc` module. They are assigned to the fields `List`, `mediabag`,
`utils`, and `text`, respectively.

* Text.Pandoc.Lua (Albert Krewinkel):

+ Split `StackInstances` into smaller Marshaling modules.
+ Get `CommonState` from Lua global. This allows more control over
the common state from within Lua scripts.

* LaTeX template:

+ Support the `subject` metadata variable (#5289, Pascal Wagler).
+ Add `\frontmatter`, `\mainmatter`, `\backmatter`
for book classes (#5306).

* epub3 template: Add titlepage class to section (#5269).

* HTML5 template: Add ARIA role `doc-toc` for table of contents (#4213).

* Make `--metadata-file` use selected extensions (#5279, #5272, Mauro Bieg).

* Text.Pandoc.Shared:

+ Remove `withTempDir` [API change].
+ Add new exported function `defaultUserDataDirs` [API change].
+ Add `filterIpynbOutput` [API change].
+ `compactify`: Avoid lists with a mix of Plain and Para elements

* Text.Pandoc.Translations: reorder alphabetically and remove `Author`
(#5334, Mauro Bieg).

* Text.Pandoc.Extensions:

+ More carefully groom ipynb default extensions.
+ Add `all_symbols_escapable` to `githubMarkdownExtensions`.

* Text.Pandoc.PDF:

+ Use system temp directory when possible (#1192). Previously we
created temp dirs in the working directory, partly (a) because
there were problems using the system temp directory on Windows,
when their pathnames included tildes, and partly (b) because
programs like `` would not be allowed to write to
directories outside the working directory in restricted mode. We
now (a) use the system temp dir except when the path includes
tildes, and (b) set `TEXMFOUTPUT` when creating the PDF, so that
subsidiary programs can use the system temp directory. This
addresses problems that occurred when pandoc was used in a
synced directory (such as Dropbox).
+ Change types of subsidiary functions to PandocIO, to allow
warnings to be threaded through (#5343).

* Text.Pandoc.MIME: add WebP (#5267, Mauro Bieg).

* Tests: avoid calling `findPandoc` multiple times.

* Old tests: remove need for temp files by using `pipeProcess`.

* Added simple ipynb reader/writer tests (#5274).

* Rearrange `--help` output in a more rational way, with common
options at the beginning and options grouped by function (#5336).

* trypandoc: Add JATS and other missing formats (Arfon Smith, #5291).

* Add missing copyright notices and remove license boilerplate (#4592,
Albert Krewinkel).

* Use latest basement/foundation on 32bit windows.

* Use latest skylighting (#5328). Custom syntax definitions no
longer try to load `language.dtd`.

* Require texmath

* Use latest pandoc-citeproc (

* MANUAL.txt:

+ Clarify variable substitution indentation in templates (#5338,
Agustín Martín Barbero).
+ Reorder custom-styles section (#5324, Mauro Bieg).

pandoc (2.6)

* Support ipynb (Jupyter notebook) as input and output format.

+ Add `ipynb` as input and output format (extension `.ipynb`).
+ Added Text.Pandoc.Readers.Ipynb [API change].
+ Added Text.Pandoc.Writers.Ipynb [API change].
+ Add `PandocIpynbDecodingError` constructor to Text.Pandoc.Error.Error
[API change].
+ Depend on ipynb library.
+ Note: there is no template for ipynb.

* Add DokuWiki reader (#1792, Alexander Krotov). This adds
Text.Pandoc.Readers.DokuWiki [API change], and adds `dokuwiki`
as an input format.

* Implement task lists (#3051, Mauro Bieg). Added `task_lists`
extension. Task lists are supported from markdown and gfm input.
They should work, to some degree, in all output formats, though in
most formats you'll get a bullet list with a unicode character for
the box. In HTML, you get checkboxes and in LaTeX/PDF output, a
box is used as the list marker. API changes:

+ Added constructor `Ext_task_lists` to `Extension`.
+ Added `taskListItemFromAscii` and `taskListItemToAscii` to

* Allow some command line options to take URL in addition to FILE.
`--include-in-header`, `--include-before-body`, `--include-after-body`.

* HTML reader:

+ Handle empty `start` attribute (see #5162).
+ Treat `textarea` as a verbatim environment (#5241) and preserve

* RST reader:

+ Change treatment of `number-lines` directive (Brian Leung,
#5207). Directives of this type without numeric inputs should
not have a `startFrom` attribute; with a blank value, the
writers can produce extra whitespace.
+ Removed superfluous `sourceCode` class on code blocks (#5047).
+ Handle `sourcecode` directive as synonynm for `code` (#5204).

* Markdown reader:

+ Remove `sourceCode` class for literate Haskell code blocks
(#5047). Reverse order of `literate` and `haskell` classes on
code blocks when parsing literate Haskell, so `haskell` is
+ Treat `<textarea>` as a verbatim environment (#5241).

* Org reader:

+ Handle `minlevel` option differently (#5190, Brian Leung).
When `minlevel` exceeds the original minimum level observed in
the file to be included, every heading should be shifted
+ Allow for case of `:minlevel == 0` (#5190).
+ Fix treatment of links to images (#5191, Albert Krewinkel).
Links with descriptions which are pointing to images are no
longer parsed as inline images, but as links.
+ Add support for #+SELECT_TAGS (Brian Leung).
+ Separate filtering logic from conversion function (Brian Leung).

* TWiki reader: Fix performance issue with underscores (#3921).

* MediaWiki reader: use `_` instead of `-` in auto-identifiers (#4731).
We may not still be exactly matching mediawiki's algorithm.

* LaTeX reader:

+ Remove `sourceCode` class for literate Haskell code blocks
(#5047). Reverse order of `literate` and `haskell` classes on
code blocks when parsing literate Haskell, so `haskell` is
+ Support `\DeclareMathOperator` (#5149).
+ Support `\inputminted` (#5103).
+ Support `\endinput` (#5233).
+ Allow includes with dots like `cc_by_4.0`. Previously the `.0`
was interpreted as a file extension, leading pandoc not to add
`.tex` (and thus not to find the file). The new behavior matches
tex more closely.

* Man reader:

+ Use `mapLeft` from Shared instead of defining own.

* Docx reader (Jesse Rosenthal):

+ Handle level overrides (#5134).

* Docx writer:

+ Support custom properties (#3024, #5252, Agustín Martín Barbero).
Also supports additional core properties: `subject`, `lang`,
`category`, `description`.
+ Make Level into a real type, instead of an alias for a tuple
(Jesse Rosenthal).

* ICML writer (Mauro Bieg):

+ Support custom-styles (#5137, see #2106).
+ Support unnumbered headers (#5140).

* Texinfo writer: Use header identifier for anchor if present (#4731).
Previously we were overwriting an existing identifier with a new one.

* Org writer: Preserve line-numbering for example and code blocks
(Brian Leung).

* Man/Ms writers: Don't escape `-` as `\-`. The `\-` gets rendered
in HTML and PDF as a unicode minus sign.

* Ms writer: Ensure we have a newline after .EN in disply math (#5251).

* RST writer: Don't wrap simple table header lines (#5128).

* Asciidoc writer: Shorter delimiters for tables, blockquotes
(#4364). This matches asciidoctor reference docs.

* Dokuwiki writer: Remove automatic `:` prefix before internal image
links (#5183, Damien Clochard). This prevented users from making
relative image links.

* Zimwiki writer: remove automatic colon prefix before internal
images (#5183, Damien Clochard).

* MediaWiki writer: fix caption, use 'thumb' instead of 'frame'
(#5105). Captions used to have the word 'caption' prepended; this
has been removed. Also, 'thumb' is used instead of 'frame' to
allow images to be resized.

* reveal.js writer:

+ Ensure that we don't get > 2 levels of section nesting,
even with slide level > 2 (#5168).
+ If slide level == N but there is no N-level header, make
sure the next header with level > N gets treated as a slide
and put in a section, rather than remaining loose (#5168).

* Markdown writer:

+ Make `plain` RawBlocks pass through in `plain` output.
+ Include needed whitespace after HTML figure (#5121).
We use HTML for a figure in markdown dialects that can't
represent it natively.

* Commonmark writer:

+ Fix handling of SoftBreak with `hard_line_breaks` (#5195).
+ Implement `--toc` (`writerTableOfContents`)
in commonmark/gfm writers (#5172).

* EPUB writer:

+ Ensure that picture transforms are done on metadata too.
+ Small fixes to `nav.xhtml`:
Add 'landmarks' id attribute to the landmarks nav.
Replace old default CSS removing numbers from ol.toc li
with new rules that match `nav#toc ol, nav#landmarks ol`.
We keep the `toc` class on `ol` for backwards compatibility.

* LaTeX writer:

+ Make raw content marked `beamer` pass through in `beamer`
output (pandoc/lua-filters#40).
+ Beamer: avoid duplicated `fragile` property in some cases
+ Add `#` special characters for listings (#4939).
This character needs special handling in `\lstinline`.

* RTF writer: use `toTableOfContents` from Shared to replace
old duplicated code.

* Pptx writer:

+ Support custom properties. Also supports additional core
properties: `subject`, `category`, `description` (#5252,
Agustín Martín Barbero).
+ Use `toTableOfContents` from Shared to replace old duplicated code.

* ODT writer (Augustín Martín Barbero):

+ Fix typo in custom properties (#2839).
+ Improve standard properties, including the following core
properties: `generator` (Pandoc/VERSION), `description`, `subject`,
`keywords`, `initial-creator` (from authors), `creation-date`
(actual creation date) (#5252).

* Custom writers:

+ Allow '-' in filenames for custom lua writers (#5187).
+ sample.lua: add `SingleQuoted`, `DoubleQuoted` (#5104).
+ sample.lua: Add a missing `>` (MichaWiedenmann).

* reveal.js template: Add `zoomKey` config (#4249).

* HTML5 template: Remove unnecessary type="text/css" on style and
link for HTML5 (#5146).

* LaTeX template (Andrew Dunning, except where noted):

+ Prevent fontspec from scaling `mainfont` to match the default
font, Latin Modern. A main font set to 12pt could
previously appear between 11pt to 13pt depending on its design.
To return to the earlier rendering, use `-V
mainfontoptions="Scale=MatchLowercase"` (#5212, #5218).
+ Display monospaced fonts without TeX ligatures when using
`--pdf-engine=lualatex`. It now matches the behaviour of other
engines (#5212, #5218).
+ Remove the deprecated `romanfont` variable. The functionality of
`mainfont` is identical (#5218).
+ Render `\subtitle` with the standard document classes.
Previously, `subtitle` only appeared when using the KOMA-Script
classes or Beamer (#5213, #5244).
+ Use Babel instead of Polyglossia for LuaLaTeX. This avoids
several language selection problems, notably with retaining
French spacing conventions when switching to a verbatim
environment or another language; and in printing Greek text
without hyphenation (#5193).
+ Use the [`xurl`]( package if
available, improving the appearance of URLs by allowing them to
break at additional points (#5193).
+ Use [`bookmark`]( if available
to correct heading levels in PDF bookmarks: see the [KOMA-Script
3.26 release notes]( (#5193).
+ Require the [`xcolor`]( package to
avoid a possible error when using additional packages alongside
footnotes in tables (#5193, closes #4861).
+ Remove obsolete `fixltx2e` package, which has no functionality
with TeX Live 2015 or later (#5193).
+ Allow multiple `fontfamilies.options` (#5193, closes #5194).
+ Restrict `institute` variable to Beamer (#5219).
+ Use [`footnotehyper`](
package if available to make footnotes in tables compatible
with `hyperref` (#5234).
+ Number parts and chapters in book classes only if the
`numbersections` variable is set, for consistency with other
output formats. To return to the previous behaviour, use
`-V numbersections -V secnumdepth=0` (#5235).
+ Reindent file (#5193).
+ Use built-in parskip handling with KOMA-Script classes (#5143, Enno).
+ Set default listings language for lua, assembler (#5227, John
MacFarlane). Otherwise we get an error when trying to compile code
with lua or assembler code. To change the default dialect
(currenty 5.3 for lua and x86masm for assembler), you can use
`--include-in-header` to inject something like

* Text.Pandoc.Readers: Changed types of `readJSON`; it now runs
in an instance of PandocMonad, like the other readers and
writers. [API change]

* Text.Pandoc.Writers: Changed types of `writeJSON`; it now runs
in an instance of PandocMonad, like the other readers and
writers. [API change]

* Text.Pandoc.Error: Added `PandocUTF8DecodingError` constructor
for `PandocError`. [API change]

* Text.Pandoc.Writers.Shared - add `toTableOfContents`. [API change]
This is refactored out from the Markdown writer.
It can be used in other writers to create a table of contents.

* Improve error messages for UTF-8 decoding errors. Now we give
the filename and byte offset (#4765).

* Text.Pandoc.XML: Strip out illegal XML characters in
`escapeXMLString` (#5119).

* Text.Pandoc.Process: update `pipeProcess` (Albert Krewinkel).
The implementation of `pipeProcess` was rewritten to fix sporadic
failures caused by prematurely closed pipes.

* Use `safeRead` instead of `read` everywhere in the code
(John MacFarlane, Mauro Bieg, #5162, #5180).

* Text.Pandoc.SelfContained: Decompress `.svgz` when
converting to `data:` URI (#5163, Alexander Krotov).

* Text.Pandoc.Parsing: Remove unused `HasHeaderMap`
(#5175, Alexander Krotov). [API change]

* Normalize Windows paths to account for change in ghc 8.6 (#5127).
When pandoc is compiled with ghc 8.6, Windows paths are treated
differently, and paths beginning `\\server` no longer work.
This commit rewrites such patsh to `\\?\UNC\server` which works.
The change operates at the level of argument parsing, so it
only affects the command line program.

* Simplify/fix reading of `--metadata` values on command line
to avoid problems relating to hvr/HsYAML#7 (#5177).

* data/pandoc.lua: auto-fix nested constructor arguments (Albert
Krewinkel). Incorrect types to pandoc element constructors are
automatically converted to the correct types when possible. This was
already done for most constructors, but conversions are now also done
for nested types (like lists of lists).

* Removed custom Setup.hs, use build-type: simple. The only thing we
gained from the custom build was automatic installation of the man
page when using 'cabal install'. But custom builds cause problems,
e.g., with cross-compilation. Installation of the man page is
better handled by packagers. Note to packagers (e.g. Debian): it
may be necessary to add a step installing the man page with the
next release.

* Allow latest http-client, tasty, zip-archive, Glob.

* Require skylighting >= 0.7.5, adding support for sml, J,

* Tests: Cleaned up `findPandoc` in `Tests.Helpers`, so it
works well with cabal v2.


+ Use button for installer links (John MacFarlane, Mauro Bieg,
+ Fix links and bump required stack version (max).
+ Improve installation notes on associated software (Andrew
Dunning). Includes explanation of how to install related tools
with package managers (since the method of installing
`rsvg-convert` is not obvious).

* doc/ improve documentation of org features (Albert

* doc/ use 3rd level headers for module fields.


+ Clarify automatic identifiers (#5201). We remove
non-alphanumerics. This includes, e.g., emojis.
+ Fix example for Div with id (Geoffrey Ely).
+ Update list of LaTeX packages used.
+ Make it clear that `hard_line_breaks` works in `gfm` (see #5195).
+ Mention `raw_attribute` in documentation for `raw_html` and
`raw_tex` (#5240, thanks to @eiro).
+ Clarify that `$sep$` must come right before `$endfor$` in templates
(#5243, Lev Givon).
+ Document metadata support for docx, odt, pptx writers (Agustín
Martín Barbero).
+ Reorganize template variables (#5249, Andrew Dunning).
Add additional headings to categorize variables, and
alphabetize when there is large number; add more examples.
+ Document `date-meta` template variable (#5260, Tristan Stenner).

* trypandoc: Fix CSS and viewport.

pandoc (2.5)

* Text.Pandoc.App: split into several unexported submodules (Albert
Krewinkel): Text.Pandoc.App.FormatHeuristics,
Text.Pandoc.App.Opt, Text.Pandoc.App.CommandLineOptions,
Text.Pandoc.App.OutputSettings. This is motivated partly by the
desire to reduce recompilations when something is modified,
since App previously depended on virtually every other module.

* Text.Pandoc.Extensions

+ Semantically, `gfm_auto_identifiers` is now a modifier of
`auto_identifiers`; for identifiers to be set, `auto_identifiers` must
be turned on, and then the type of identifier produced depends on
`gfm_auto_identifiers` and `ascii_identifiers` are set. Accordingly,
`auto_identifiers` is now added to `githubMarkdownExtensions` (#5057).
+ Remove `ascii_identifiers` from `githubMarkdownExtensions`.
GitHub doesn't seem to strip non-ascii characters any more.

* Text.Pandoc.Lua.Module.Utils (Albert Krewinkel)

+ Test AST object equality via Haskell (#5092). Equality of Lua
objects representing pandoc AST elements is tested by unmarshalling
the objects and comparing the result in Haskell.
A new function `equals` which performs this test has been added to the
`pandoc.utils` module.
+ Improve stringify. Meta value strings (MetaString)
and booleans (MetaBool) are now converted to the literal string and the
lowercase boolean name, respectively. Previously, all values of these
types were converted to the empty string.

* Text.Pandoc.Parsing: Remove Functor and Applicative constraints where Monad
already exists (Alexander Krotov).

* Text.Pandoc.Pretty: Don't render BreakingSpace at end of line
or beginning of line (#5050).

* Text.Pandoc.Readers.Markdown

+ Fix parsing of citations, quotes, and underline emphasis
after symbols. Starting with pandoc 2.4, citations, quoted inlines,
and underline emphasis were no longer recognized after certain
symbols, like parentheses (#5099, #5053).
+ In pandoc 2.4, a soft break after an abbreviation would be
relocated before it to allow for insertion of a nonbreaking
space after the abbreviation. This behavior is here reverted.
A soft break after an abbreviation will remain, and no nonbreaking
space will be added. Those who care about this issue should take care not
to end lines with an abbreviation, or to insert nonbreaking spaces

* Text.Pandoc.Readers.FB2: Do not throw error for unknown elements in
`<body>` (Alexander Krotov). Some libraries include custom elements
in their FB2 files.

* Text.Pandoc.Readers.HTML

+ Allow `tfoot` before body rows (#5079).
+ Parse `<small>` as a Span with class "small" (#5080).
+ Allow thead containing a row with `td` rather than `th` (#5014).

* Text.Pandoc.Readers.LaTeX

+ Cleaned up handling of dimension arguments. Allow decimal points,
preceding space.
+ Don't allow arguments for verbatim, etc.
+ Allow space before bracketed options.
+ Allow optional arguments after `\\` in tables.
+ Improve parsing of `\tiny`, `\scriptsize`, etc. Parse as raw,
but know that these font changing commands take no arguments.

* Text.Pandoc.Readers.Muse

+ Trim whitespace before parsing grid table cells (Alexander Krotov).
+ Add grid tables support (Alexander Krotov).

* Text.Pandoc.Shared

+ For bibliography match Div with id `refs`, not class `references`.
This was a mismatch between pandoc's docx, epub, latex, and markdown
writers and the behavior of pandoc-citeproc, which actually looks for a
div with id `refs` rather than one with class `references`.
+ Exactly match GitHub's identifier generating algorithm (#5057).
+ Add parameter for `Extensions` to `uniqueIdent` and
`inlineListToIdentifier` (#5057). [API change]
This allows these functions to be sensitive to the settings of
`Ext_gfm_auto_identifiers` and `Ext_ascii_identifiers`, and allows us to
use `uniqueIdent` in the CommonMark reader, replacing custom code. It
also means that `gfm_auto_identifiers` can now be used in all formats.

* Text.Pandoc.Writers.AsciiDoc

+ Use `.`+ as list markers to support nested ordered lists (#5087).
+ Support list number styles (#5089).
+ Render Spans using `[#id .class]#contents#` (#5080).

* Text.Pandoc.Writers.CommonMark

+ Respect `--ascii` (#5043, quasicomputational).
+ Make sure `--ascii` affects quotes, super/subscript.

* Text.Pandoc.Writers.Docx

+ Fix bookmarks to headers with long titles (#5091).
Word has a 40 character limit for bookmark names. In addition, bookmarks
must begin with a letter. Since pandoc's auto-generated identifiers may
not respect these constraints, some internal links did not work. With
this change, pandoc uses a bookmark name based on the SHA1 hash of the
identifier when the identifier isn't a legal bookmark name.
+ Add bookmarks to code blocks (Nikolay Yakimov).
+ Add bookmarks to images (Nikolay Yakimov).
+ Refactor common bookmark creation code into a function (Nikolay Yakimov).

* Text.Pandoc.Writers.EPUB: Handle calibre metadata (#5098).
Nodes of the form

<meta name="calibre:series" content="Classics on War and Politics"/>

are now included from an epub XML metadata file. You can also include
this information in your YAML metadata, like so:

series: Classics on War and Policitics

In addition, ibooks-specific metadata can now be included via an XML
file. (Previously, it could only be included via YAML metadata, see

* Text.Pandoc.Writers.HTML: Use plain `"` instead of `&quot;` outside of

* Text.Pandoc.Writers.ICML: Consolidate adjacent strings, inc. spaces.
This avoids splitting up the output unnecessarily into separate elements.

* Text.Pandoc.Writers.LaTeX: Don't emit `[<+->]` unless beamer output, even
if `writerIncremental` is True (#5072).

* Text.Pandoc.Writers.Muse (Alexander Krotov).

+ Output tables as grid tables if they have multi-line cells.
+ Indent simple tables only on the top level.
+ Output tables with one column as grid tables.
+ Add support for `--reference-location`.
+ Internal improvements.

* Text.Pandoc.Writers.OpenDocument: Fix list indentation (Nils Carlson,
#5095). This was a regression in pandoc 2.4.

* Text.Pandoc.Writers.RTF: Fix warnings for skipped raw inlines.

* Text.Pandoc.Writers.Texinfo: Add blank line before `@menu` section (#5055).

* Text.Pandoc.XML: in `toHtml5Entities`, prefer shorter entities
when there are several choices for a particular character.

* data/abbreviations

+ Add additional abbreviations (Andrew Dunning)
Many of these borrowed from the Chicago Manual of Style 10.42,
'Scholarly abbreviations'.

* Templates

+ Asciidoc template: add :lang: to title header is lang is set in
metadata (#5088).

* pandoc.cabal: Add cabal flag `derive_json_via_th` (Albert Krewinkel)
Disabling the flag will cause derivation of ToJSON and FromJSON
instances via GHC Generics instead of Template Haskell. The flag is
enabled by default, as deriving via Generics can be slow (see #4083).

* trypandoc:

+ Tweaked drop-down lists.
+ Put link to site in footer.
+ Preselect output format.
+ Update on change of in or out format.
+ Add man input format.

* MANUAL.txt:

+ Fix outdated description of latex_macros extension.
+ Clarified placement of bibliography.
+ Added "A note on security."
+ Fix note on curly brace syntx for locators.
+ Document new explicit syntax for citeproc locators.
+ Remove confusing cross-links for some extensions.
+ Don't put pandoc in code ticks in heading.
+ Document that `--ascii` works for gfm and commonmark too.
+ Add `man` to `--from` options.

* doc/ various improvements (Mauro Bieg).

pandoc (2.4)

[new features]

* New input format `man` (Yan Pashkovsky, John MacFarlane).

[behavior changes]

* `--ascii` is now implemented in the writers, not in Text.Pandoc.App,
via the new `writerPreferAscii` field in `WriterOptions`.
Now the `write*` functions for Docbook, HTML, ICML, JATS, LaTeX,
Ms, Markdown, and OPML are sensitive to `writerPreferAscii`.
Previously the to-ascii translation was done in Text.Pandoc.App,
and thus not available to those using the writer functions

* `--ascii` now works with Markdown output. HTML5 character reference
entities are used.

* `--ascii` now works with LaTeX output. 100% ASCII output can't be
guaranteed, but the writer will use commands like `\"{a}` and `\l`
whenever possible, to avoid emiting a non-ASCII character.

* For HTML5 output, `--ascii` now uses HTML5 character reference
entities rather than numerical entities.

* Improved detection of format based on extension (in Text.Pandoc.App).
We now ensure that if someone tries to convert a file for a
format that has a pandoc writer but not a reader, it won't just
default to markdown.

* Add viz. to abbreviations file (#5007, Nick Fleisher).

* AsciiDoc writer: always use single-line section headers,
instead of the old underline style (#5038). Previously the single-line
style would be used if `--atx-headers` was specified, but
now it is always used.

* RST writer: Use simple tables when possible (#4750).

* CommonMark (and gfm) writer: Add plain text fallbacks. (#4528,
quasicomputational). Previously, the writer would unconditionally
emit HTML output for subscripts, superscripts, strikeouts (if the
strikeout extension is disabled) and small caps, even with
`raw_html` disabled. Now there are plain-text (and, where
possible, fancy Unicode) fallbacks for all of these corresponding
(mostly) to the Markdown fallbacks, and the HTML output is only
used when `raw_html` is enabled.

* Powerpoint writer: support raw openxml (Jesse Rosenthal, #4976).
This allows raw openxml blocks and inlines to be used in the pptx
writer. Caveats: (1) It's up to the user to write
well-formed openxml. The chances for corruption, especially with
such a brittle format as pptx, is high. (2) Because of
the tricky way that blocks map onto shapes, if you are using
a raw block, it should be the only block on a slide
(otherwise other text might end up overlapping it). (3) The
pptx ooxml namespace abbreviations are different from the
docx ooxml namespaces. Again, it's up to the user to get it
right. Unzipped document and ooxml specification should be

* With `--katex` in HTML formats, do not use the autorenderer (#4946).
We no longer surround formulas with `\(..\)` or `\[..\]`. Instead,
we tell katex to convert the contents of span elements with
class "math". Since math has already been identified, this
avoids wasted time parsing for LaTeX delimiters. Note, however,
that this may yield unexpected results if you have span elements
with class "math" that don't contain LaTeX math.
Also, use latest version of KaTeX by default (0.9.0).

* The man writer now produces ASCII-only output, using groff escapes,
for portability.

* ODT writer:

+ Add title, author and date to metadata; any remaining
metadata fields are added as `meta:user-defined` tags.
+ Implement table caption numbering (#4949, Nils Carlson).
Captioned tables are numbered and labeled with format "Table 1:
caption", where "Table" is replaced by a translation, depending
on the value of `lang` in metadata. Uncaptioned tables are not
+ OpenDocument writer: Implement figure numbering in captions (#4944,
Nils Carlson). Figure captions are now numbered 1, 2, 3, ...
The format in the caption is "Figure 1: caption" and so on
(where "Figure" is replaced by a translation, depending on the
value of `lang` in the metadata). Captioned figures are numbered
consecutively and uncaptioned figures are not enumerated. This
is necessary in order for LibreOffice to generate an
Illustration Index (Table of Figures) for included figures.

* RST reader: Pass through fields in unknown directives as div attributes
(#4715). Support `class` and `name` attributes for all directives.

* Org reader: Add partial support for `#+EXCLUDE_TAGS` option. (#4284,
Brian Leung). Headers with the corresponding tags should not
appear in the output.

* Log warnings about missing title attributes now include a
suggestion about how to fix the problem (#4909).

* Lua filter changes (Albert Krewinkel):

+ Report traceback when an error occurs. A proper Lua traceback is
added if either loading of a file or execution of a filter
function fails. This should be of help to authors of Lua filters
who need to debug their code.

+ Allow access to pandoc state (#5015). Lua filters and custom
writers now have read-only access to most fields of pandoc's
internal state via the global variable `PANDOC_STATE`.

+ Push ListAttributes via constructor (Albert Krewinkel).
This ensures that ListAttributes, as present in OrderedList elements,
have additional accessors (viz. `start`, `style`, and `delimiter`).

+ Rename ReaderOptions fields, use snake_case. Snake case is used
in most variable names, using camelCase for these fields was an
oversight. A metatable is added to ensure that the old field
names remain functional.

+ Iterate over AST element fields when using `pairs`. This makes
it possible to iterate over all ield names of an AST element by
using a generic `for` loop with pairs`:

for field_name, field_content in pairs(element) do

Raw table fields of AST elements should be considered an
implementation detail and might change in the future. Accessing
element properties should always happen through the fields
listed in the Lua filter docs.

Note that the iterator currently excludes the `t`/`tag` field.

+ Ensure that MetaList elements behave like Lists. Methods usable
on Lists can also be used on MetaList objects.

+ Fix MetaList constructor (Albert Krewinkel). Passing a MetaList
object to the constructor `pandoc.MetaList` now returns the
passed list as a MetaList. This is consistent with the
constructor behavior when passed an (untagged) list.

* Custom writers: Custom writers have access to the global variable
`PANDOC_DOCUMENT`(Albert Krewinkel, #4957). The variable contains
a userdata wrapper around the full pandoc AST and exposes two
fields, `meta` and `blocks`. The field content is only
marshaled on-demand, performance of scripts not accessing the
fields remains unaffected.

[API changes]

* Text.Pandoc.Options: add `writerPreferAscii` to `WriterOptions`.

* Text.Pandoc.Shared:

+ Export `splitSentences`. This was previously duplicated in the Man and
Ms writers.
+ Add `ToString` typeclass (Alexander Krotov).

* New exported module Text.Pandoc.Filter (Albert Krewinkel).

* Text.Pandoc.Parsing

+ Generalize `gridTableWith` to any `Char` Stream (Alexander Krotov).
+ Generalize `readWithM` from `[Char]` to any `Char` Stream
that is a `ToString` instance (Alexander Krotov).

* New exposed module Text.Pandoc.Filter (Albert Krewinkel).

* Text.Pandoc.XML: add `toHtml5Entities`.

* New exported module Text.Pandoc.Readers.Man (Yan Pashkovsky, John

* Text.Pandoc.Writers.Shared

+ Add exported functions `toSuperscript` and
`toSubscript` (quasicomputational, #4528).
+ Remove exported functions `metaValueToInlines`,
`metaValueToString`. Add new exported functions
`lookupMetaBool`, `lookupMetaBlocks`, `lookupMetaInlines`,
`lookupMetaString`. Use these whenever possible for uniformity
in writers (Mauro Bieg, #4907). (Note that
removed function `metaValueToInlines` was in previous
released versions.)
+ Add `metaValueToString`.

* Text.Pandoc.Lua

+ Expose more useful internals (Albert Krewinkel):

- `runFilterFile` to run a Lua filter from file;
- data type `Global` and its constructors; and
- `setGlobals` to add globals to a Lua environment.

This module also contains `Pushable` and `Peekable` instances
required to get pandoc's data types to and from Lua. Low-level
Lua operation remain hidden in Text.Pandoc.Lua.

+ Rename `runPandocLua` to `runLua` (Albert Krewinkel).

+ Remove `runLuaFilter`, merging this into Text.Pandoc.Filter.Lua's
`apply` (Albert Krewinkel).

[bug fixes and under-the-hood improvements]

* Text.Pandoc.Parsing

+ Make `uri` accept any stream with Char tokens (Alexander Krotov).
+ Rewrite `uri` without `withRaw` (Alexander Krotov).
+ Generalize `parseFromString` and `parseFromString'` to any
streams with Char token (Alexander Krotov)
+ Rewrite `nonspaceChar` using `noneOf` (Alexander Krotov)

* Text.Pandoc.Shared: Reimplement `mapLeft` using `Bifunctor.first`
(Alexander Krotov).

* Text.Pandoc.Pretty: Simplify `Text.Pandoc.Pretty.offset`
(Alexander Krotov).

* Text.Pandoc.App

+ Work around HXT limitation for --syntax-definition with windows
drive (#4836).
+ Always preserve tabs for man format. We need it for tables.
+ Split command line parsing code into a separate unexported
module, Text.Pandoc.App.CommandLineOptions (Albert Krewinkel).

* Text.Pandoc.Readers.Roff: new unexported module for tokenizing
roff documents.

* New unexported module Text.Pandoc.RoffChar, provided character
escape tables for roff formats.

* Text.Pandoc.Readers.HTML: Fix `htmlTag` and `isInlineTag` to
accept processing instructions (#3123, regression since 2.0).

* Text.Pandoc.Readers.JATS: Use `foldl'` instead of `maximum` to
account for empty lists (Alexander Krotov).

* Text.Pandoc.Readers.RST: Don't allow single-dash separator in
headerless table (#4382).

* Text.Pandoc.Readers.Org: Parse empty argument array in inline src
blocks (Brian Leung).

* Text.Pandoc.Readers.Vimwiki:

+ Get rid of `F`, `runF` and `stateMeta'` in favor of `stateMeta`
(Alexander Krotov).
+ Parse `Text` without converting to `[Char]` (Alexander Krotov).

* Text.Pandoc.Readers.Creole: Parse `Text` without converting to
`[Char]` (Alexander Krotov).

* Text.Pandoc.Readers.LaTeX

+ Allow space at end of math after `\` (#5010).
+ Add support for `nolinkurl` command (#4992, Brian Leung).
+ Simplified type on `doMacros'`.
+ Tokenize before pulling tokens, rather than after (#4408). This
has some performance penalty but is more reliable.
+ Make macroDef polymorphic and allow in inline context.
Otherwise we can't parse something like `\lowercase{\def\x{Foo}}`.
I have actually seen tex like this in the wild.
+ Improved parsing of `\def`, `\let`. We now correctly parse:
\fooi +\fooii

\fooi +\fooii
+ Improve parsing of `\def` argspec.
+ Skip `\PackageError` commands (see #4408).
+ Fix bugs omitting raw tex (#4527). The default is `-raw_tex`,
so no raw tex should result unless we explicitly say `+raw_tex`.
Previously some raw commands did make it through.
+ Moved `isArgTok` to Text.Pandoc.Readers.LaTeX.Parsing.
+ Moved `babelLangToBCP`, `polyglossiaLangToBCP` to new module,
Text.Pandoc.Readers.LaTeX.Lang (unexported).
+ Simplified accent code using unicode-transforms.
New dependency on unicode-transforms package for normalization.
+ Allow verbatim blocks ending with blank lines (#4624).
+ Support `breq` math environments: `dmath`, `dgroup`, `darray`.
This collects some of the general-purpose code from the LaTeX reader,
with the aim of making the module smaller.

* Text.Pandoc.Readers.Markdown

+ Fix awkward soft break movements before abbreviations (#4635).
+ Add updateStrPos in a couple places where needed.

* Text.Pandoc.Readers.Docx: Trigger bold/italic with bCs, iCs
(#4947). These are variants for "complex scripts" like Arabic
and are now treated just like b, i (bold, italic).

* Text.Pandoc.Readers.Muse (Alexander Krotov)

+ Try to parse lists before trying to parse table.
This ensures that tables inside lists are parsed correctly.
+ Forbid whitespace after opening and before closing markup
+ Parse page breaks.
+ Simplify `museToPandocTable` to get rid of partial functions.
+ Allow footnotes to start with empty line.
+ Make sure that the whole text is parsed.
+ Allow empty headers. Previously empty headers caused parser to
terminate without parsing the rest of the document.
+ Allow examples to be indented with tabs.
+ Remove indentation from examples indicated by `{{{` and `}}}`.
+ Fix parsing of empty cells.
+ Various changes to internals.
+ Rewrite some parsers in applicative style.
+ Avoid tagsoup dependency.
+ Allow table caption to contain `+`.

* Text.Pandoc.Writers.LaTeX

+ Add newline if math ends in a comment (#4880). This prevents the
closing delimiter from being swalled up in the comment.
+ With `--listings`, don't pass through org-babel attributes (#4889).
+ With `--biblatex`, use `\autocite` when possible (#4960).
`\autocites{a1}{a2}{a3}` will not collapse the entries. So, if
we don't have prefixes and suffixes, we use instead
+ Fix description lists contining highlighted code (#4662).

* Text.Pandoc.Writers.Man

+ Don't wrap `.SH` and `.SS` lines (#5019).
+ Avoid unnecessary `.RS`/`.RE` pair in definition lists with
one paragraph definitions.
+ Moved common groff functions to Text.Pandoc.Writers.Groff.
* Fix strong/code combination on man (should be `\f[CB]` not
`\f[BC]`, see #4973).
+ Man writer: use `\f[R]` instead of `\f[]` to reset font
(Alexander Krotov, #4973).
+ Move `splitSentences` to Text.Pandoc.Shared.

* Text.Pandoc.Writers.Docx

+ Add framework for custom properties (#3034). So far, we don't
actually write any custom properties, but we have the
infrastructure to add this.

+ Handle tables in table cells (#4953). Although this is not
documented in the spec, some versions of Word require a `w:p`
element inside every table cell. Thus, we add one when the
contents of a cell do not already include one (e.g. when a table
cell contains a table).

* Text.Pandoc.Writers.AsciiDoc: Prevent illegal nestings.
Adjust header levels so that n+1 level headers are only
found under n level headers, and the top level is 1.

* Text.Pandoc.Writers.OpenDocument: Improve bullet/numbering
alignment (#4385). This change eliminates the large gap we used
to have between bullet and text, and also ensures that numbers
in numbered lists will be right-aligned.

* Text.Pandoc.Writers.ZimWiki

+ Number ordered list items sequentially, rather than always
with 1 (#4962).
+ Remove extra indentation on lists (#4963).

* Text.Pandoc.Writers.EPUB: Use metadata field `css` instead of
`stylesheet` (Mauro Bieg, #4990).

* Text.Pandoc.Writers.Markdown: Ensure blank between raw block and
normal content (#4629). Otherwise a raw block can prevent a
paragraph from being recognized as such.

* Text.Pandoc.Writers.Ms

+ Removed old `escapeBar`. We don't need this now that we use
`@` for math delim.
+ Moved common code to Text.Pandoc.Writers.Roff and to
+ Move `splitSentences` to Text.Pandoc.Shared (to avoid duplication
with the man writer).

* Text.Pandoc.Writers.Muse (Alexander Krotov).

+ Add support for grid tables.
+ Fix Muse writer style.
+ Use `length` instead of `realLength` to calculate definition
indentation. Muse parsers don't take character width into
account when calculating indentation.
+ Do not insert newline before lists.
+ Use lightweight markup after `</em>` tag.

* New unexported module Text.Pandoc.Writers.Roff, providing functions
useful for all roff format writers (man, ms).

* Text.Pandoc.Lua

+ Move globals handling to separate module Text.Pandoc.Lua.Global
(Albert Krewinkel).

+ Lua filter internals: push Shared.Element as userdata (Albert
Krewinkel). Hierarchical Elements were pushed to Lua as plain
tables. This is simple, but has the disadvantage that marshaling
is eager: all child elements will be marshaled as part of the
object. Using a Lua userdata object instead allows lazy access
to fields, causing content marshaling just (but also each time)
when a field is accessed. Filters which do not traverse the full
element contents tree become faster as a result.

[default template changes]

* LaTeX template:

+ Add variable `hyperrefoptions` (#4925, Mathias Walter).
+ Add variable `romanfont`, `romanfontoptions` (#4665, OvidiusCicero).

* AsciiDoc template: use single-line style for title.

* revealjs template: Fix typo in the javascript plugin (#5006,
Yoan Blanc).

* Text.Pandoc.Lua.Util: add missing docstring to `defineHowTo`
(Albert Krewinkel).

* data/pandoc.lua: add datatype ListAttributes (Albert Krewinkel)

* data/sample.lua: replace custom pipe function with pandoc.utils.pipe
(Albert Krewinkel).

[documentation improvements]


+ Add chromeos install instructions (#4958) (Evan Pratten).
+ Add note about TinyTeX.

* MANUAL.txt

+ Change `groff` -> `roff`.
+ Implement `--ascii` for Markdown writer.
+ Clarify LaTeX image dimensions output (Mauro Bieg).

* doc/ added skeleton (Mauro Bieg, #3288).

* doc/ Added title to to avoid warning.

* doc/ merge type references into main document,
fix description of Code.text (Albert Krewinkel).

[build infrastructure improvements]

* Makefile

+ Makefile: added quick-cabal, full-cabal targets.
+ Make .msi download targets insensitive to order of appveyor builds.

* Update benchmarks for ghc 8.6.1.

* pandoc.cabal:

+ Enable more compiler warnings (Albert Krewinkel).
+ Make base lower bound 4.8.
+ Bump upper bound for QuickCheck.
+ Bump upper bound for binary.
+ Updated version bounds for containers and haddock-library (#4974).
+ Added docx/docPropos/custom.xml to cabal data-files.
+ Require skylighting 0.7.4 (#4920).
+ New dependency on unicode-transforms package for normalization.

* Improved .travis.yml testing and test with GHC 8.6.1 (Albert Krewinkel).

* Added `tools/`.

* Added test/ for testing the man reader on real man pages.

pandoc (2.3.1)

* RST reader:

+ Parse RST inlines containing newlines (#4912, Francesco Occhipinti).
This eliminates a regression introduced after pandoc 2.1.1, which
caused inline constructions containing newlines not to be recognized.
+ Fix bug with internal link targets (#4919). They were gobbling up
indented content underneath.

* Markdown reader: distinguish autolinks in the AST. With this change,
autolinks are parsed as Links with the `uri` class. (The same is true
for bare links, if the `autolink_bare_uris` extension is enabled.)
Email autolinks are parsed as Links with the `email` class. This
allows the distinction to be represented in the AST.

* Org reader:

+ Force inline code blocks to honor export options (Brian Leung).
+ Parse empty argument array in inline src blocks (Brian Leung).

* Muse reader (Alexander Krotov):

+ Added additional tests.
+ Do not allow code markup to be followed by digit.
+ Remove heading level limit.
+ Simplify `<literal>` tag parsers
+ Parse Text instead of String. Benchmark shows 7% improvement.
+ Get rid of HTML parser dependency.
+ Various code improvements.

* ConTeXt writer: change `\` to `/` in Windows image paths (#4918).
We do this in the LaTeX writer, and it avoids problems.
Note that `/` works as a LaTeX path separator on Windows.

* LaTeX writer:

+ Add support for multiprenote and multipostnote arguments
with `--biblatex` (Brian Leung, #4930). The multiprenotes occur before
the first prefix of a multicite, and the multipostnotes follow the
last suffix.
+ Fix a use of `last` that might take empty list. If you ran with
`--biblatex` and have an empty document (metadata but no blocks),
pandoc would previously raise an error because of the use of
`last` on an empty list.

* RTF writer: Fix build failure with ghc-8.6.1 caused by missing
MonadFail instance (Jonas Scholl).

* ODT Writer: Improve table header row style handling (Nils Carlson).
This changes the way styles for cells in the header row
and normal rows are handled in ODT tables.
Previously a new (but identical) style was generated for
every table, specifying the style of the cells within the table.
After this change there are two style definitions for table cells,
one for the cells in the header row, one for all other cells.
This doesn't change the actual styles, but makes post-processing
changes to the table styles much simpler as it is no longer
necessary to introduce new styles for header rows and there are
now only two styles where there was previously one per table.

* HTML writer:

+ Don't add `uri` class to presumed autolinks. Formerly the `uri`
class was added to autolinks by the HTML writer, but it had to
guess what was an autolink and could not distinguish
`[](` from
`<>`. It also incorrectly recognized
`[pandoc](pandoc)` as an autolink. Now the HTML writer
simply passes through the `uri` attribute if it is present,
but does not add anything.
+ Avoid adding extra section nestings for revealjs.
Previously revealjs title slides at level (slidelevel - 1)
were nested under an extra section element, even when
the section contained no additional (vertical) content.
That caused problems for some transition effects.
+ Omit unknown attributes in EPUB2 output. For example,
`epub:type` attributes should not be passed through,
or the epub produced will not validate.

* JATS writer: remove 'role' attribute on 'bold' and 'sc' elements (#4937).
The JATS spec does not allow these.

* Textile writer: don't represent `uri` class explicitly
for autolinks (#4913).

* Lua filters (Albert Krewinkel):

+ Cleanup filter execution code.
+ Better error on test failure.

* HTML, Muse reader tests: reduce time taken by round-trip test.

* Added cabal.project.

* MANUAL: `epub:type` is only useful for epub3 (Maura Bieg).

* Use hslua v1.0.0 (Albert Krewinkel).

* Fix `translations/ru` to use modern Russian orthography
(Ivan Trubach).

* Build Windows binary using ghc 8.6.1 and cabal new-build. This
fixes issues with segfaults in the 32-bit Windows binaries (#4283).

pandoc (2.3)

* Add `--metadata-file` option (Mauro Bieg, #1960), which allows
users to specify metadata in a YAML file, regardless of the
input format (#1960).

* Text.Pandoc.Writers.Shared: export `isDisplayMath` (API change).

* Text.Pandoc.Readers.Markdown: export `yamlToMeta` (API change,
Mauro Bieg).

* Text.Pandoc.Readers.LaTeX.Types:

+ New type `ArgSpec` (API change).
+ Second parameter of `Macro` constructor is now `[ArgSpec]` instead of
`Int` (API change).

* Markdown reader:

+ Use `tex` instead of `latex` for raw tex-ish content. We can't
always tell if it's LaTeX, ConTeXt, or plain TeX. Better just to
use `tex` always. Note that if `context` or `latex` specifically
is desired, you can still force that in a markdown document by
using the raw attribute. Note that this change may affect some
filters, if they assume that raw tex parsed by the Markdown reader
will be RawBlock (Format `latex`). In most cases it should be
trivial to modify the filters to accept `tex` as well.
+ Refactor and reorganize YAML code (Mauro Bieg).
+ Make `example_lists` work for interrupted lists
without `startnum` (#4908).

* HTML reader:

+ Parse `<script type="math/tex` tags as math (#4877).
These are used by MathJax in some configurations.
+ Extract spaces inside links instead of trimming them
(Alexander Krotov, #4845).
+ Added round-trip tests (Alexander Krotov).
+ Make parsing sensitive to the `raw_tex` extension (#1126). This now
allows raw LaTeX environments, `\ref`, and `\eqref` to be parsed
(which is helpful for translation HTML documents using MathJaX).

* Org reader (Albert Krewinkel):

+ Respect export option `p` for planning info. Inclusion of planning info
(`*DEADLINE*`, `*SCHEDULED*`, and `*CLOSED*`) can be controlled via the
`p` export option: setting the option to `t` will add all planning
information in a *Plain* block below the respective headline.
+ Org reader internals: don't rely on RecordWildCards and ViewPatterns
ghc extensions.
+ Strip planning info from output. Planning info is parsed, but not
included in the output (as is the default with Emacs Org-mode, #4867).

* LaTeX reader:

+ Split some general-purpose definitions into a separate, unexported
module, Text.Pandoc.Readers.LaTeX.Parsing. This is to avoid
out-of-memory errors compiling the LaTeX reader.
+ Handle parameter patterns for `\def` (#4768, #4771).
For example: `\def\foo#1[#2]{#1 and #2}`.
+ Allow `%` characters in URLs. This affects `\href` and `\url` (#4832).
+ Fixed parsing of `\texorpdfstring`. We were returning the wrong
argument as the content.
+ Support `blockcquote`, `foreignblockquote`, `foreigncblockquote`,
`hyphenblockquote`, `hyphencblockquote`, `enquote*`, `foreignquote`,
`hyphenquote` from `csquotes` (#4848). Note that `foreignquote`
will be parsed as a regular Quoted inline (not using the quotes
appropriate to the foreign language).
+ Support more text-mode accents (#4652). Add support for `\|`, `\b`, `\G`,
`\h`, `\d`, `\f`, `\r`, `\t`, `\U`, `\i`, `\j`, `\newtie`,
`\textcircled`. Also fall back to combining characters when composed
characters are not available.
+ Resolve `\ref` for figure numbers.
+ Support `mintinline` (#4365, Marc Schreiber).
+ Fix siunitx unit commands so they are only recognized in
siunitx contexts (#4842). For example, `\l` outside of an
siunitx context should be l-slash, not l (for liter).
+ Fix double `unnumbered` class (#4838). The `unnumbered` class
was being included twice for starred sections.

* RST reader: Don't skip link definitions after comments (#4860).

* Muse reader (Alexander Krotov):

+ Close the `</quote>` in indented tag test.
There is already a separate test for unclosed `</quote>`.
+ Autonumber sections in the correct order. Parsing now stops
at each section header to ensure the header is registered before
parsing of the next section starts.
+ Move duplicate code into `headingStart` function.
+ Allow newline after opening `*` or `**`.
+ Don't allow digits after closing marker in lightweight markup
This change makes reader more compatible with Emacs Muse.
+ Parse `<verse>` tag in one pass instead of using
`parseFromString`. This change makes it possible to have
verbatim `</verse>` tag inside verse.

* ODT reader: deal gracefully with missing `<office:font-face-decls/>`
(#4336). This allows pandoc to parse ODT document produced by KDE's

* Muse writer (Alexander Krotov):

+ Output headers without asterisks if not on the top level.
+ Never wrap definition list terms.
+ Set `envInsideBlock = True` when rendering notes.
+ Use `""` instead of `[]` for empty String.
+ Check for whitespace in the beginning and end of Str's.
+ Escape `-`, `;` and `>` in the beginning of strings.
+ Escape list markers in the beginning of notes.
+ Normalize inline list before testing if tags should be used.
+ Use tags instead of lightweight markup for empty strings.
+ Use lightweight markup when possible.
+ Escape empty strings. This guarantees that `conditionalEscapeString`
never returns empty string.
+ Wrap conditionalEscapeString result into `Muse` type.
This removes the need to pass `envInsideLinkDescription` to it.
+ Separate `shouldEscapeString` function.
+ Simplify inline list rendering.
+ Replace newlines in strings with spaces.

* Docx writer:

+ Add MetaString case for abstract, subtitle (#4900, Mauro Bieg).
+ Properly handle display math in spans (#4826). This isn't a
complete solution, since other nestings of display math may still
cause problems, but it should work for what is by far the most
common case.

* HTML writer:

+ Always output `<dt>` element, even if it is empty (#4883,
Alexander Krotov).
+ Don't prefix `epub:` attributes with `data-`.

* Org writer: Don't escape literal `_`, `^` (#4882). Org doesn't
recognize these escapes.

* ODT writer: Fix percentage image scaling (#4881, Nils Carlson).
Image scaling was broken when a width was set to a percentage.

* EPUB writer: set `epub:type` on body element in each chapter,
depending on the `epub:type` of the first section (#4823). This
only affects epub3. See

* FB2 writer: put `coverpage` element between title and date rather than in
`document-info` element (#4854).

* Markdown writer: Escape `~` if strikeout extension enabled (#4840).

* Haddock writer: Use proper format for latex math in haddock (#4571, Joe
Hermaszewski). Inline math in `\(..\)`, display math in `\[..\]`,
tex is now used. Previously we'd "fake it with unicode" and fall
back to tex when that didn't work. But newer haddock versions
support latex math.

* TEI writer:

+ Ensure that title element is always present, even if empty (#4839).
+ Put author tags in the template, rather than adding them in
the writer (#4839).

* LaTeX writer/template: be sensitive to `filecolor` variable (#4822).
`linkcolor` only affects internal links, and `urlcolor` only
affects linked URLs. For external links, the option to use is

* ConTeXt writer: output raw `tex` blocks as well as `context` (#969).

* RST writer:

+ Use `.. raw:: latex` for `tex` content.
+ Use `.. container` for generic Divs, instead of raw HTML.
+ Render Divs with admonition classes as admonitions (#4833).
Also omit Div with class `admonition-title`. These are generated
by the RST reader and should be omitted on round-trip.

* Text.Pandoc.PDF: fix message printed when rsvg-convert is not available
(#4855, Antonio Terceiro).

* HTML5 template: add the `title-block-header` identifier to the
`header` element, to make it easier to style precisely (#4767,
J. B. Rainsberger).

* OpenDocument template: Remove unnecessary indenting of TOC title (#4798,
José de Mattos Neto).

* latex template: Add support for $toc-title$ to LaTeX (and PDF)
(#4853, Wandmalfarbe).

* TEI template: improve `publicationStmt`. Add support for
`publisher`, `address`, `pubPlace`, and `date` variables.

* beamer template: Support "toc-title" (#4835, Cyril Roelandt).

* Text.Pandoc.Extensions: Fix haddock on `Ext_footnotes` (Chris Martin).

* Lua: cleanup Lua utils, remove unused functions (Albert Krewinkel).

* MANUAL.txt:

+ Clarify that `--biblatex/--natbib` don't work directly for PDF
+ Document `epub:type` attribute (Mauro Bieg, #4901)
+ Clarify when `--resource-path` has an effect.
+ More detail on customization in syntax highlighting section.
+ Document encoding issue with `--listings` (#4871, Damien Clochard).
+ Remove docs on removed `--katex-stylesheet` (Mauro Bieg, #4862).
+ Use https for context wiki links (#4910).


+ Link to lua-filters repository (#4874).
+ Fix mistake in REPL instructions for stack. (#4849, Brian Leung).

* add links to filters, and to lua-filters repository


+ Indicate that cabal >= 2.0 is needed.
+ Added chocolatey installation method (#4844, Miodrag Milić).

* Travis: exclude round-trip tests, except for nightly test which can fail.

* Use latest texmath, pandoc-citeproc.

* Use a patched version of foundation until
is fixed.

* Clean up appveyor build and Windows package creation.
We now use 64-bit stack and ghc 8.4.3, lts-12 for the 64-bit build. The
WiX-based msi is now 64-bit for 64-bit builds (fixing #4795).

* Remove obsolete

* Added additional compiler warnings in Makefile and CI builds.

pandoc (

* Markdown reader: Properly handle boolean values in YAML metadata (#4819).
This fixes a regression in 2.2.3, which cause boolean values to
be parsed as MetaInlines instead of MetaBool.

We here record another undocumented (but desirable) change in 2.2.3:
numerical metadata fields are now parsed as MetaInlines rather than

pandoc (

* Markdown reader: Fix parsing of embedded mappings in YAML metadata
(#4817). This fixes a regression in 2.2.3 which caused embedded
mappings (e.g. mappings in sequences) not to work in YAML metadata.

pandoc (2.2.3)

* RST reader: improve parsing of inline interpreted text roles (#4811).

+ Use a Span with class "title-reference" for the default
title-reference role.
+ Use `B.text` to split up contents into `Space`s, `SoftBreak`s,
and `Str`s for `title-reference`.
+ Use Code with class "interpreted-text" instead of Span and Str for
unknown roles. (The RST writer has also been modified to round-trip
this properly.)
+ Disallow blank lines in interpreted text.
+ Backslash-escape now works in interpreted text.
+ Backticks followed by alphanumerics no longer end interpreted text.
+ Remove support for nested inlines (Francesco Occhipinti).
RST does not allow nested emphasis, links, or other inline
constructs. This fixes several bugs (#4581, #4561, #4792).

* Org reader: fix parsers relying on `parseFromString` (#4784, Albert
Krewinkel). Emphasis was not parsed when it followed directly after
some block types (e.g., lists).

* Markdown reader: Allow unquoted numbers and booleans as YAML mapping
keys. Previously in 2.2.2 you could not do
0: bar
but only
'0': bar
With this change, both forms work.

* DocBook reader: metadata handling improvements.
Now we properly parse title and subtitle elements that are direct
children of book and article (as well as children of bookinfo,
articleinfo, or info). We also now use the `subtitle` metadata
field for subtitles, rather than tacking the subtitle on to the

* RST writer:

+ Allow images to be directly nested within links (#4810, Francesco
+ Use `titleblock` instead of `title` variable for title block (#4803,
Francesco Occhipinti). `titleblock` contains a properly formatted
title and subtitle (using top-level headers). `title` and
`subtitle` variables are still available and just contain the
title and subtitle text. Note that this change will require an
update to custom rst templates.
+ Render Code with class "interpreted-text" as interpreted text role.

* MediaWiki writer: Avoid extra blank line in tables with empty cells
(#4794). Note that the old output is semantically identical, but the
new output looks better.

* Lua Utils module: add function `blocks_to_inlines` (#4799, Albert
Krewinkel). Exposes a function converting which flattenes a list of
blocks into a list of inlines. An example use case would be the
conversion of Note elements into other inlines.

* RST template: use `titleblock` instead of `title`. Users of
custom RST templates will want to update this.

* LaTeX template: Moved some beamer code in default.latex template.
This change allows beamer themes to change the template and font (as
Metropolis does) (#4450).

* Better error message on `-t pdf -o out.pdf` (#1155, Mauro Bieg).

* Added test case for #4669 to repository.

* Fix broken link for cabal-install (#4806, ChanHoHo).

* MANUAL.txt:

+ Add beamer info for slide backgrounds (#4802, John Muccigrosso).
+ Clarify when `csquotes` is used in LaTeX writer (#4514).
+ Add `commonmark` to list of output formats where `raw_tex` has an
effect (see #4527).

pandoc (

* Fix regression finding templates in user data directory (#4777).
Under version 2.2.1 and prior pandoc found latex templates in the
templates directory under the data directory, but this broke in 2.2.2.

* Fix for bug in parsing `\input` in `rawLaTeXBlock` and
`rawLaTeXInline` (#4781). (This primarily affects the markdown
reader, and other readers that accept raw tex.)
Starting in 2.2.2, everything after an `\input` (or `\include`)
in a markdown file would be parsed as raw LaTeX.


+ Clarify `gfm` vs `markdown_github` (#4783, Mauro Bieg).
* Use `keywords` instead of `tags` in YAML metadata example (#4779).
Unlike `tags`, `keywords` is used in some of the writers
and default templates.

* Add missing `rollingLinks` option to revealjs template (#4778,
Igor Khorlo).

pandoc (2.2.2)

* Use HsYAML instead of yaml for translations and YAML metadata (#4747).
yaml wraps a C library; HsYAML is pure Haskell. Advances #4535.

Note: HsYAML implements YAML 1.2, in which the valid true
values are `true`, `True`, `TRUE`. This means a change in
the semantics of YAML metadata that could affect users:
`y`, `yes`, and `on` no longer count as true values.

* Fix regression: make `--pdf-engine` work with full paths (#4681, Mauro

* CommonMark reader: Handle `ascii_identifiers` extension (#4742,
Anders Waldenborg). Non-ascii characters were not stripped from
identifiers even if the `ascii_identifiers` extension was
enabled (which is is by default for gfm).

* TikiWiki reader: Improve list parsing (#4722, Mauro Bieg).
Remove trailing Space from list items. Parse lists that have no space
after marker.

* LaTeX reader:

+ Treat `lilypond` as a verbatim environment (#4725).
+ Parse figure label into Image id (#4700, Mauro Bieg).
+ Beamer: Allow "noframenumbering" option (#4696, Raymond Ehlers).
+ Allow spaces around `\graphicspath` arguments (#4698).
+ Handle includes without surrounding blanklines (#4553).
In addition, `\input` can now be used in an inline context,
e.g. to provide part of a paragraph, as it can in LaTeX.
+ In `rawLaTeXBlock`, handle macros that resolve to a
`\begin` or `\end` (#4667).
+ In `rawLaTeXBlock`, don't expand macros in macro definitions (#4653).
Note that this only affected LaTeX in markdown.
+ Tighten up reading of beamer overlay specifications (#4669).
Ideally we'd turn these on only when reading beamer, but currently
beamer is not distinguished from latex as an input format.
This commit also activates parsing of overlay specifications
after commands in general (e.g. `\item`), since they can occur
in many contexts in beamer.
+ Parse more siunitx unit commands (#4296, #4773).
+ Be more forgiving in key/value option parsing (#4761).

* Markdown reader:

+ Allow empty code spans, e.g. `` ` ` ``.
+ Emojis are now wrapped in Spans with class `emoji` and
attribute `data-emoji` (Anders Waldenborg, #4743).
This allows the writer to handle them in a special way
(e.g. using a special font, or just rendering the
emoji name).

* Muse reader (Alexander Krotov, except where indicated):

+ Get rid of non-exhaustive pattern match warning (Mauro Bieg).
+ Add support for floating images.
+ Add support for images with specified width.
+ Parse image URLs without "guard" and "takeExtension".
+ Split link and image parsing into separate functions.
+ Parse links starting with "URL:" explicitly instead of trying to strip
"URL:" prefix after parsing.

* Texinfo writer: Use `@sup` and `@sub` instead of custom macros (#4728,
Alexander Krotov).

* Markdown writer: Preserve `implicit_figures` with attributes, even if
`implicit_attributes` is not set, by rendering in raw HTML (#4677).

* Markdown and commonmark/github writers now respect the `emoji`
extension. So, `-f markdown+emoji -t markdown+emoji` now leaves
`:smile:` as `:smile:` rather than converting it to a smile

* Docx writer: Be sensitive to `toc` in YAML metadata (#4645).

* ODT/OpenDocument writer: Make internal links work (#4358).
This adds proper bookmarks to the headers with non-null IDs.

* EPUB writer: Properly escape pagetitle. Previously we weren't
escaping `&` and other XML characters in the pagetitle, so a title
containing a `&` would be invalid.

* AsciiDoc Writer: Eescape square brackets at start of line (#4545,
Mauro Bieg).

* RST writer:

+ Don't treat 'example' as a syntax name (#4748).
This fixes conversions from org with example blocks.
+ Support `--number-sections` via the `section-numbering`
directive in standalone output.

* reveal.js writer and template: reuse mathjax URL
provided by the argument to `--mathjax` or the normal pandoc default,
rather than a hard-coded one in the template (#4701).

* LaTeX writer:

+ Properly handle footnotes in table captions (#4683).
Refactored code from figure captions to use in both places.
+ In beamer output, fix single digit column percentage (#4690, Mauro

* FB2 writer (Alexander Krotov):

+ Convert Plain to Para in annotation (#2424).
+ Fix order of items in title-info (#2424).

* Custom writer: fix error message on script failure (Albert Krewinkel).
Error messages produced by Lua were not displayed by Pandoc.

* Text.Pandoc.Emoji now exports `emojiToInline`, which returns a Span
inline containing the emoji character and some attributes with metadata
(class `emoji`, attribute `data-emoji` with emoji name). (API change,
Anders Waldenborg, #4743).

* Text.Pandoc.PDF:

+ Revert fix for #4484 (only compress images on last run, #4755).
This will mean some increase in the time it takes to
produce an image-heavy PDF with xelatex, but it will
make tables of contents correct, which is more important.
+ Fix logic error in `runTeXProgram`. We were running the tex program
one more time than requested. This should speed up PDF production.

* Allow `--template` to take a URL as argument.

* Text.Pandoc.Highlighting: Add missing re-export of `breezeDark`
highlighting style (#4687, Adrian Sieber, API change).

* Clarify macOS install in (#4661). Make the binary package
installer the recommended method, and note that on some older versions of
macOS, homebrew installs from source and takes a lot of disk space
(#4664, Ian).


+ Clarify EPUB linked media (#4756, Mauro Bieg)
+ Update manual for "true" YAML values. Now that we're using HsYAML and
YAML 1.2, the valid true values are `true`, `True`, `TRUE`. NOTE!
`y`, `yes`, `on` no longer count as true values.
+ Document `-F` as alias for `--filter` (thanks to Gandalf Saxe).
+ Update manual on how math is rendered in LaTeX.
+ Add proxy description (#4131, Mauro Bieg).
+ Clarify that `--toc` requires `--standalone` (#4703).
+ Update citation styles link (#4699, wiefling).

* In API docs, clarify how `Ext_east_asian_line_breaks` extension works
(kaizshang91). Note that it will not take effect when readers/writers
are called as libraries (#4674).

* Improved translations/fr (#4766, lux-lth).

* Removed inadvertently added `.orig` files from repository (#4648).

* Remove `network-uri` flag and use 'Network.Socket'.
This removes a compiler warning. There is no need for the old
`network-uri` flag, since network 2.6 was released in 2014.

* Add stack.lts10.yaml, stack.lts11.yaml. use lts-12 in stack.yaml.

* Bump upper bounds for dependent packages.

* Exclude foundation 0.0.21 for ghc 7.10. Otherwise cabal gets
confused because of the way ghc 7.10 is excluded in foundation's
cabal file. This can be removed when haskell-foundation/foundation#500
is fixed.

* Require cabal-version >= 2.0. This is needed for haddock-library.

pandoc (2.2.1)

* Restored and undeprecated gladtex for HTML math (#4607).

+ Added `GladTeX` constructor to `Text.Pandoc.Options.HTMLMathMethod`
[API change, reverts removal in v2.2]
+ Restored and undeprecated `--gladtex` option, removed in v2.2.

* LaTeX reader: Handle `$` in `/text{..}` inside math (#4576).

* Org reader: Fix image filename recognition (Albert Krewinkel).
Use a function from the filepath library to check whether a string is a
valid file name. The custom validity checker that was used before gave
wrong results (e.g. for absolute file paths on Windows,

* FB2 reader: Replace some errors with warnings (Alexander Krotov).

* HTML writer:

+ Strip links from headers when creating TOC (#4340).
Otherwise the TOC entries will not link to the sections.
+ Fix regression with tex math environments in HTML + MathJax (#4639).

* Muse writer (Alexander Krotov): Add support for left-align and
right-align classes (#4542).

* Docx writer: Support underline (#4633).

* Text.Pandoc.Parsing: Lookahead for non-whitespace after
`singleQuoteStart` and `doubleQuoteStart` (#4637).

* `test-pandoc-utils.lua`: more robust testing on both windows
and \*nix. Previously the pipe tests were only run if
`\bin/false` and `/bin/sed` were present, which they aren't
in default MacOS and Windows systems. Fixed by using `tr`
and `false`, which should always be in the path on a \*nix
system, and `find` and `echo` for Windows.

* Text.Pandoc.Shared: add `uriPathToPath`.
This adjusts the path from a file: URI in a way that is sensitive
to Windows/Linux differences. Thus, on Windows,
`/c:/foo` gets interpreted as `c:/foo`, but on Linux,
`/c:/foo` gets interpreted as `/c:/foo`. See #4613.

* Use `uriPathToPath` with file: URIs (#4613).

* Revert piping HTML to pdf-engine (Mauro Bieg, #4413). Use a temp
file as before.

* Text.Pandoc.Class: Catch IO errors when writing media files
and issue a warning, rather than an error (Francesco Occhipinti, #4559).

* Don't lowercase custom writer filename (Alexander Krotov, #4610).

* MANUAL (Mauro Bieg):

+ Clarify truthiness in template variables (#2281).
+ Clarify pipe table width calculation (#4520).

* ConTeXt template: New Greek fallback typeface (Pablo Rodríguez, #4405).
CMU Serif gives better typographic results than the previous
Greek fallback DejaVu Serif.

* Make HTML template polyglot (#4606, OvidiusCicero), by making
`<link rel="stylesheet" href="$css$">` self-closing.

* Use texmath 0.11, allowing better translation of non-ASCII
characters in math (#4642).

pandoc (2.2)

* New input format: `fb2` (FictionBook2) (Alexander Krotov).

* Make `--ascii` work for all XML formats (ICML, OPML, JATS,...),
and for `ms` and `man`.

* Remove deprecated `--latexmathml`, `--gladtex`, `--mimetex`, `--jsmath`, `-m`,
`--asciimathml` options.

* New module Text.Pandoc.Readers.FB2, exporting readFB2 (Alexander
Krotov, API change).

* Markdown reader:

+ Allow empty key-value attributes, like `title=""` (#2944).
+ Handle table w/o following blank line in fenced div (#4560).
+ Remove "fallback" for `doubleQuote` parser. Previously the
parser tried to be efficient -- if no end double quote was found,
it would just return the contents. But this could backfire in a
case `**this should "be bold**`, since the fallback would return
the content `"be bold**` and the closing boldface delimiter
would never be encountered.
+ Improve computation of the relative width of the last column in a
multiline table, so we can round-trip tables without constantly
shrinking the last column.

* EPUB reader:

+ Fix images with space in file path (#4344).

* LaTeX reader:

+ Properly resolve section numbers with `\ref` and chapters (#4529).
+ Parse sloppypar environment (#4517, Marc Schreiber).
+ Improve handling of raw LaTeX (for markdown etc.) (#4589, #4594).
Previously there were some bugs in how macros were handled.
+ Support `\MakeUppercase`, `\MakeLowercase', `\uppercase`, `\lowercase`,
and also `\MakeTextUppercase` and `\MakeTextLowercase` from textcase

* Textile reader:

+ Fixed tables with no body rows (#4513).
Previously these raised an exception.

* Mediawiki reader:

+ Improve table parsing (#4508). This fixes detection of table
attributes and also handles `!` characters in cells.

* DocBook reader:

+ Properly handle title in `section` element (#4526).
Previously we just got `section_title` for `section` (though `sect1`,
`sect2`, etc. were handled properly).
+ Read tex math as output by asciidoctor (#4569, Joe Hermaszewski).

* Docx reader:

+ Combine adjacent CodeBlocks with the same attributes into
a single CodeBlock. This prevents a multiline codeblock in
Word from being read as different paragraphs.

* RST reader:

+ Allow < 3 spaces indent under directives (#4579).
+ Fix anonymous redirects with backticks (#4598).

* Muse reader (Alexander Krotov):

+ Add support for Text::Amuse multiline headings.
+ Add `<math>` tag support.
+ Add support for `<biblio>` and `<play>` tags.
+ Allow links to have empty descriptions.
+ Require block `<literal>` tags to be on separate lines.
+ Allow `-` in anchors.
+ Allow verse to be indented.
+ Allow nested footnotes.
+ Internal improvements.

* Muse writer (Alexander Krotov):

+ Escape `>` only at the beginning of a line.
+ Escape `]` in image title.
+ Escape `]` brackets in URLs as `%5D`.
+ Only escape brackets when necessary.
+ Escape ordered list markers.
+ Do not escape list markers unless preceded by space.
+ Escape strings starting with space.
+ Escape semicolons and markers after line break.
+ Escape `;` to avoid accidental comments.
+ Don't break headers, line blocks and tables with line breaks.
+ Correctly output empty headings.
+ Escape horizontal rule only if at the beginning of the line.
+ Escape definition list terms starting with list markers.
+ Place header IDs before header.
+ Improve span writing.
+ Do not join Spans in normalization.
+ Don't align ordered list items.
+ Remove key-value pairs from attributes before normalization.
+ Enable `--wrap=preserve` for all tests by default.
+ Reduced `<verbatim>` tags in output.
+ Internal changes.

* RST writer:

+ Use more consistent indentation (#4563). Previously we
used an odd mix of 3- and 4-space indentation. Now we use 3-space
indentation, except for ordered lists, where indentation must
depend on the width of the list marker.
+ Flatten nested inlines (#4368, Francesco Occhipinti).
Nested inlines are not valid RST syntax, so we flatten them following
some readability criteria discussed in #4368.

* EPUB writer:

+ Ensure that `pagetitle` is always set, even when structured titles
are used. This prevents spurious warnings about empty title
elements (#4486).

* FB2 writer (Alexander Krotov):

+ Output links inline instead of producing notes. Previously all links
were turned into footnotes with unclickable URLs inside.
+ Allow emphasis and notes in titles.
+ Don't intersperse paragraph with empty lines.
+ Convert metadata value `abstract` to book annotation.
+ Use `<empty-line />` for `HorizontalRule` rather than `LineBreak`.
FB2 does not have a way to represent line breaks inside paragraphs;
previously we used `<empty-line />` elements, but these are not allowed
inside paragraphs.

* Powerpoint writer (Jesse Rosenthal):

+ Handle Quoted Inlines (#4532).
+ Simplify code with `ParseXml`.
+ Allow fallback options when looking for placeholder type.
+ Check reference-doc for all layouts.
+ Simplify speaker notes logic.
+ Change notes state to a simpler per-slide value.
+ Remove `Maybe` from `SpeakerNotes` in `Slide`. `mempty`
means no speaker notes.
+ Add tests for improved speaker notes.
+ Handle speaker notes earlier in the conversion process.
+ Keep notes with related blocks (#4477). Some blocks automatically
split slides (imgs, tables, `column` divs). We assume that any
speaker notes immediately following these are connected to these
elements, and keep them with the related blocks, splitting after them.
+ Remove `docProps/thumbnail.jpeg` in data dir (Jesse Rosenthal, #4588).
It contained a nonfree ICC color calibration profile and is not needed
for production of a powerpoint document.

* Markdown writer:

+ Include a blank line at the end of the row in a single-row multiline
table, to prevent it from being interpreted as a simple table (#4578).

* CommonMark writer:

+ Correctly ignore LaTeX raw blocks when `raw_tex` is not
enabled (#4527, quasicomputational).

* EPUB writer:

+ Add `epub:type="footnotes"` to notes section in EPUB3 (#4489).

* LaTeX writer:

+ In beamer, don't use format specifier for default ordered lists
(#4556). This gives better results for styles that put ordered list
markers in boxes or circles.
+ Update `\lstinline` delimiters (#4369, Tim Parenti).

* Ms writer:

+ Use `\f[R]` rather than `\f[]` to reset font (#4552).
+ Use `\f[BI]` and `\f[CB]` in headers, instead of `\f[I]` and `\f[C]`,
since the header font is automatically bold (#4552).
+ Use `\f[CB]` rather than `\f[BC]` for monospace bold (#4552).
+ Create pdf anchor for a Div with an identifier (#4515).
+ Escape `/` character in anchor ids (#4515).
+ Improve escaping for anchor ids: we now use _uNNN_ instead of uNNN
to avoid ambiguity.

* Man writer:

+ Don't escape U+2019 as `'` (#4550).

* Text.Pandoc.Options:

+ Removed `JsMath`, `LaTeXMathML`, and `GladTeX` constructors from
`Text.Pandoc.Options.HTMLMathMethod` [API change].

* Text.Pandoc.Class:

+ `writeMedia`: unescape URI-escaping in file path. This avoids
writing things like `file%20one.png` to the file system.

* Text.Pandoc.Parsing:

+ Fix `romanNumeral` parser (#4480). We previously accepted 'DDC'
as 1100.
+ `uri`: don't treat `*` characters at end as part of URI (#4561).

* Text.Pandoc.MIME:

+ Use the alias `application/eps` for EPS (#2067).
This will ensure that we retain the eps extension after reading the
image into a mediabag and writing it again.

* Text.Pandoc.PDF:

+ Use `withTempDir` in `html2pdf`.
+ With `xelatex`, don't compress images til the last run (#4484).
This saves time for image-heavy documents.
+ Don't try to convert EPS files (#2067). `pdflatex` converts them
itself, and JuicyPixels can't do it.
+ For `pdflatex`, use a temp directory in the working directory.
Otherwise we can have problems with the EPS conversion pdflatex
tries to do, which can't operate on a file above the working
directory without `--shell-escape`.

* Changes to tests to accommodate changes in pandoc-types.
In <> we changed
the table builder to pad cells. This commit changes tests
(and two readers) to accord with this behavior.

* Set default extensions for `beamer` same as `latex`.

* LaTeX template:

+ Add `beameroption` variable (#4359, Étienne Bersac).
+ Use `pgfpages` package; this is needed for notes on second
screen in beamer (Étienne Bersac).
+ Add `background-image` variable (#4601, John Muccigrosso).

* reveal.js template: Add `background-image` variable (#4600,
John Muccigrosso).

* ms template: Fix date. Previously `.ND` was used, but this only
works if you have a title page, which we don't. Thanks to @teoric.

* Removed pragmas for unused extensions (#4506, Anabra).

* Fix bash completion for `--print-default-data-file` (#4549).
Previously this looked in the filesystem, even if pandoc
was compiled with `embed_data_files` (and sometimes it looked
in a nonexistent build directory). Now the bash completion
script just includes a hard-coded list of data file names.


+ Clarify template vs metadata variables (#4501, Mauro Bieg).
+ Fix raw content example (#4479, Mauro Bieg).
+ Specify that you use html for raw output in epub.
+ Add examples for raw docx blocks (#4472, Tristan Stenner).
The documentation states that the target format name should match
the output format, which isn't the case for `docx`/`openxml` and
some others.
+ Don't say that `empty_paragraphs` affects markdown output (#4540).
+ Consolidate input/output format documentation (#4577, Mauro Bieg).

* New README template. Take in/out formats from manual.

* Fix example in lua-filters docs (#4459, HeirOfNorton).

* Use the `-threaded` GHC flag when building benchmarks (#4587,
Francesco Occhipinti).

* Bump temporary upper bound to 1.4.

* Use pandoc-citeproc

* Use texmath- (fixes escapes in math in ms, #4597).

* Removed old lib directory. This was used for something long ago,
but plays no role now.

* Removed unneeded data file `LaTeXMathML.js`.

* Create 64- and 32-bit versions of Windows binary packages.

pandoc (2.1.3)

* Docx reader (Jesse Rosenthal):

+ Add tests for nested smart tags.
+ Parse nested smart tags.
+ Make unwrapSDT into a general `unwrap` function that can unwrap both
nested SDT tags and smartTags. This makes the SmartTags constructor in
the Docx type unnecessary, so we remove it (#4446).
+ Remove unused `docxWarnings` (Alexander Krotov).

* RST reader: Allow unicode bullet characters (#4454).

* Haddock reader: Better table handling, using haddock-library's
new table support, if compiled against a version that
includes it. Note that tables with col/rowspans will not
translate well into Pandoc.

* Muse reader (Alexander Krotov):

+ Require closing tag to have the same indentation as opening.
+ Do not reparse blocks inside unclosed block tag (#4425).
+ Parse `<class>` tag (supported by Emacs Muse).
+ Do not produce empty Str element for unindented verse lines.
+ Don't allow footnote references inside links.
+ Allow URL to be empty.
+ Require that comment semicolons are in the first column (#4551).
+ Various internal improvements.

* LaTeX reader:

+ Add support to parse unit string of `\SI` command (closes #4296,
Marc Schreiber).

* Haddock writer: In the writer, we now render tables always as
grid tables, since Haddock supports these.

* DokuWiki writer: rewrite backSlashLineBreaks (#4445, Mauro Bieg).

* Docx writer: Fixed formatting of `DefaultStyle` ordered lists in
docx writer. We want decimal for the top level, not lower roman.

* RST writer:

+ Strip whitespace at beginning and ending of inline containers
(#4327, Francesco Occhipinti).
+ Filter out empty inline containers (#4434). There is nothing in
RST that corresponds to e.g. `Emph []`, so we just ignore elements
like this (Francesco Occhipinti).

* Muse writer (Alexander Krotov):

+ Support spans with anchors.
+ Replace smallcaps with emphasis before normalization.
+ Output smallcaps as emphasis.
+ Expand Cite before list normalization.
+ Write empty inline lists as `<verbatim></verbatim>`.
+ Remove empty Str from the beginning of inline lists during normalization.
+ Escape "-" to avoid creating bullet lists.
+ Fix math expansion for more than one expression per paragraph.
+ Expand math before inline list normalization.

* Dokuwiki writer: fix LineBreaks in Tables (#4313, Mauro Bieg).

* Ms writer:

+ Asciify pdf anchors, since unicode anchors don't work (#4436).
Internal links should be converted automatically, so this shouldn't
affect users directly.
+ Don't escape hyphens as `\-`; that's for a minus sign (#4467).

* Beamer writer: put hyperlink after `\begin{frame}` and not in the title
(#4307). If it's in the title, then we get a titlebar on slides with
the `plain` attribute, when the id is non-null. This fixes a regression
in 2.0.

* EPUB writer: Remove notes from TOC in nav.xhtml (#4453, Mauro Bieg).

* JATS writer: Remove extraneous, significant whitespace (#4335,
Nokome Bentley).

* html2pdf: inject base tag with current working directory (#4413, Mauro
Bieg). This helps ensure that linked resources are included.

* Add Semigroup instances for everything for which we defined a
Monoid instance previously (API change):

+ Text.Pandoc.Class.FileTree.
+ Text.Pandoc.Translations.Translations.
+ Text.Pandoc.Extensions.Extensions.
+ Text.Pandoc.Readers.Odt.StyleReader.Styles.
+ Text.Pandoc.Pretty.Doc.
+ Text.Pandoc.MediaBag.MediaBag.

* Add custom Prelude to give clean code for Monoid and Semigroup
that works with ghc 7.10-8.4. The custom Prelude (`prelude/Prelude`)
is used for ghc versions < 8.4. `NoImplicitPrelude` is used
in all source files, and Prelude is explicitly imported
(this is necessary for ghci to work properly with the custom prelude).

* Text.Pandoc.Writers.Shared (Francesco Occhipinti):

+ Export `stripLeadingTrailingSpace`.
+ Don't wrap lines in grid tables when `--wrap=none` (#4320).
+ `gridTable`: Don't wrap lines in tables when `--wrap=none`. Instead,
expand cells, even if it results in cells that don't respect relative
widths or surpass page column width. This change affects RST,
Markdown, and Haddock writers.

* Raise error if someone tries to print docx, odt, etc. template (#4441).

* LaTeX template: Provide `bidi` package's option using
`\PassOptionsToPackage` (#4357, Václav Haisman). This avoid a
clash when `polyglossia` loads it first and then it is loaded again
for XeLaTeX.

* ConTeXt template: Added `pdfa` variable to generate PDF/A (#4294, Henri
Menke). Instructions on how to install the ICC profiles on ConTeXt
standalone can be found in the wiki:
If the ICC profiles are not available the log will contain error

* Use latest pandoc-types, skylighting

* Use latest pandoc-citeproc in binary package.

* Bump upper bound for time, criterion, haddock-library, exceptions,
http-types, aeson, haddock-library.

* Bump upper bound tasty-quickcheck 0.10 (#4429, Felix Yan).

* pandoc.cabal: fix up other-extensions and language fields.
Language is now consistently `Haskell2010`, and other-extensions
is consistently `NoImplicitPrelude`. Everything else to be specified
in the module header as needed.

* Removed `old-locale` flag and Text.Pandoc.Compat.Time.
This is no longer necessary since we no longer support ghc 7.8.

* Make `weigh-pandoc` into a benchmark program.
Remove `weigh-pandoc` flag. `weigh-pandoc` is now built (and run)
automatically when you build (and run) benchmarks.

* MANUAL: add instructions for background images reveal.js (#4325, John

* appveyor: use VS 2013 environment instead of VS 2015 for Windows builds.

pandoc (2.1.2)

* Markdown reader:

+ Fix parsing bug with nested fenced divs (#4281). Previously we allowed
"nonindent spaces" before the opening and closing `:::`, but this
interfered with list parsing, so now we require the fences to be flush with
the margin of the containing block.

* Commonmark reader:

+ `raw_html` is now on by default. It can be disabled explicitly
using `-f commonmark-raw_html`.

* Org reader (Albert Krewinkel):

+ Move citation tests to separate module.
+ Allow changing emphasis syntax (#4378). The characters allowed before
and after emphasis can be configured via `#+pandoc-emphasis-pre` and
`#+pandoc-emphasis-post`, respectively. This allows to change which
strings are recognized as emphasized text on a per-document or even
per-paragraph basis. Example:

#+pandoc-emphasis-pre: "-\t ('\"{"
#+pandoc-emphasis-post: "-\t\n .,:!?;'\")}["

* LaTeX reader:

+ Fixed comments inside citations (#4374).
+ Fix regression in package options including underscore (#4424).
+ Make `--trace` work.
+ Fixed parsing of `tabular*` environment (#4279).

* RST reader:

+ Fix regression in parsing of headers with trailing space (#4280).

* Muse reader (Alexander Krotov):

+ Enable `<literal>` tags even if amuse extension is enabled.
Amusewiki disables <literal> tags for security reasons.
If user wants similar behavior in pandoc, RawBlocks and RawInlines
can be removed or replaced with filters.
+ Remove space prefix from `<literal>` tag contents.
+ Do not consume whitespace while looking for closing end tag.
+ Convert alphabetical list markers to decimal in round-trip test.
Alphabetical lists are an addition of Text::Amuse.
They are not present in Emacs Muse and can be ambiguous
when list starts with "i.", "c." etc.
+ Allow `<quote>` and other tags to be indented.
+ Allow single colon in definition list term.
+ Fix parsing of verse in lists.
+ Improved parsing efficiency. Avoid `parseFromString`.
Lists are parsed in linear instead of exponential time now.
+ Replace ParserState with MuseState.
+ Prioritize lists with roman numerals over alphabetical lists.
This is to make sure "i." starts a roman numbered list,
instead of a list with letter "i" (followed by "j", "k", ...").
+ Fix directive parsing.
+ Parse definition lists with multiple descriptions.
+ Parse next list item before parsing more item contents.
+ Fixed a bug: headers did not terminate lists.
+ Move indentation parsing from `definitionListItem` to `definitionList`.
+ Paragraph indentation does not indicate nested quote.
Muse allows indentation to indicate quotation or alignment,
but only on the top level, not within a <quote> or list.
+ Require that block tags are on separate lines.
Text::Amuse already explicitly requires it anyway.
+ Fix matching of closing inline tags.
+ Various internal changes.
+ Fix parsing of nested definition lists.
+ Require only one space for nested definition list indentation.
+ Do not remove trailing whitespace from `<code>`.
+ Fix parsing of trailing whitespace. Newline after whitespace now
results in softbreak instead of space.

* Docx reader (Jesse Rosenthal, except where noted):

+ Handle nested sdt tags (#4415).
+ Don't look up dependent run styles if `+styles` is enabled.
+ Move pandoc inline styling inside custom-style span.
+ Read custom styles (#1843). This will read all paragraph and
character classes as divs and spans, respectively. Dependent styles
will still be resolved, but will be wrapped with appropriate style
tags. It is controlled by the `+styles` extension (`-f docx+styles`).
This can be used in conjunction with the `custom-style` feature in the
docx writer for a pandoc-docx editing workflow. Users can convert from
an input docx, reading the custom-styles, and then use that same input
docx file as a reference-doc for producing an output docx file. Styles
will be maintained across the conversion, even if pandoc doesn't
understand them.
+ Small change to Fields hyperlink parser. Previously, unquoted string
required a space at the end of the line (and consumed it). Now we
either take a space (and don't consume it), or end of input.
+ Pick table width from the longest row or header (Francesco Occhipinti,

* Muse writer (Alexander Krotov):

+ Change verse markup: `> ` instead of `<verse>` tag.
+ Remove empty strings during inline normalization.
+ Don't indent nested definition lists.
+ Use unicode quotes for quoted text.
+ Write image width specified in percent in Text::Amuse mode.
+ Don't wrap displayMath into `<verse>`.
+ Escape nonbreaking space (`~~`).
+ Join code with different attributes during normalization.
+ Indent lists inside Div.
+ Support definitions with multiple descriptions.

* Powerpoint writer (Jesse Rosenthal):

+ Use table styles This will use the default table style in the
reference-doc file. As a result they will be easier when using
in a template, and match the color scheme.
+ Remove empty slides. Because of the way that slides were split, these
could be accidentally produced by comments after images. When animations
are added, there will be a way to add an empty slide with either
incremental lists or pauses.
+ Implement syntax highlighting. Note that background colors can't
be implemented in PowerPoint, so highlighting styles that require
these will be incomplete.
+ New test framework for pptx. We now compare the output of the
Powerpoint writer with files that we know to (a) not be corrupt,
and (b) to show the desired output behavior (details below).
+ Add `notesMaster` to `presentation.xml` if necessary.
+ Ignore links and (end)notes in speaker notes.
+ Output speaker notes.
+ Read speaker note templates conditionally. If there are speaker
notes in the presentation, we read in the notesMasters templates
from the reference pptx file.
+ Fix deletion track changes (#4303, Jesse Rosenthal).

* Markdown writer: properly escape @ to avoid capture as citation

* LaTeX writer:

+ Put hypertarget inside figure environment (#4388).
This works around a problem with the endfloat package and
makes pandoc's output compatible with it.
+ Fix image height with percentage (#4389). This previously caused
the image to be resized to a percentage of textwidth, rather than

* ConTeXt writer (Henri Menke):

+ New section syntax and support `--section-divs` (#2609).
`\section[my-header]{My Header}` ->
`\section[title={My Header},reference={my-header}]`.
The ConTeXt writer now supports the `--section-divs` option to
write sections in the fenced style, with `\startsection` and
+ xtables: correct wrong usage of caption (Henri Menke).

* Docx writer:

+ Fix image resizing with multiple images (#3930, Andrew Pritchard).
+ Use new golden framework (Jesse Rosenthal).
+ Make more deterministic to facilitate testing (Jesse Rosenthal).
- `getUniqueId` now calls to the state to get an incremented digit,
instead of calling to P.uniqueHash.
- we always start the PRNG in mkNumbering/mkAbstractNum with the same
seed (1848), so our randoms should be the same each time.
+ Fix ids in comment writing (Jesse Rosenthal). Comments from
`--track-changes=all` were producing corrupt docx, because the
writer was trying to get id from the `(ID,_,_)` field of
the attributes, and ignoring the "id" entry in the key-value pairs. We
now check both.

* Ms writer: Added papersize variable.

* TEI writer:

+ Use `height` instead of `depth` for images (#4331).
+ Ensure that id prefix is always used.
+ Don't emit `role` attribute; that was a leftover from the
Docbook writer.
+ Use 'xml:id', not 'id' attribute (#4371).

* AsciiDoc writer:

+ Do not output implicit heading IDs (#4363, Alexander
Krotov). Convert to `asciidoc-auto_identifiers` for old behaviour.

* RST writer:

+ Remove `blockToRST'` moving its logic into `fixBlocks`
(Francesco Occhipinti).
+ Insert comment between lists and quotes (#4248, Francesco Occchipinti).

* RST template: remove definition of 'math' role as raw.
This used to be needed prior to v 0.8 of docutils, but
now math support is built-in.

* Slides: Use divs to set incremental/non-incremental (#4381,
Jesse Rosenthal). The old method (list inside blockquote) still
works, but we are encouraging the use of divs with class
`incremental` or `nonincremental`.

* Text.Pandoc.ImageSize:

+ Make image size detection for PDFs more robust (#4322).
+ Determine image size for PDFs (#4322).
+ EMF Image size support (#4375, Andrew Pritchard).

* Text.Pandoc.Extensions:

+ Add `Ext_styles` (Jesse Rosenthal, API change). This will be used in
the docx reader (defaulting to off) to read pargraph and character
styles not understood by pandoc (as divs and spans, respectively).
+ Made `Ext_raw_html` default for `commonmark` format.

* Text.Pandoc.Parsing:

+ Export `manyUntil` (Alexander Krotov, API change).
+ Export improved `sepBy1` (Alexander Krotov).
+ Export list marker parsers: `upperRoman`, `lowerRoman`,
`decimal`, `lowerAlpha`, `upperAlpha` (Alexander Krotov, API change).

* Tests/Lua: fix tests on windows (Albert Krewinkel).

* Lua: register script name in global variable (#4393). The name of the Lua
script which is executed is made available in the global Lua variable
`PANDOC_SCRIPT_FILE`, both for Lua filters and custom writers.

* Tests: Abstract powerpoint tests out to OOXML tests (Jesse Rosenthal).
There is very little pptx-specific in these tests, so we abstract out
the basic testing function so it can be used for docx as well. This
should allow us to catch some errors in the docx writer that slipped
by the roundtrip testing.

* Lua filters: store constructors in registry (Albert Krewinkel). Lua
functions used to construct AST element values are stored in the Lua
registry for quicker access. Getting a value from the registry is much
faster than getting a global value (partly to idiosyncrasies of hslua);
this change results in a considerable performance boost.

* Documentation:

+ `doc/` Add draft of Org-mode documentation (Albert Krewinkel).
+ `doc/`: document global vars set for filters
(Albert Krewinkel).
+ mention Stack version. (#4343, Adam Brandizzi).
+ MANUAL: add documentation on custom styles (Jesse Rosenthal).
+ MANUAL.txt: Document incremental and nonincremental divs (Jesse
Rosenthal). Blockquoted lists are still described, but fenced divs are
presented in preference.
+ MANUAL.txt: document header and footer variables (newmana).
+ MANUAL.txt: self-contained implies standalone (#4304, Daniel Lublin).
+ label was renamed. (#4310, Alexander Brandizzi).

* Require tagsoup 0.14.3 (#4282), fixing HTML tokenization bug.

* Use latest texmath.

* Use latest pandoc-citeproc.

* Allow exceptions 0.9.

* Require aeson-pretty 0.8.5 (#4394).

* Bump blaze-markup, blaze-html lower bounds to 0.8, 0.9 (#4334).

* Update tagsoup to 0.14.6 (Alexander Krotov, #4282).

* Removed ghc-prof-options. As of cabal 1.24, sensible defaults are used.

* Update default.nix to current nixpkgs-unstable for hslua-0.9.5 (#4348,

pandoc (2.1.1)

* Markdown reader:

+ Don't coalesce adjacent raw LaTeX blocks if they are separated by a
blank line. See lierdakil/pandoc-crossref#160.
+ Improved `inlinesInBalancedBrackets` (#4272, jgm/pandoc-citeproc#315).
The change both improves performance and fixes a regression whereby
normal citations inside inline notes and figure captions were not
parsed correctly.

* RST reader:

+ Better handling for headers with an anchor (#4240). Instead of creating a
Div containing the header, we put the id directly on the header.
This way header promotion will work properly.
+ Add aligned environment when needed in math (#4254). ``
uses an `align*` environment for math in `.. math::` blocks, so this
math may contain line breaks. If it does, we put the math in an
`aligned` environment to simulate ``'s behavior.

* HTML reader:

+ Fix col width parsing for percentages < 10% (#4262, n3fariox).

* LaTeX reader:

+ Advance source position at end of stream.
+ Pass through macro defs in `rawLaTeXBlock` even if the `latex_macros`
extension is set (#4246). This reverts to earlier behavior and is
probably safer on the whole, since some macros only modify things in
included packages, which pandoc's macro expansion can't modify.
+ Fixed pos calculation in tokenizing escaped space.
+ Allow macro definitions inside macros (#4253). Previously we went into
an infinite loop with
+ Fix inconsistent column widths (#4238). This fixes a bug whereby column
widths for the body were different from widths for the header in some

* Docx reader (Jesse Rosenthal):

+ Parse hyperlinks in `instrText` tags (#3389, #4266). This was a form of
hyperlink found in older versions of word. The changes introduced for
this, though, create a framework for parsing further fields in MS Word
(see the spec, ECMA-376-1:2016, §17.16.5, for more on these fields).
We introduce a new module, `Text.Pandoc.Readers.Docx.Fields` which
contains a simple parsec parser. At the moment, only simple hyperlink
fields are accepted, but that can be extended in the future.

* Muse reader (Alexander Krotov):

+ Parse `~~` as non-breaking space in Text::Amuse mode.
+ Refactor list parsing.

* Powerpoint writer (Jesse Rosenthal):

+ Change reference to `notesSlide` to `endNotesSlide`.
+ Move image sizing into `picProps`.
+ Improve table placement.
+ Make our own `_rels/.rels` file.
+ Import reference-doc images properly.
+ Move `Presentation.hs` out of `PandocMonad`.
+ Refactor into separate modules. T.P.W.Powerpoint.Presentation
defines the Presentation datatype and goes Pandoc->Presentation;
T.P.W.Pandoc.Output goes Presentation->Archive.
Text.Pandoc.Writers.Powerpoint a thin wrapper around the two modules.
+ Avoid overlapping blocks in column output.
+ Position images correctly in two-column layout.
+ Make content shape retrieval environment-aware.
+ Improve image handling. We now determine image and caption placement
by getting the dimensions of the content box in a given layout.
This allows for images to be correctly sized and positioned in a
different template. Note that images without captions and headers are
no longer full-screened. We can't do this dependably in different
layouts, because we don't know where the header is (it could be to
the side of the content, for example).
+ Read presentation size from reference file. Our presentation size is
now dependent on the reference/template file we use.
+ Handle (sub)headers above slidelevel correctly. Above the slidelevel,
subheaders will be printed in bold and given a bit of extra space
before them. Note that at the moment, no distinction is made between
levels of headers above the slide header, though that can be changed.
+ Check for required files. Since we now import from reference/dist
file by glob, we need to make sure that we're getting the files we
need to make a non-corrupt Powerpoint. This performs that check.
+ Improve templating using `--reference-doc`. Templating should work
much more reliably now.
+ Include Notes slide in TOC.
+ Set notes slide header to slide-level.
+ Add table of contents. This is triggered by the `--toc` flag. Note
that in a long slide deck this risks overrunning the text box. The user
can address this by setting `--toc-depth=1`.
+ Set notes slide number correctly.
+ Clean up adding metadata slide. We want to count the slide numbers
correctly if it's in there.
+ Add anchor links. For anchor-type links (`[foo](#bar)`) we produce
an anchor link. In powerpoint these are links to slides, so we keep
track of a map relating anchors to the slides they occur on.
+ Make the slide number available to the blocks. For anchors,
block-processing functions need to know what slide number
they're in. We make the `envCurSlideId` available to blocks.
+ Move `curSlideId` to environment.
+ Allow setting `toc-title` in metadata.
+ Link notes to endnotes slide.

* Markdown writer:

+ Fix cell width calculation (#4265). Previously we could get
ever-lengthening cell widths when a table was run repeatedly through
`pandoc -f markdown -t markdown`.

* LaTeX writer:

+ Escape `&` in lstinline (Robert Schütz).

* ConTeXt writer:

+ Use xtables instead of Tables (#4223, Henri Menke).
Default to xtables for context output. Natural Tables are used
if the new `ntb` extension is set.

* HTML writer:

+ Fixed footnote backlinks with `--id-prefix` (#4235).

* `Text.Pandoc.Extensions`: Added `Ext_ntb` constructor (API change,
Henri Menke).

* `Text.Pandoc.ImageSize`: add derived `Eq` instance to `Dimension`
(Jesse Rosenthal, API change).

* Lua filters (Albert Krewinkel):

+ Make `PANDOC_READER_OPTIONS` available.
The options which were used to read the document are made available to
Lua filters via the `PANDOC_READER_OPTIONS` global.
+ Add lua module `pandoc.utils.run_json_filter`, which runs a JSON filter
on a Pandoc document.
+ Refactor filter-handling code into `Text.Pandoc.Filter.JSON`,
`Text.Pandoc.Filter.Lua`, and `Text.Pandoc.Filter.Path`.
+ Improve error messages. Provide more context about the task
which caused an error.

* data/pandoc.lua (Albert Krewinkel):

+ Accept singleton inline as a list. Every constructor which accepts a
list of inlines now also accepts a single inline element for
+ Accept single block as singleton list. Every constructor which accepts
a list of blocks now also accepts a single block element for
convenience. Furthermore, strings are accepted as shorthand for
`{pandoc.Str "text"}` in constructors.
+ Add attr, listAttributes accessors. Elements with
attributes got an additional `attr` accessor. Attributes were
accessible only via the `identifier`, `classes`, and `attributes`,
which was in conflict with the documentation, which indirectly states
that such elements have the an `attr` property.
+ Drop `_VERSION`. Having a `_VERSION` became superfluous, as this
module is closely tied to the pandoc version, which is available via
+ Fix access to Attr components. Accessing an Attr value (e.g.,
` Attr().classes`) was broken; the more common case of accessing it via
an Inline or Block element was unaffected by this.

* Move `metaValueToInlines` to from Docx writer to
`Text.Pandoc.Writers.Shared`, so it can be used by other writers
(Jesse Rosenthal).

* MANUAL.txt:

+ Clarify otherlangs in LaTeX (#4072).
+ Clarify `latex_macros` extension.
+ Recommend use of `raw_attribute` extension in header includes (#4253).

* Allow latest QuickCheck, tasty, criterion.

* Remove custom prelude and ghc 7.8 support.

* Reduce compiler noise (exact paths for compiled modules).

pandoc (2.1)

* Allow filters and lua filters to be interspersed (#4196). Previously
we ran all lua filters before JSON filters. Now we run filters in
the order they are presented on the command line, whether lua or JSON.
There are two incompatible API changes: The type of `applyFilters`
has changed, and `applyLuaFilters` has been removed. `Filter` is
also now exported.

* Use latest skylighting and omit the `missingIncludes` check, fixing
a major performance regression in earlier releases of the 2.x series
(#4226). Behavior change: If you use a custom syntax definition that
refers to a syntax you haven't loaded, pandoc will now complain when
it is highlighting the text, rather than doing a check at the start.
This change dramatically speeds up invocations of pandoc on short

* Text.Pandoc.Class: make `FileTree` opaque (don't export
`FileTree` constructor). This forces users to interact with it using
`insertInFileTree` and `getFileInfo`, which normalize file names.

* Markdown reader:

+ Rewrite `inlinesInBalancedBrackets`. The rewrite is much more
direct, avoiding `parseFromString`. And it performs significantly
better; unfortunately, parsing time still increases exponentially
(see #1735).
+ Avoid parsing raw tex unless `\` + letter seen. This seems to
help with the performance problem, #4216.

* LaTeX reader: Simplified a check for raw tex command.

* Muse reader (Alexander Krotov):

+ Enable round trip test (#4107).
+ Automatically translate `#cover` into `#cover-image`.
Amusewiki uses #cover directive to specify cover image.

* Docx reader (Jesse Rosenthal):

+ Allow for insertion/deletion of paragraphs (#3927).
If the paragraph has a deleted or inserted paragraph break (depending
on the track-changes setting) we hold onto it until the next
paragraph. This takes care of accept and reject. For this we introduce
a new state which holds the ils from the previous para if necessary.
For `--track-changes=all`, we add an empty span with class
`paragraph-insertion`/`paragraph-deletion` at the end of the paragraph
prior to the break to be inserted or deleted.
+ Remove unused anchors (#3679). Docx produces a lot of anchors with
nothing pointing to them---we now remove these to produce cleaner
output. Note that this has to occur at the end of the process
because it has to follow link/anchor rewriting.
+ Read multiple children of `w:sdtContents`.
+ Combine adjacent anchors. There isn't any reason to have numerous
anchors in the same place, since we can't maintain docx's
non-nesting overlapping. So we reduce to a single anchor.
+ Improved tests.

* Muse writer (Alexander Krotov): don't escape URIs from AST

* Docx writer:

+ Removed redundant subtitle in title (Sebastian Talmon).
+ `firstRow` table definition compatibility for Word 2016 (Sebastian
Talmon). Word 2016 seems to use a default value of "1" for table
headers, if there is no firstRow definition (although a default
value of 0 is documented), so all tables get the first Row formatted
as header. Setting the parameter to 0 if the table has no header
row fixes this for Word 2016
+ Fix custom styles with spaces in the name (#3290).

* Powerpoint writer (Jesse Rosenthal):

+ Ignore Notes div for parity with other slide outputs.
+ Set default slidelevel correctly. We had previously defaulted to
slideLevel 2. Now we use the correct behavior of defaulting to the
highest level header followed by content. We change an expected test
result to match this behavior.
+ Split blocks correctly for linked images.
+ Combine adjacent runs.
+ Make inline code inherit code size. Previously (a) the code size
wasn't set when we force size, and (b) the properties was set from
the default, instead of inheriting.
+ Simplify `replaceNamedChildren` function.
+ Allow linked images. The following markdown:
`[![Image Title](image.jpg)](`
will now produce a linked image in the resulting PowerPoint file.
+ Fix error with empty table cell. We require an empty `<a:p>` tag,
even if the cell contains no paragraphs---otherwise PowerPoint
complains of corruption.
+ Implement two-column slides. This uses the columns/column div
format described in the pandoc manual. At the moment, only two
columns (half the screen each) are allowed. Custom widths are not
+ Added more tests.

* OpenDocument/ODT writers: improved rendering of formulas (#4170, oltolm).

* Lua filters (Albert Krewinkel):

+ `data/pandoc.lua`: drop 'pandoc-api-version' from Pandoc objects
+ The current pandoc-types version is made available to Lua programs in
the global `PANDOC_API_VERSION`. It contains the version as a list of
+ The pandoc version available as a global `PANDOC_VERSION` (a list
of numbers).
+ `data/pandoc.lua`: make `Attr` an `AstElement`.
+ `data/pandoc.lua`: make all types subtypes of `AstElement`.
`Pandoc`, `Meta`, and `Citation` were just plain functions and did
not set a metatable on the returned value, which made it difficult
to amend objects of these types with new behavior. They are now
subtypes of AstElement, meaning that all their objects can gain
new features when a method is added to the behavior object
(e.g., `pandoc.Pandoc.behavior`).
+ `data/pandoc.lua`: split type and behavior tables. Clearly distinguish
between a type and the behavioral properties of an instance of that
type. The behavior of a type (and all its subtypes) can now be
amended by adding methods to that types `behavior` object, without
exposing the type objects internals. E.g.:
pandoc.Inline.behavior.frob = function () print'42' end
local str = pandoc.Str'hello'
str.frob() -- outputs '42'
+ `data/pandoc.lua`: fix Element inheritance. Extending all elements
of a given type (e.g., all inline elements) was difficult, as the
table used to lookup unknown methods would be reset every time a
new element of that type was created, preventing recursive property
lookup. This is was changed in that all methods and attributes of
supertypes are now available to their subtypes.
+ `data/pandoc.lua`: fix attribute names of Citation (#4222). The
fields were named like the Haskell fields, not like the documented,
shorter version. The names are changed to match the documentation
and Citations are given a shared metatable to enable simple
+ `data/pandoc.lua`: drop function `pandoc.global_filter`.
+ Bump `hslua` version to 0.9.5. This version fixes a bug that made it
difficult to handle failures while getting lists or a Map from Lua.
A bug in pandoc, which made it necessary to always pass a tag when
using MetaList or MetaBlock, is fixed as a result. Using the pandoc
module's constructor functions for these values is now optional
(if still recommended).
+ Stop exporting `pushPandocModule` (API change). The introduction
of `runPandocLua` renders direct use of this function obsolete.
+ Update generation of module docs for lua filters.
+ `Lua.Module.Utils`: make stringify work on `MetaValues` (John
MacFarlane). I'm sure this was intended in the first place,
but currently only `Meta` is supported.

* Improve benchmarks.

+ Set the default extensions properly.
+ Improve benchmark argument parsing. You can now say
`make bench BENCHARGS="markdown latex reader"` and both the
markdown and latex readers will be benchmarked.

* MANUAL.txt simplify and add more structure (Mauro Bieg).

* Generate from template and MANUAL.txt.
`make` will generate the after changes
to MANUAL.txt have been made.

* Update copyright notices to include 2018 (Albert Krewinkel).

pandoc (2.0.6)

* Added `jats` as an input format.

+ Add Text.Pandoc.Readers.JATS, exporting `readJATS` (API
change) (Hamish Mackenzie).
+ Improved citation handling in JATS reader. JATS citations
are now converted to pandoc citations, and JATS ref-lists
are converted into a `references` field in metadata, suitable
for use with pandoc-citeproc. Thus a JATS article with embedded
bibliographic information can be processed with pandoc and
pandoc-citeproc to produce a formatted bibliography.

* Allow `--list-extensions` to take an optional FORMAT argument.
This lists the extensions set by default for the selected FORMAT.
The extensions are now alphabetized, and the `+` or `-`
indicating the default setting comes before, rather than after,
the extension.

* Markdown reader:

+ Preserve original whitespace between blocks.
+ Recognize `\placeformula` as context.
+ Be pickier about table captions. A caption starts with a `:` which
can't be followed by punctuation. Otherwise we can falsely interpret
the start of a fenced div, or even a table header line like
`:--:|:--:`, as a caption.
+ Always use four space rule for example lists. It would be awkward
to indent example list contents to the first non-space character after
the label, since example list labels are often long. Thanks to
Bernhard Fisseni for the suggestion.
+ Improve raw tex parsing. Note that the Markdown reader is also
affected by the `latex_macros` extension changes described below
under the LaTeX reader.

* LaTeX reader:

+ `latex_macros` extension changes (#4179). Don't pass through macro
definitions themselves when `latex_macros` is set. The macros
have already been applied. If `latex_macros` is enabled, then
`rawLaTeXBlock` in Text.Pandoc.Readers.LaTeX will succeed in parsing
a macro definition, and will update pandoc's internal macro map
accordingly, but the empty string will be returned.
+ Export `tokenize`, `untokenize` (API change).
+ Use `applyMacros` in `rawLaTeXBlock`, `rawLaTeXInline`.
+ Refactored `inlineCommand`.
+ Fix bug in tokenizer. Material following `^^` was
dropped if it wasn't a character escape. This only affected
invalid LaTeX, so we didn't see it in the wild, but it appeared
in a QuickCheck test failure.
+ Fix regression in LateX tokenization (#4159). This mainly affects the
Markdown reader when parsing raw LaTeX with escaped spaces.
+ Add tests of LaTeX tokenizer.
+ Support `\foreignlanguage` from babel.
+ Be more tolerant of `&` character (#4208). This allows us to parse
unknown tabular environments as raw LaTeX.

* Muse reader (Alexander Krotov):

+ Parse anchors immediately after headings as IDs.
+ Require that note references does not start with 0.
+ Parse empty comments correctly.

* Org reader (Albert Krewinkel):

+ Fix asterisks-related parsing error (#4180).
+ Support minlevel option for includes (#4154). The level of headers
in included files can be shifted to a higher level by specifying a
minimum header level via the `:minlevel` parameter. E.g.
`#+include: "" :minlevel 1` will shift the headers in such that the topmost headers become level 1 headers.
+ Break-up org reader test file into multiple modules.

* OPML reader:

+ Enable raw HTML and other extensions by default for notes
(#4164). This fixes a regression in 2.0. Note that extensions can
now be individually disabled, e.g. `-f opml-smart-raw_html`.

* RST reader:

+ Allow empty list items (#4193).
+ More accurate parsing of references (#4156). Previously we erroneously
included the enclosing backticks in a reference ID (#4156). This
change also disables interpretation of syntax inside references, as
in docutils. So, there is no emphasis in `` `my *link*`_ ``.

* Docx reader:

+ Continue lists after interruption (#4025, Jesse Rosenthal).
Docx expects that lists will continue where they left off after an
interruption and introduces a new id if a list is starting again. So
we keep track of the state of lists and use them to define a "start"
attribute, if necessary.
+ Add tests for structured document tags unwrapping (Jesse Rosenthal).
+ Preprocess Document body to unwrap `w:sdt` elements (Jesse Rosenthal,

* Plain writer:

+ Don't linkify table of contents.

* RST writer:

+ Fix anchors for headers (#4188). We were missing an `_`.

* PowerPoint writer (Jesse Rosenthal):

+ Treat lists inside BlockQuotes as lists. We don't yet produce
incremental lists in PowerPoint, but we should at least treat lists
inside BlockQuotes as lists, for compatibility with other slide formats.
+ Add ability to force size. This replaces the more specific
`blockQuote runProp`, which only affected the size of blockquotes. We
can use this for notes, etc.
+ Implement notes. This currently prints all notes on a final slide.
Note that at the moment, there is a danger of text overflowing the
note slide, since there is no logic for adding further slides.
+ Implement basic definition list functionality to PowerPoint writer.
+ Don't look for default template file for Powerpoint (#4181).
+ Add pptx to isTextFormat list. This is used to check standalone
and not writing to the terminal.
+ Obey slide level option (Jesse Rosenthal).
+ Introduce tests.

* Docx writer:

+ Ensure that `distArchive` is the one that comes with pandoc
(#4182). Previously a `reference.docx` in `~/.pandoc` (or the user data
dir) would be used instead, and this could cause problems because a
user-modified docx sometimes lacks vital sections that we count
on the `distArchive` to supply.

* Org writer:

+ Do not wrap "-" to avoid accidental bullet lists (Alexander Krotov).
+ Don't allow fn refs to wrap to beginning of line (#4171, with help from
Alexander Krotov). Otherwise they can be interpreted as footnote

* Muse writer (Alexander Krotov):

+ Don't wrap note references to the next line (#4172).

* HTML writer:

+ Use br elements in line blocks instead of relying on CSS
(#4162). HTML-based templates have had the custom CSS for
`div.line-block` removed. Those maintaining custom templates will want
to remove this too. We still enclose line blocks in a div with class

* LaTeX writer:

+ Use `\renewcommand` for `\textlatin` with babel (#4161).
This avoids a clash with a deprecated `\textlatin` command defined
in Babel.
+ Allow fragile=singleslide attribute in beamer slides (#4169).
+ Use `\endhead` after `\toprule` in headerless tables (#4207).

* FB2 writer:

+ Add cover image specified by `cover-image` meta (Alexander Krotov,

* JATS writer (Hamish Mackenzie):

+ Support writing `<fig>` and `<table-wrap>` elements
with `<title>` and `<caption>` inside them by using Divs with class set
to one of `fig`, `table-wrap` or `caption` (Hamish Mackenzie). The
title is included as a Heading so the constraint on where Heading can
occur is also relaxed.
+ Leave out empty alt attributes on links.
+ Deduplicate image mime type code.
+ Make `<p>` optional in `<td>` and `<th>` (#4178).
+ Self closing tags for empty xref (#4187).
+ Improve support for code language.

* Custom writer:

+ Use init file to setup Lua interpreter (Albert Krewinkel).
The same init file (`data/init`) that is used to setup the Lua
interpreter for Lua filters is also used to setup the interpreter of
custom writers.lua.
+ Define instances for newtype wrapper (Albert Krewinkel). The custom
writer used its own `ToLuaStack` instance definitions, which made
it difficult to share code with Lua filters, as this could result
in conflicting instances. A `Stringify` wrapper is introduced to
avoid this problem.
+ Added tests for custom writer.
+ Fixed definition lists and tables in `data/sample.lua`.

* Fixed regression: when target is PDF, writer extensions were being
ignored. So, for example, `pandoc -t latex-smart -o file.pdf`
did not work properly.

* Lua modules (Albert Krewinkel):

+ Add `pandoc.utils` module, to hold utility functions.
+ Create a Haskell module Text.Pandoc.Lua.Module.Pandoc to
define the `pandoc` lua module.
+ Make a Haskell module for each Lua module. Move definitions for the
`pandoc.mediabag` modules to a separate Haskell module.
+ Move `sha1` from the main `pandoc` module to `pandoc.utils`.
+ Add function `pandoc.utils.hierarchicalize` (convert list of
Pandoc blocks into (hierarchical) list of Elements).
+ Add function `pandoc.utils.normalize_date` (parses a date and
converts it (if possible) to "YYYY-MM-DD" format).
+ Add function `pandoc.utils.to_roman_numeral` (allows conversion
of numbers below 4000 into roman numerals).
+ Add function `pandoc.utils.stringify` (converts any AST element
to a string with formatting removed).
+ `data/init.lua`: load `pandoc.utils` by default
+ Turn pipe, read into full Haskell functions. The `pipe` and `read`
utility functions are converted from hybrid lua/haskell functions
into full Haskell functions. This avoids the need for intermediate
`_pipe`/`_read` helper functions, which have dropped.
+ pandoc.lua: re-add missing MetaMap function. This was a bug
introduced in version 2.0.4.

* Text.Pandoc.Class: Add `insertInFileTree` [API change]. This gives
a pure way to insert an ersatz file into a `FileTree`. In addition, we
normalize paths both on insertion and on lookup.

* Text.Pandoc.Shared: export `blocksToInlines'` (API change, Maura Bieg).

* Text.Pandoc.MIME: Add opus to MIME type table as audio/ogg (#4198).

* Text.Pandoc.Extensions: Alphabetical order constructors for
`Extension`. This makes them appear in order in `--list-extensions`.

* Allow lenient decoding of latex error logs, which are not always
properly UTF8-encoded (#4200).

* Update latex template to work with recent versions of beamer.
The old template produced numbered sections with some recent
versions of beamer. Thanks to Thomas Hodgson.

* Updated reference.docx (#4175). Instead of just "Hello, world", the
document now contains exemplars of most of the styles that have an
effect on pandoc documents. This makes it easier to see the effect
of style changes.

* Removed `default.theme` data file (#4096). It is no longer needed now
that we have `--print-highlight-style`.

* Added `stack.lts9.yaml` for building with lts 9 and ghc 8.0.2.
We still need this for the alpine static linux build, since
we don't have ghc 8.2.2 for that yet.

* Removed `stack.pkg.yaml`. We only really need `stack.yaml`; we
can put flag settings for pandoc-citeproc there.

* Makefile: Add 'trypandoc' and 'pandoc-templates' targets to
make releases easier.

* MANUAL.txt:

+ Add note on what formats have `+smart` by default.
+ Use native syntax for custom-style (#4174, Mauro Bieg).
+ Introduce dedicated Extensions section, since some extensions
affect formats other than markdown (Mauro Bieg, #4204).
+ Clarify default html output for `--section-divs` (Richard Edwards).

* say that Text.Pandoc.JSON comes form pandoc-types.
Closes jgm/pandoc-website#16.

* Delete removed `-S` option from command (#4151, Georger Araújo).

pandoc (2.0.5)

* Fix a bug in 2.0.4, whereby pandoc could not read the theme files
generated with `--print-highlight-style` (#4133). Improve JSON
serialization of styles.

* Fix CSS issues involving line numbers (#4128).
Highlighted code blocks are now enclosed in a div with class `sourceCode`.
Highlighting CSS no longer sets a generic color for pre and code; we only
set these for class `sourceCode`.

* `--pdf-engine-opt`: fix bug where option order was reversed (#4137).

* Add PowerPoint (pptx) writer (Jesse Rosenthal).
It works following the standard Pandoc conventions for making other
sorts of slides. Caveats:

+ Syntax highlighting is not yet implemented. (This is difficult
because there are no character classes in Powerpoint.)
+ Footnotes and Definition lists are not yet implemented. (Notes will
usually take the form of a final slide.
+ Image placement and auto-resizing has a few glitches.
+ Reference powerpoint files don't work dependably from the command
line. This will be implemented, but at the moment users are advised
to change themes from within Powerpoint.

* Create shared Text.Pandoc.Writers.OOXML module (Jesse Rosenthal).
This is for functions used by both Powerpoint and Docx writers.

* Add default pptx data for Powerpoint writer (Jesse Rosenthal).

* Add `empty_paragraphs` extension.

+ Deprecate `--strip-empty-paragraphs` option. Instead we now
use an `empty_paragraphs` extension that can be enabled on
the reader or writer. By default, disabled.
+ Add `Ext_empty_paragraphs` constructor to `Extension`.
+ Revert "Docx reader: don't strip out empty paragraphs."
This reverts commit d6c58eb836f033a48955796de4d9ffb3b30e297b.
+ Implement `empty_paragraphs` extension in docx reader and writer,
opendocument writer, HTML reader and writer.
+ Add tests for `empty_paragraphs` extension.

* Markdown reader:

+ Don't parse native div as table caption (#4119).
+ Improved computation of column widths in pipe tables.
Pipe tables with lines longer than the text width (as set
by `--columns`) are now scaled to text width, with the relative
widths of columns determined by the ratios between the
header lines. Previously we computed column widths using
the ratio of header line lengths to column width, so that
tables with narrow header lines were extremely thin, which
was very rarely the desired result.

* LaTeX reader: fix `\` before newline (#4134). This should be a space,
as long as it's not followed by a blank line. This has been fixed at the
tokenizer level.

* Muse reader (Alexander Krotov):

+ Add test for `#disable-tables` directive in Emacs mode.
+ Don't allow emphasis to be preceded by letter.
+ Add underline support in Emacs Muse mode..
+ Support multiline directives in Amusewiki mode

* Man writer: omit internal links (#4136). That is, just print the link
text without the URL.

* Markdown reader: accept processing instructions as raw HTML (#4125).

* Lua filters (Albert Krewinkel):

+ Use script to initialize the interpreter. The file `init.lua` is
used to initialize the Lua interpreter which is used in Lua filters.
This gives users the option to require libraries which they want to
use in all of their filters, and to extend default modules.
+ Fix package loading for Lua 5.1. The list of package searchers is
named `package.loaders` in Lua 5.1 and LuaJIT, and `package.searchers`
in Lua 5.2 and later.
+ Refactor lua module handling. The integration with Lua's package/module
system is improved: A pandoc-specific package searcher is prepended to
the searchers in `package.searchers`. The modules `pandoc` and
`pandoc.mediabag` can now be loaded via `require`.
+ Bump lower bound of hslua. The release hslua 0.9.3 contains a new
function which makes using Haskell functions as package loaders much

* reveal.js template: add title-slide identifier to title slide (#4120).
This allows it to be styled more easily.

* LaTeX template: Added support for `pagestyle` variable (#4135,
Thomas Hodgson)

* Add `-threaded` to ghc-options for executable (#4130, fixes a build
error on linux).

pandoc (2.0.4)

* Add `--print-highlight-style` option. This generates a JSON version
of a highlighting style, which can be saved as a `.theme` file, modified,
and used with `--highlight-style` (#4106, #4096).

* Add `--strip-empty-paragraphs` option. This works for any input format.
It is primarily intended for use with docx and odt documents where
empty paragraphs have been used for inter-paragraph spaces.

* Support `--webtex` for `gfm` output.

* Recognize `.muse` file extension.

* Support beamer `\alert` in LaTeX reader. Closes #4091.

* Docx reader: don't strip out empty paragraphs (#2252).
Users who have a conversion pipeline from docx may want to consider adding
`--strip-empty-paragraphs` to the command line.

* Org reader (Albert Krewinkel): Allow empty list items (#4090).

* Muse reader (Alexander Krotov):

+ Parse markup in definition list terms.
+ Allow definition to end with EOF.
+ Make code blocks round trip.
+ Drop common space prefix from list items.
+ Add partial round trip test.
+ Don't interpret XML entities.
+ Remove `nested`.
+ Parse `~~` as non-breaking space in Emacs mode.
+ Correctly remove indentation from notes. Exactly one space is
required and considered to be part of the marker.
+ Allow list items to be empty.
+ Add ordered list test.
+ Add more multiline definition tests.
+ Don't allow blockquotes within lists.
+ Fix reading of multiline definitions.
+ Add inline `<literal>` support.
+ Concatenate inlines of the same type

* Docx writer: allow empty paragraphs (#2252).

* CommonMark/gfm writer:

+ Use raw html for native divs/spans (#4113). This allows a pandoc
markdown native div or span to be rendered in gfm using raw html tags.
+ Implement `raw_html` and `raw_tex` extensions. Note that `raw_html`
is enabled by default for `gfm`, while `raw_tex` is disabled by default.

* Muse writer (Alexander Krotov):

+ Test that inline math conversion result is normalized.
Without normalization this test produced
+ Improve inline list normalization and move to writer.
+ Escape hash symbol.
+ Escape `----` to avoid accidental horizontal rules.
+ Escape only `</code>` inside code tag.
+ Additional `<verbatim>` is not needed as `<code>` is verbatim already.

* LaTeX writer:

+ Allow specifying just width or height for image size.
Previously both needed to be specified (unless the image was
being resized to be smaller than its original size).
If height but not width is specified, we now set width to
textwidth. If width but not height is specified, we now set
height to textheight. Since we have `keepaspectratio`, this
yields the desired result.
+ Escape `~` and `_` in code with `--listings` (#4111).

* HTML writer: export `tagWithAttributes`. This is a helper allowing
other writers to create single HTML tags.

* Let papersizes `a0`, `a1`, `a2`, ... be case-insensitive by
converting the case as needed in LaTeX and ConTeXt writers.

* Change `fixDisplayMath` from `Text.Pandoc.Writers.Shared`
so that it no longer produces empty `Para`'s as an artifact.

* `Text.Pandoc.Shared.blocksToInlines`: rewrote using builder.
This gives us automatic normalization, so we don't get
for example two consecutive Spaces.

* Include default CSS for 'underline' class in HTML-based templates.

* revealjs template: add `tex2jax` configuration for the
math plugin. With the next release of reveal.js, this will
fix the problem of `$`s outside of math contexts being
interpreted as math delimiters (#4027).

* `pandoc.lua` module for use in lua filters (Albert Krewinkel):

+ Add basic lua List module (#4099, #4081). The List module is
automatically loaded, but not assigned to a global variable. It can be
included in filters by calling `List = require 'List'`. Lists of blocks,
lists of inlines, and lists of classes are now given `List` as a metatable,
making working with them more convenient. E.g., it is now possible to
concatenate lists of inlines using Lua's concatenation operator `..`
(requires at least one of the operants to have `List` as a metatable):

function Emph (emph)
local s = {pandoc.Space(), pandoc.Str 'emphasized'}
return pandoc.Span(emph.content .. s)

The `List` metatable is assigned to the tables which get passed to
the constructors `MetaBlocks`, `MetaInline`, and `MetaList`. This
enables the use of the resulting objects as lists.
+ `Lua/StackInstances`: push Pandoc and Meta via constructor.
Pandoc and Meta elements are now pushed by calling the respective
constructor functions of the pandoc Lua module. This makes serialization
consistent with the way blocks and inlines are pushed to lua and allows
to use List methods with the `blocks` value.
+ Add documentation for pandoc.List in ``.

* Use latest tagsoup. This fixes a bug in parsing HTML tags with
`&` (but not a valid entity) following them (#4094, #4088).

* Use skylighting, fixing the color of unmarked code text
when `numberLines` is used (#4103).

* Make `normalizeDate` more forgiving (Mauro Bieg, #4101), not
requiring a leading 0 on single-digit days.

* Fix `--help` output for `--highlight-style` to include `FILE` (Mauro
Bieg, #4095).

* Clearer deprecation warning for `--latexmathml, --asciimathml, -m`.
Previously we only mentioned `--latexmathml`, even if `-m` was

* Changelog: fix description of lua filters in 2.0 release
(Albert Krewinkel). Lua filters were initially run *after* conventional
(JSON) filters. However, this was changed later to make it easier to deal
with files in the mediabag. The changelog is updated to describe that
feature of the 2.0 release correctly.

* Change Generic JSON instances to TemplateHaskell (Jasper Van der Jeugt,
#4085). This reduces compile time and memory usage significantly.

* ``: Added tikz filter example.

* Create alternative zip file for macOS binaries.

* Create alternative zip file for Windows binaries.

* Update since we now provide zips for binaries.

* Relax `http-types` dependency (Justus Sagemüller, #4084).

* Add ``, `` to docs. These used to live in
the website repo.

* Add `packages` target to Makefile.

* Bump bounds for binary, http-types, tasty-hunit

pandoc (2.0.3)

* Lua filters: preload text module (Albert Krewinkel, #4077).
The `text` module is preloaded in lua. The module contains some UTF-8
aware string functions, implemented in Haskell. The module is loaded on
request only, e.g.:

text = require 'text'
function Str (s)
s.text = text.upper(s.text)
return s

* Allow table-like access to attributes in lua filters (Albert Krewinkel,
#4071). Attribute lists are represented as associative lists in Lua. Pure
associative lists are awkward to work with. A metatable is attached to
attribute lists, allowing to access and use the associative list as if
the attributes were stored in as normal key-value pair in table.
Note that this changes the way `pairs` works on attribute lists. Instead
of producing integer keys and two-element tables, the resulting iterator
function now returns the key and value of those pairs. Use `ipairs` to
get the old behavior. Warning: the new iteration mechanism only works if
pandoc has been compiled with Lua 5.2 or later (current default: 5.3).

* Text.Pandoc.Parsing.uri: allow `&` and `=` as word characters (#4068).
This fixes a bug where pandoc would stop parsing a URI with an
empty attribute: for example, `&a=&b=` wolud stop at `a`.
(The uri parser tries to guess which punctuation characters
are part of the URI and which might be punctuation after it.)

* Introduce `HasSyntaxExtensions` typeclass (Alexander Krotov, #4074).

+ Added new `HasSyntaxExtensions` typeclass for `ReaderOptions` and
+ Reimplemented `isEnabled` function from `Options.hs` to accept both
`ReaderOptions` and `WriterOptions`.
+ Replaced `enabled` from `CommonMark.hs` with new `isEnabled`.

* Add `amuse` extension (Alexander Krotov) to enable Amuse wiki
behavior for `muse`. New `Ext_amuse` constructor for
`Extension`. Note: this is switched on by default; for
Emacs behavior, use `muse-amuse`.

* Muse reader (Alexander Krotov):

+ Count only one space as part of list item marker.
+ Produce SoftBreaks on newlines. Now wrapping can be preserved
with `--wrap=preserve`.
+ Add Text::Amuse footnote extensions. Footnote end is indicated by
indentation, so footnotes can be placed anywhere in the text,
not just at the end of it.
+ Accept Emacs Muse definition lists when `-amuse`.
Emacs Muse does not require indentation.

* HTML reader:

+ Ensure we don't produce level 0 headers (#4076), even for chapter
sections in epubs. This causes problems because writers aren't set
up to expect these.
+ Allow spaces after `\(` and before `\)` with `tex_math_single_backslash`.
Previously `\( \frac{1}{a} < \frac{1}{b} \)` was not parsed as math in
`markdown` or `html` `+tex_math_single_backslash`.
+ Parse div with class `line-block` as LineBlock.
+ Don't fail with block-level content in figcaption (Mauro Bieg, #4183).

* MANUAL: clarify that math extensions work with HTML.
Clarify that `tex_math_dollars` and `tex_math_single_backslash`
will work with HTML as well as Markdown.

* Creole reader: Fix performance issue for longer lists (Sascha Wilde,

* RST reader: better support for 'container' directive (#4066).
Create a div, incorporate name attribute and classes.

* LaTeX reader:

+ Support column specs like `*{2}{r}` (#4056). This is equivalent to
`rr`. We now expand it like a macro.
+ Allow optional args for parbox (#4056).
+ Allow optional arguments on `\footnote` (#4062).

* EPUB writer: Fixed path for cover image (#4069). It was previously
`media/media/imagename`, and should have been `media/imagename`.

* Markdown writer: fix bug with doubled footnotes in grid tables

* LaTeX template: include natbib/biblatex after polyglossia (#4073).
Otherwise we seem to get an error; biblatex wants polyglossia
language to be defined.

* Added examples to lua filters documentation.

pandoc (2.0.2)

* Deprecated ancient HTML math methods: `--latexmathml`, `--gladtex`,
`--mimetex`, `--jsmath`.

* Fixed URIs in `data/jats.csl`. They were being rendered twice,
leading to invalid XML in default JATS output with pandoc-citeproc.

* ``: use real-world man page filter as example.

* Add lua filter functions `walk_inline` and `walk_block`
in the pandoc module, to apply filters inside particular
inline and block elements.

* Refactored some code from `Text.Pandoc.Lua.PandocModule`
into new internal module `Text.Pandoc.Lua.Filter`.

* Markdown reader:

+ Allow fenced code blocks to be indented 1-3 spaces (#4011).
This brings our handling of them into alignment with CommonMark's.
+ Fix YAML metadata with "chomp" (`|-`). Previously if a
YAML block under `|-` contained a blank line, pandoc would
not parse it as metadata.

* Removed `etc.` from abbreviations file. Often `etc.` ends a
sentence, and we want the period to be treated as a
sentence-ending period.

* Fix regression with `--metadata` (#4054). Values specified with
`--metadata` should replace a metadata value set in the document
itself, rather than creating a list including a new value.

* EPUB writer:

+ Fix EPUB OCF structure. #3720 had been improperly implemented.
+ Fix modified paths for raw HTML tags (src, poster, etc.)
(#4050, #4055). This had not been updated for the new EPUB
container layout, with a separate text/ subdirectory.
+ Fix image paths with empty `--epub-subdirectory`.

* Miscellaneous code cleanup (Alexander Krotov).

* Use pandoc-types 1.17.3, which adds `Walkable` instances
for `[Block] Block` and `[Inline] Inline`.

* Remove obsolete `stack.full.yaml` (#4052).

* Change to using pandoc-citeproc 0.12.1 in binary packages.

* Consolidate math output method documentation (#4049, Mauro Bieg).

* `MANUAL.txt`: fix header level of "Extension: emoji" (Albert Krewinkel).

* Use lua filter to generate man page from `MANUAL.txt`, replacing old
Haskell filters. This is easier and faster.

* Improved ``.

* Update commands to extract deb archive on Linux (#4043, Salim B).

pandoc (

* Improved fix to #3989 (parsing of HTML tags containing
`>` in an attribute or comment). The previous fix (in 2.0.1) only
worked in certain cases.

* FB2 writer (Alexander Krotov):

+ Add `unrecognised` genre to `<title-info>`
(Alexander Krotov). XML schema requires at least one genre.
+ Remove `<annotation>` from `<body>`.

* CommonMark writer: fix strikethrough for `gfm` (#4038).

* Use texmath 0.10, which adds support for a wider range of
symbols and fixes default column alignments in MathML
and OMML.

* Highlighting fixes, using skylighting

+ Fix invalid CSS.
+ Support `lineAnchors` (or `line-anchors`) in HTML code blocks.
+ Ensure that code lines don't get duplicate identifiers (#4031).
The line identifiers are built using the code block's identifier
as a prefix. If the code block has null identifier, we use
`cb1`, `cb2`, etc.

* Added a few abbreviations to `data/abbreviations`,
and sorted the list (#3984, Wandmalfarbe).

* Improved support for columns in HTML writer (#4028).

+ Remove `width` attribute from the `div`.
+ Remove space between `<div class="column">` elements,
since this prevents columns whose widths sum to 100%
(the space takes up space).
+ Move as much as possible of the CSS to the template.
+ Ensure that all the HTML-based templates (including epub)
contain the CSS for columns.
+ Columns default to 50% width unless they are given a width
attribute. So if you want two equal-width columns, you
can use a div with class `column` and no `width` attribute.

* SelfContained: use `base64` for css links with media attribute (#4026).
This fixes `--self-contained` with S5.

* Improve `pandoc-template-mode.el` (Vaclav Haisman).

* Issue INFO, not WARNING, when a .sty file cannot be
read in LaTeX reader. It is normally not an issue requiring
a fix from the user if .sty files are not found.

* MacOS instructions needed xar -f (adam234).

* MANUAL.txt:

+ Clarify that --setext-headers doesn't affect gfm output (#4035).
+ Clarify what is needed to open and close a div in `fenced_divs`
(#4039, Tristano Ajmone).
+ Removed reference to `default.beamer` in docs (#4024).
Also added mention of other templates affecting PDF output
with different settings.

pandoc (2.0.1)

* Fixed regression in parsing of HTML comments in markdown and other
non-HTML formats (`Text.Pandoc.Readers.HTML.htmlTag`) (#4019).
The parser stopped at the first `>` character, even if it wasn't
the end of the comment.

* Creole reader (Sascha Wilde):

+ Fix some minor typos and formatting.
+ Add additional test on nowiki-block after para.
+ Fix lists with trailing white space.

* LaTeX reader: handle `%` comment right after command.
For example, `\emph%`.

* Markdown reader: make sure fenced div closers work in lists.
Previously the following failed:

::: {.class}
1. one
2. two

and you needed a blank line before the closing `:::`.

* Make `fenced_divs` affect the Markdown writer. If `fenced_divs` is
enabled, Divs will be rendered as fenced divs.

* LaTeX/Beamer writer: support "blocks" inside columns and other Divs

* HTML Writer: consistently use dashed class-names (Mauro Bieg, #3556).
Note: this change may require some changes in CSS rules.
`footnoteRef` has become `footnote-ref`, `titleslide` has
become `title-slide`, and `footnoteBack` has become `footnote-back`.

* JATS writer: Properly pass through author metadata (#4020).

* FB2 writer (Alexander Krotov):

+ Write blocks outside of `<p>` in definitions.
+ Make bullet lists consistent with ordered lists, repeating
the marker for the outer list rather than indenting sublists,
since indentation does not work in readers.
+ Add new style FB2 tests.

* `Text.Pandoc.ImageSize`: Add `Millimeter` constructor to `Dimension`
(#4012) [API change]. Now sizes given in 'mm' are no longer converted
to 'cm'.

* Revise documentation of small caps syntax (Andrew Dunning, #4013).

* Fix broken reference links in manual (Andrew Dunning, #4014)

* Fixed example of slide columns structure in changelog (#4015).
Also documented this feature in MANUAL.txt.

pandoc (

* EPUB writer:

+ Fixed filepaths for nonstandard epub-subdirectory values.
+ Ensure that epub2 is recognized as a non-text format,
so that a template is used.
+ Don't include "prefix" attribute for ibooks for epub2.
It doesn't validate.
+ Fix stylesheet paths; previously we had an incorrect
stylesheet path for the cover page and nav page.

* LaTeX reader:

+ Insert space when needed in macro expansion (#4007).
Sometimes we need to insert a space after a control sequence
to prevent it merging with a following letter.
+ Allow unbraced arguments for macros (#4007).
+ Allow body of macro definition to be unbraced (#4007).

* Linux package build: ensure that pandoc-citeproc is statically linked.

* trypandoc: add native, ms.

pandoc (2.0)

[new features]

* New output format `ms` (groff ms). Complete support, including
tables, math, syntax highlighting, and PDF bookmarks. The writer uses
texmath's new eqn writer to convert math to eqn format, so a ms file
produced with this writer should be processed with `groff -ms -e` if
it contains math.

* New output format `jats` (Journal Article Tag Suite). This is an XML
format used in archiving and publishing articles. Note that a
URI-encoded CSL stylesheet (`data/jats.csl`) is added automatically
unless a stylesheet is specified using `--css`.

* New output format `gfm` (GitHub-flavored CommonMark) (#3841).
This uses bindings to GitHub's fork of cmark, so it should parse
gfm exactly as GitHub does (excepting certain postprocessing
steps, involving notifications, emojis, etc.). `markdown_github`
has been deprecated in favor of `gfm`.

* New output format `muse` (Emacs Muse) (Alexander Krotov, #3489).

* New input format `gfm` (GitHub-flavored CommonMark) (#3841).
This uses bindings to GitHub's fork of cmark. `markdown_github`
has been deprecated in favor of `gfm`.

* New input format `muse` (Emacs Muse) reader (Alexander Krotov, #3620).

* New input format `tikiwiki` (TikiWiki markup) (rlpowell, #3800).

* New input format `vimwiki` (Vimwiki markup) (Yuchen Pei, #3705).
Note that there is a new data file, `data/vimwiki.css`, which can
be used to display the HTML produced by this reader and
pandoc's HTML writer in the style of vimwiki's own HTML

* New input format `creole` (Creole 1.0) (#3994, Sascha Wilde).

* New syntax for Divs, with `fenced_divs` extension enabled by
default (#168). This gives an attractive, plain-text way to create
containers for block-level content.

* Added new syntax for including raw content in any output format,
enabled by the `raw_attribute` extension (which is on by default
for `markdown` and `multimarkdown`). The syntax is the same as
for fenced code blocks or code inlines, only with `{=FORMAT}` for
attributes, where `FORMAT` is the name of the output format
(e.g., `ms`, `html`).

* Implement multicolumn support for slide formats (#1710).
The structure expected is:

:::::::::::::: {.columns}
::: {.column width="40%"}
::: {.column width="60%"}

Support has been added for beamer and all HTML slide formats.

* Allows line comments in templates, beginning with `$--` (#3806).
(Requires doctemplates 0.2.1.)

* Add `--eol=crlf|lf|native` flag and writer option to control line endings
(Stefan Dresselhaus, #3663, #2097).

* Add `--log` option to save log messages in JSON format to a file (#3392).

* Add `--request-header` option, to set request headers when pandoc
makes HTTP requests to fetch external resources. For example:
`--request-header User-Agent:blah`.

* Added lua filters (Albert Krewinkel, #3514). The new `--lua-filter`
option works like `--filter` but takes pathnames of special lua filters
and uses the lua interpreter baked into pandoc, so that no external
interpreter is needed. Note that lua filters are all applied before
regular filters, regardless of their position on the command line.
For documentation of lua filters, see `doc/`.

* Set `PANDOC_READER_OPTIONS` in environment where filters are run.
This contains a JSON representation of `ReaderOptions`, so filters
can access it.

* Support creation of pdf via groff `ms` and pdfroff.
`pandoc -t ms -o output.pdf input.txt`.

* Support for PDF generation via HTML and `weasyprint` or `prince`
(Mauro Bieg, #3909). `pandoc -t html5 -o output.pdf --pdf-engine=prince`.

* Added `--epub-subdirectory` option (#3720). This specifies the
subdirectory in the OCF container that holds the EPUB specific content.
We now put all EPUB related content in an `EPUB/` subdirectory by default
(later this will be configurable).

EPUB/ <<--configurable-->>
fonts/ <<--static-->>
media/ <<--static-->>
styles/ <<--static-->>
text/ <<--static-->>

* Added `--resource-path=SEARCHPATH` command line option (#852).
SEARCHPATH is separated by the usual character, depending on OS
(: on unix, ; on windows). Default resource path is just working
directory. However, the working directory must be explicitly
specified if the `--resource-path` option is used.

* Added `--abbreviations=FILE` option for custom abbreviations file
(#256). Default abbreviations file (`data/abbreviations`) contains
a list of strings that will be recognized by pandoc's
Markdown parser as abbreviations. (A nonbreaking space will
be inserted after the period, preventing a sentence space in
formats like LaTeX.) Users can override the default by putting a file
abbreviations in their user data directory (`~/.pandoc` on *nix).

* Allow a theme file as argument to `--highlight-style`.
Also include a sample, `default.theme`, in `data/`.

* Allow `--syntax-definition` option for dynamic loading of syntax
highlighting definitions (#3334).

* Lists in `markdown` by default now use the CommonMark variable
nesting rules (#3511). The indentation required for a block-level
item to be included in a list item is no longer fixed, but is
determined by the first line of the list item. To be included in
the list item, a block must be indented to the level of the first
non-space content after the list marker. Exception: if are 5 or more
spaces after the list marker, then the content is interpreted as an
indented code block, and continuation paragraphs must be indented
two spaces beyond the end of the list marker. See the CommonMark
spec for more details and examples.

Documents that adhere to the four-space rule should, in most cases,
be parsed the same way by the new rules. Here are some examples
of texts that will be parsed differently:

- a
- b

will be parsed as a list item with a sublist; under the four-space
rule, it would be a list with two items.

- a


Here we have an indented code block under the list item, even though it
is only indented six spaces from the margin, because it is four spaces
past the point where a continuation paragraph could begin. With the
four-space rule, this would be a regular paragraph rather than a code

- a


Here the code block will start with two spaces, whereas under
the four-space rule, it would start with `code`. With the four-space
rule, indented code under a list item always must be indented eight
spaces from the margin, while the new rules require only that it
be indented four spaces from the beginning of the first non-space
text after the list marker (here, `a`).

This change was motivated by a slew of bug reports from people
who expected lists to work differently (#3125, #2367, #2575, #2210,
#1990, #1137, #744, #172, #137, #128) and by the growing prevalance
of CommonMark (now used by GitHub, for example). Those who
prefer the old behavior can use `-f markdown+four_space_rule`.

* Added `four_space_rule` extension. This triggers the old pandoc
parsing rule for content nested under list items (the "four space

* Added `spaced_reference_links` extension (#2602). It allows whitespace
between the two parts of a reference link: e.g.

[a] [b]

[b]: url

This was previously enabled by default; it is now forbidden by default.

* Add `space_in_atx_header` extension (#3512). This is enabled by default
in pandoc and GitHub markdown but not the other flavors.
This requires a space between the opening #'s and the header
text in ATX headers (as CommonMark does but many other implementations
do not). This is desirable to avoid falsely capturing things like




* Add `sourcefile` and `outputfile` template variables (Roland Hieber,

* Allow ibooks-specific metadata in epubs (#2693). You can now have
the following fields in your YAML metadata, and it will be treated
appropriately in the generated EPUB:

version: 1.3.4
specified-fonts: false
ipad-orientation-lock: portrait-only
iphone-orientation-lock: landscape-only
binding: true
scroll-axis: vertical

[behavior changes]

* Reader functions no longer presuppose that CRs have been
stripped from the input. (They strip CRs themselves, before
parsing, to simplify the parsers.)

* Added support for translations (localization) (#3559).
Currently this only affects the LaTeX reader, for things
like `\figurename`. Translation data files for 46 languages
can be found in `data/translations`.

* Make `--ascii` work with DocBook output too.

* Rename `--latex-engine` to `--pdf-engine`,
and `--latex-engine-opt` to `--pdf-engine-opt`.

* Removed `--parse-raw` and `readerParseRaw`. These were confusing.
Now we rely on the `+raw_tex` or `+raw_html` extension with latex or html
input. Thus, instead of `--parse-raw -f latex` we use `-f latex+raw_tex`,
and instead of `--parse-raw -f html` we use `-f html+raw_html`.

* With `--filter` R filters are now recognized, even if they are
not executable (#3940, #3941, Andrie de Vries).

* Support SVG in PDF output, converting with `rsvg2pdf` (#1793).

* Make epub an alias for epub3, not epub2.

* Removed `--epub-stylesheet`; use `--css` instead (#3472, #847).
Multiple stylesheets may be used. Stylesheets will be taken both from
`--css` and from the `stylesheet` metadata field (which can contain
either a file path or a list of them).

* `--mathml` and MathML in HTMLMathMethod no longer take an argument.
The argument was for a bridge JavaScript that used to be necessary
in 2004. We have removed the script already.

* `--katex` improvements. The latest version is used, and the
autoload script is loaded by default.

* Change MathJax CDN default since old one is shutting down (#3544).
Note: The new URL requires a version number, which we'll have
to update manually in subsequent pandoc releases in order to
take advantage of mathjax improvements.

* `--self-contained`: don't incorporate elements with `data-external="1"`
(#2656). You can leave an external link as it is by adding the attribute
data-external="1" to the element. Pandoc will then not try to
incorporate its content when `--self-contained` is used. This is
similar to a feature already supported by the EPUB writer.

* Allow `--extract-media` to work with non-binary input formats
(#1583, #2289). If `--extract-media` is supplied with a non-binary
input format, pandoc will attempt to extract the contents of all
linked images, whether in local files, data: uris, or external uris.
They will be named based on the sha1 hash of the contents.

* Make `papersize: a4` work regardless of the case of `a4`.
It is converted to `a4` in LaTeX and `A4` in ConTeXt.

* Make `east_asian_line_breaks` affect all readers/writers (#3703).

* Underlined elements are now treated consistently by readers
(#2270, hftf); they are always put in a Span with class `underline`.
This allows the user to treat them differently from other emphasis,
using a filter. Docx, Org, Textile, Txt2Tags, and HTML readers
have been changed.

* Improved behavior of `auto_identifiers` when there are explicit ids
(#1745). Previously only autogenerated ids were added to the list
of header identifiers in state, so explicit ids weren't taken
into account when generating unique identifiers. Duplicated
identifiers could result. This simple fix ensures that explicitly given
identifiers are also taken into account.

* Use `table-of-contents` for contents of toc, make `toc` a boolean
(#2872). Changed markdown, rtf, and HTML-based templates accordingly.
This allows you to set `toc: true` in the metadata; this
previously produced strange results in some output formats.
For backwards compatibility, `toc` is still set to the
toc contents. But it is recommended that you update templates
to use `table-of-contents` for the toc contents and `toc`
for a boolean flag.

* Change behavior with binary format output to stdout.
Previously, for binary formats, output to stdout was disabled
unless we could detect that the output was being piped (and not
sent to the terminal). Unfortunately, such detection is not
possible on Windows, leaving windows users no way to pipe binary
output. So we have changed the behavior in the following way:

+ Output to stdout is allowed when it can be determined that
the output is being piped (on non-Windows platforms).
+ If the `-o` option is not used, binary output is never sent
to stdout by default; instead, an error is raised.
+ If `-o -` is used, binary output is sent to stdout, regardless
of whether it is being piped. This works on Windows too.

* Better error behavior: uses of `error` have been replaced by
raising of `PandocError`, which can be trapped and handled by the
calling program.

* Removed `hard_line_breaks` extension from `markdown_github` (#3594).
GitHub has two Markdown modes, one for long-form documents like READMEs
and one for short things like issue comments. In issue comments, a line
break is treated as a hard line break. In README, wikis, etc., it is
treated as a space as in regular Markdown. Since pandoc is more likely to
be used to convert long-form documents from GitHub Markdown,
`-hard_line_breaks` is a better default.

* Include `backtick_code_blocks` extension in `mardkown_mmd` (#3637).

* Escape `MetaString` values (as added with `-M/--metadata` flag) (#3792).
Previously they would be transmitted to the template without any
escaping. Note that `--M title='*foo*'` yields a different result from

title: *foo*

In the latter case, we have emphasis; in the former case, just
a string with literal asterisks (which will be escaped
in formats, like Markdown, that require it).

* Allow `em`, `cm`, `in` for image height/width in HTML, LaTeX (#3450).

* HTML writer: Insert `data-` in front of unsupported attributes. Thus,
a span with attribute `foo` gets written to HTML5 with `data-foo`, so
it is valid HTML5. HTML4 is not affected. This will allow us to use
custom attributes in pandoc without producing invalid HTML. (With help
from Wandmalfarbe, #3817.)

* Plain writer: improved super/subscript rendering. We now
handle more non-digit characters for which there are
sub/superscripted unicode characters. When unicode
sub/superscripted characters are not available, we use
`_(..)` or `^(..)` (#3518).

* Docbook, JATS, TEI writers: print INFO message when omitting interior
header (#3750). This only applies to section headers inside list items,
e.g., which were otherwise silently omitted.

* Change to `--reference-links` in Markdown writer (#3701). With
`--reference-location` of `section` or `block`, pandoc will now repeat
references that have been used in earlier sections. The Markdown
reader has also been modified, so that *exactly* repeated references
do not generate a warning, only references with the same label but
different targets. The idea is that, with references after every block,
one might want to repeat references sometimes.

* ODT/OpenDocument writer:

+ Support `lang` attribute (#1667).
+ Added support for `--toc` (#2836). Thanks to @anayrat.

* Docx writer:

+ `lang` meta, see #1667 (Mauro Bieg, #3515).
+ Change `FigureWithCaption` to `CaptionedFigure` (iandol, #3658).
+ Use `Table` rather than `Table Normal` for table style (#3275).
`Table Normal` is the default table style and can't be modified.
+ Pass through comments (#2994). We assume that comments are defined as
parsed by the docx reader:

I want <span class="comment-start" id="0" author="Jesse Rosenthal"
date="2016-05-09T16:13:00Z">I left a comment.</span>some text to
have a comment <span class="comment-end" id="0"></span>on it.

We assume also that the id attributes are unique and properly
matched between comment-start and comment-end.
+ Bookmark improvements. Bookmark start/end now surrounds content rather
than preceding it. Bookmarks generated for Div with id
+ Add `keywords` metadata to docx document properties (Ian).

* RST writer: support unknown interpreted text roles by
parsing them as `Span` with `role` attributes (#3407). This
way they can be manipulated in the AST.

* HTML writer:

+ Line block: Use class instead of style attribute (#1623). We now
issue `<div class="line-block">` and include a default definition
for `line-block` in the default templates, instead of hard-coding a
`style` on the div.
+ Add class `footnoteBack` to footnote back references (Timm Albers).
This allows for easier CSS styling.
+ Render SmallCaps as span with smallcaps class (#1592), rather than
using a style attribute directly. This gives the user more flexibility
in styling small caps in CSS.
+ With reveal.js we use `data-src` instead of `src` for images for
lazy loading.
+ Special-case `.stretch` class for images in reveal.js (#1291).
Now in reveal.js, an image with class `stretch` in a paragraph
by itself will stretch to fill the whole screen, with no
caption or figure environment.

* Added warnings for non-rendered blocks to writers.

* Writers now raise an error on template failure.

* When creating a PDF via LaTeX, warn if the font is missing some
characters (#3742).

* Remove initial check for PDF-creating program (#3819).
Instead, just try running it and raise the exception if it
isn't found at that point. This improves things for users of Cygwin
on Windows, where the executable won't be found by `findExecutable`
unless `.exe` is added. The same exception is raised as before, but
at a later point.

* Readers issue warning for duplicate header identifiers (#1745).
Autogenerated header identifiers are given suffixes so as not to clash
with previously used header identifiers. But they may still coincide with
an explicit identifier that is given for a header later in the document,
or with an identifier on a div, span, link, or image. We now issue
a warning in this case, so users can supply an explicit identifier.

* CommonMark reader now supports `emoji`, `hard_line_breaks`, `smart`,
and `raw_html` extensions.

* Markdown reader:

+ Don't allow backslash + newline to affect block structure (#3730).
Note that as a result of this change, the following, which formerly
produced a header with two lines separated by a line break, will
now produce a header followed by a paragraph:

# Hi\

This may affect some existing documents that relied on
this undocumented and unintended behavior. This change makes pandoc
more consistent with other Markdown implementations, and with itself
(since the two-space version of a line break doesn't work inside ATX
headers, and neither version works inside Setext headers).

+ A space between the opening # and the header is now required for
pandoc and GitHub markdown (but not the other flavors). Disable
the `space_in_atx_header` extension (#3512) to restore the
original behavior.

* Org reader (Albert Krewinkel, unless noted):

+ Support `table.el` tables (#3314).
+ Support macros (#3401).
+ Support the `#+INCLUDE:` file inclusion mechanism (#3510).
Recognized include types are `example`, `export`, `src`, and
normal org file inclusion. Advanced features like line numbers
and level selection are not implemented yet.
+ Interpret more meta value as inlines. The values of the following
meta variables are now interpreted using org-markup instead of
treating them as pure strings: `keywords` (comma-separated list of
inlines), `subtitle` (inline values), `nocite` (inline values, can
be repeated).
+ Support `\n` export option (#3940). This turns all newlines in the
text into hard linebreaks.

* RST reader:

+ Improved admonition support (#223). We no longer add an
`admonition` class, we just use the class for the type of admonition,
`note` for example. We put the word corresponding to the label in
a paragraph inside a `Div` at the beginning of the admonition with
class `admonition-title`. This is about as close as we can get to
RST's own output.
+ Initial support of `.. table` directive. This allows adding captions
to tables.
+ Support `.. line-block` directive. This is deprecated but may still
be in older documents.
+ Support scale and align attributes of images (#2662).
+ Implemented implicit internal header links (#3475).
+ Support RST-style citations (#853). The citations appear at the end
of the document as a definition list in a special div with id
`citations`. Citations link to the definitions.
+ Recurse into bodies of unknown directives (#3432).
In most cases it's better to preserve the content than
to emit it. This isn't guaranteed to have good results;
it will fail spectacularly for unknown raw or verbatim directives.
+ Handle chained link definitions (#262). For example,

.. _hello:
.. _goodbye:

Here both `hello` and `goodbye` should link to ``.
+ Support anchors (#262). E.g.


.. _hello:


This is supported by putting "paragraph" in a `Div` with id `hello`.
+ Support `:widths:` attribute for table directive.
+ Implement csv-table directive (#3533). Most attributes are supported,
including `:file:` and `:url:`.
+ Support unknown interpreted text roles by parsing them as Span
with "role" attributes (#3407). This way they can be manipulated in
the AST.

* HTML reader: parse a span with class `smallcaps` as `SmallCaps`.

* LaTeX reader:

+ Implemented `\graphicspath` (#736).
+ Properly handle column prefixes/suffixes. For example, in
`\begin{tabular}{>{$}l<{$}>{$}l<{$} >{$}l<{$}}`
each cell will be interpreted as if it has a `$`
before its content and a `$` after (math mode).
+ Handle komascript `\dedication` (#1845). It now adds a
`dedication` field to metadata. It is up to the user to supply
a template that uses this variable.
+ Support all `\textXX` commands, where XX = `rm`, `tt`, `up`, `md`,
`sf`, `bf` (#3488). Spans with a class are used when there is
nothing better.
+ Expand `\newenvironment` macros (#987).
+ Add support for LaTeX subfiles package (Marc Schreiber, #3530).
+ Better support for subfigure package (#3577).
A figure with two subfigures turns into two pandoc
figures; the subcaptions are used and the main caption
ignored, unless there are no subcaptions.
+ Add support for \vdots (Marc Schreiber, #3607).
+ Add basic support for hyphenat package (Marc Schreiber, #3603).
+ Add basic `\textcolor` support (Marc Schreiber).
+ Add support for `tabularx` environment (Marc Schreiber, #3632).
+ Better handling of comments inside math environments (#3113).
This solves a problem with commented out `\end{eqnarray}` inside
an eqnarray (among other things).
+ Parse tikzpicture as raw verbatim environment if `raw_tex` extension
is selected (#3692). Otherwise skip with a warning. This is better
than trying to parse it as text!
+ Add `\colorbox` support (Marc Schreiber).
+ Set identifiers on Spans used for `\label`.
+ Have `\setmainlanguage` set `lang` in metadata.
+ Support etoolbox's `\ifstrequal`.
+ Support `plainbreak`, `fancybreak` et al from the memoir class
(bucklereed, #3833).
+ Support `\let`. Also, fix regular macros so they're expanded at the
point of use, and NOT also the point of definition. `\let` macros,
by contrast, are expanded at the point of definition. Added an
`ExpansionPoint` field to `Macro` to track this difference.
+ Support simple `\def` macros. Note that we still don't support
macros with fancy parameter delimiters, like `\def\foo#1..#2{...}`.
+ Support \chaptername, \partname, \abstractname, etc. (#3559,
obsoletes #3560).
+ Put content of `\ref`, `\label`, `\eqref` commands into `Span` with
attributes, so they can be handled in filters (Marc Schreiber, #3639)
+ Add Support for `glossaries` and `acronym` package (Marc Schreiber,
#3589). Acronyms are not resolved by the reader, but acronym and
glossary information is put into attributes on Spans so that they
can be processed in filters.
+ Use `Link` instead of `Span` for `\ref`. This makes more sense
semantically and avoids unnecessary `Span [Link]` nestings when
references are resolved.
+ Rudimentary support for `\hyperlink`.
+ Support `\textquoteleft|right`, `\textquotedblleft|right` (#3849).
+ Support `\lq`, `\rq`.
+ Implement `\newtoggle`, `\iftoggle`, `\toggletrue|false` from etoolbox
+ Support `\RN` and `\Rn`, from biblatex (bucklereed, #3854).
+ Improved support for `\hyperlink`, `\hypertarget` (#2549).
+ Support `\k` ogonek accent.
+ Improve handling of accents. Handle ogonek, and fall back correctly
with forms like `\"{}`.
+ Better support for ogonek accents.
+ Support for `\faCheck` and `\faClose` (Marc Schreiber, #3727).
+ Support for `xspace` (Marc Schreiber, #3797).
+ Support `\setmainlanguage` or `\setdefaultlanguage` (polyglossia)
and `\figurename`.
+ Better handling of `\part` in LaTeX (#1905). Now we parse chapters as
level 0 headers, and parts as level -1 headers. After parsing, we
check for the lowest header level, and if it's less than 1 we bump
everything up so that 1 is the lowest header level. So `\part` will
always produce a header; no command-line options are needed.
+ Add block version of `\textcolor` (Marc Schreiber).
+ `\textcolor` works as inline and block command (Marc Schreiber).
+ `\textcolor` will be parse as span at the beginning of a paragraph
(Marc Schreiber).
+ Read polyglossia/babel `\text(LANG){...}` (bucklereed)
+ Improved handling of include files in LaTeX reader (#3971).
Previously `\include` wouldn't work if the included file
contained, e.g., a begin without a matching end.
+ Support `\expandafter` (#3983).
+ Handle `\DeclareRobustCommand` (#3983). Currently it's just treated
as a synonym for `\newcommand`.
+ Handle `\lettrine` (Mauro Bieg).

* Math improvements due to updates in texmath:

+ Improved handling of accents and upper/lower delimiters.
+ Support for output in GNU eqn format (used with *roff).
+ Allow `\boldsymbol` + a token without braces, and similarly
with other styling commands.
+ Improve parsing of `\mathop` to allow multi-character operator names.
+ Add thin space after math operators when "faking it with

* `walk` is now used instead of `bottomUp` in the `ToJSONFilter`
instance for `a -> [a]` (pandoc-types). Note that behavior
will be slightly different, since `bottomUp`'s treatment of
a function `[a] -> [a]` is to apply it to each sublist of a
list, while walk applies it only to maximal sublists.
Usually the latter behavior is what is wanted, and the
former can be simulated when needed. But there may be
existing filters that need to be rewritten in light of the
new behavior. Performance should be improved.

* There are some changes to syntax highlighting due to revisions
in the `skylighting` library:

+ Support for `powershell` has been added, and many syntax
definitions have been updated.
+ Background colors have been added to the `kate` style.
+ The way highlighted code blocks are formatted in HTML has
been changed (David Baynard), in ways that may require
changes in hard-coded CSS affecting highlighting.
(If you haven't included hard-coded highlighting CSS in
your template, you needn't change anything.)

[API changes]

* New module `Text.Pandoc.Class` (Jesse Rosenthal, John MacFarlane).
This contains definitions of the `PandocMonad` typeclass, the
`PandocIO` and `PandocPure` monads, and associated functions.

* Changed types of all writers and readers.

+ We now use `Text` instead of `String` in the interface (#3731).
(We have not yet changed the internals of most readers to work
with `Text`, but making this change in the API now opens up a
path to doing that.)
+ The result is now of form `m a` with constraint `PandocMonad m`.
Readers and writers can be combined to form monadic values which
can be run using either `runIO` or `runPure`. If `runIO` is used,
then both readers and writers will be able to do IO when needed
(for include files, for example); if `runPure` is used,
then the functions are pure and will not touch IO.
+ Where previously you used
`writeRST def (readMarkdown def "[foo](url)")`, now you
would use
`runPure $ readMarkdown def (pack "[foo](url)") >>= writeRST def`.

* New module `Text.Pandoc.Readers` (Albert Krewinkel). This
contains reader helper functions formerly defined in the
top-level `Text.Pandoc` module.

+ Changed `StringReader` -> `TextReader`.
+ `getReader` now returns a pair of a reader and
`Extensions`, instead of building the extensions into the
reader (#3659). The calling code must explicitly set
`readerExtensions` using the `Extensions` returned. The
point of the change is to make it possible for the calling
code to determine what extensions are being used.

* New module `Text.Pandoc.Writers` (Albert Krewinkel).
This contains writer helper functions formerly defined in the
top-level `Text.Pandoc` module.

+ Changed `StringWriter` -> `TextWriter`.
+ `getWriter` now returns a pair of a reader and
`Extensions`, instead of building the extensions into the
reader (#3659). The calling code must explicitly set
`readerExtensions` using the `Extensions` returned. The
point of the change is to make it possible for the calling
code to determine what extensions are being used.

* New module `Text.Pandoc.Lua`, exporting `runLuaFilter` (Albert Krewinkel,

* New module `Text.Pandoc.App`. This abstracts out the functionality
of the command line program (`convertWithOpts`), so it can be reproduced
e.g. in a desktop or web application. Instead of exiting, we throw errors
(#3548), which are caught (leading to exit) in pandoc.hs, but allow other
users of `Text.Pandoc.App` to recover. `pandoc.hs` is now a 2-liner.
The module also exports some utility functions for parsing options
and running filters.

* New module `Text.Pandoc.Logging` (exported module) (#3392).
This now contains the `Verbosity` definition previously in
`Text.Pandoc.Options`, as well as a new `LogMessage` datatype that will
eventually be used instead of raw strings for warnings. This will enable
us, among other things, to provide machine-readable warnings if desired.
Include ToJSON instance and showLogMessage. This gives us the possibility
of both machine-readable and human-readable output for log messages.

* New module `Text.Pandoc.BCP47`, with `getLang`, `Lang(..)`, `parseBCP47`.

* New module `Text.Pandoc.Translations`, exporting `Term`,
`Translations`, `readTranslations`.

* New module `Text.Pandoc.Readers.LaTeX.Types', exporting `Macro`, `Tok`,
`TokType`, `Line`, `Column`.

* `Text.Pandoc.Error`: added many new constructors for `PandocError`.

* Expose some previously private modules (#3260). These are often
helpful to people writing their own reader or writer modules:

+ `Text.Pandoc.Writers.Shared`
+ `Text.Pandoc.Parsing`
+ `Text.Pandoc.Asciify`
+ `Text.Pandoc.Emoji`
+ `Text.Pandoc.ImageSize`
+ `Text.Pandoc.Highlighting`
* New module `Text.Pandoc.Extensions` (Albert Krewinkel):
Extension parsing and processing functions were defined in the top-level
`Text.Pandoc` module. These functions are moved to the Extensions
submodule as to enable reuse in other submodules.

* Add `Ext_raw_attribute` constructor for `Extension`.

* Add `Ext_fenced_divs` constructor for `Extension'.

* Add `Ext_four_space_rule` constructor in `Extension`.

* Add `Ext_gfm_auto_identifiers` constructor for `Extension`.

* Add `Monoid` instance for `Extensions`.

* Add `Text.Pandoc.Writers.Ms`, exporting `writeMs`.

* Add `Text.Pandoc.Writers.JATS`, exporting `writeJATS`.

* Add `Text.Pandoc.Writers.Muse`, exporting `writeMuse`.

* Add `Text.Pandoc.Readers.Muse`, exporting `readMuse`.

* Add `Text.Pandoc.Readers.TikiWiki`, exporting `readTikiWiki`.

* Add `Text.Pandoc.Readers.Vimwiki`, exporting `readVimwiki`.

* Add `Text.Pandoc.Readers.Creole`, exporting `readCreole`.

* Export `setVerbosity` from `Text.Pandoc`.

* `Text.Pandoc.Pretty`: Add `Eq` instance for `Doc`.

* `Text.Pandoc.XML`: `toEntities`: changed type to `Text -> Text`.

* `Text.Pandoc.UTF8`:

+ Export `fromText`, `fromTextLazy`, `toText`, `toTextLazy`.
Define `toString`, `toStringLazy` in terms of them.
+ Add new functions parameterized on `Newline`: `writeFileWith`,
`putStrWith`, `putStrLnWith`, `hPutStrWith`, `hPutStrLnWith`.

* `Text.Pandoc.MediaBag`: removed `extractMediaBag`.

* `Text.Pandoc.Highlighting`:

+ `highlighting` now returns an Either rather than Maybe.
This allows us to display error information returned by the skylighting
library. Display a warning if the highlighting library throws an error.
+ Add parameter for `SyntaxMap` to `highlight`.

* `Text.Pandoc.Writers.Math`:

+ Export `defaultMathJaxURL`, `defaultKaTeXURL`. This will ensure that
we only need to update these in one place.

* `Text.Pandoc.SelfContained`:

+ Removed `WriterOptions` parameter from `makeSelfContained`.
+ Put `makeSelfContained` in PandocMonad instead of IO. This removes
the need to pass MediaBag around and improves exceptions. It also
opens up the possibility of using makeSelfContained purely.
+ Export `makeDataURI`.

* `Text.Pandoc.ImageSize`:

+ Export `lengthToDim`, new function `scaleDimension`.
+ Export `inEm` from ImageSize (#3450).
+ Change `showFl` and `show` instance for `Dimension` so
extra decimal places are omitted.
+ Added `Em` as a constructor of `Dimension`.
+ Add `WriterOptions` parameter to `imageSize` signature (Mauro Bieg).

* `Text.Pandoc.Templates`:

+ Change type of `renderTemplate'`. Now it runs in `PandocMonad`
and raises a proper `PandocTemplateError` if there are problems, rather
than failing with uncatchable `error`.
+ Change signature of `getDefaultTemplate`. Now it runs in any instance
of `PandocMonad`, and returns a `String` rather than an `Either` value.
And it no longer takes a `datadir` parameter, since this can be
retrieved from `CommonState`.

* `Text.Pandoc.Options`:

+ Added `writerEpubSubdirectory` to `WriterOptions` (#3720).
The EPUB writer now takes its EPUB subdirectory from this option.
+ In `WriterOptions`, rename `writerLaTeXEngine` to `writerPdfEngine`
and `writerLaTeXArgs` to `writerPdfArgs` (Mauro Bieg, #3909).
+ Add `writerSyntaxMap` to `WriterOptions`.
+ Removed `writerEpubStylesheet` from `WriterOptions`.
+ Remove `writerUserDataDir` from `WriterOptions`. It is now carried
in `CommonState` in `PandocMonad` instances. (And thus it can be used
by readers too.)
+ Changed `writerEpubMetadata` to a `Maybe String`.
+ Removed `readerApplyMacros` from `ReaderOptions`. Now we just check
the `latex_macros` reader extension.
+ FromJSON/ToJSON instances for `ReaderOptions`.
+ In `HTMLMathMethod`, the `KaTeX` contsructor now takes only
one string (for the KaTeX base URL), rather than two.
+ Removed `writerSourceURL` from `WriterOptions`. We now use
`stSourceURL` in `CommonState`, which is set by `setInputFiles`.

* `Text.Pandoc.Shared`:

+ `tabFilter` now takes a `Text`, not `String`.
+ `openURL`: Changed type from an Either. Now it will just raise
an exception to be trapped later.
+ Remove `normalizeSpaces` (#1530).
+ Remove `warn`. (Use `report` from `Text.Pandoc.Class` instead.)
+ Export a new function `crFilter`.
+ Add `eastAsianLineBreakFilter` (previously in Markdown reader).
+ Provide custom `isURI` that rejects unknown schemes.
(Albert Krewinkel, #2713). We also export the set of known
`schemes`. The new function replaces the function of the same name
from `Network.URI`, as the latter did not check whether a scheme is
well-known. All official IANA schemes (as of 2017-05-22) are
included in the set of known schemes. The four non-official schemes
`doi`, `isbn`, `javascript`, and `pmid` are kept.
+ Remove `err`.
+ Remove `readDataFile`, `readDefaultDataFile`, `getReferenceDocx`,
`getReferenceODT`. These now live in `Text.Pandoc.Class`,
where they are defined in terms of `PandocMonad`
primitives and have different signatures.
+ Remove `openURL`. Use `openURL` from `Text.Pandoc.Class` instead.
+ Add `underlineSpan`.

* `Text.Pandoc.Readers.HTML`: export new `NamedTag` class.

* `Text.Pandoc.Readers.Markdown`: remove `readDocxWithWarnings`.
With the new API one can simply use `getLog` after running
the reader.

* `Text.Pandoc.Readers.LaTeX`: Changed types for `rawLaTeXInline`
and `rawLaTeXBlock`. (Both now return a `String`, and they are
polymorphic in state.)

[bug fixes and under-the-hood improvements]

* TEI writer: Added identifiers on `<div>` elements.

* DokuWiki reader: Better handling for code block in list item (#3824).

* Custom writer: Remove old preprocesesor conditionals (Albert Krewinkel).

* ZimWiki writer: Removed internal formatting from note and table cells,
because ZimWiki does not support it (Alex Ivkin, #3446).

* MediaWiki writer:

+ Updated list of syntax highlighting languages (#3461).
Now `r` gets you `<source>` rather than `<code>` (among others).
+ Add display attribute on `<math>` tags (#3452). This allows display
math to be rendered properly.
+ Remove newline before `</ref>` (#2652).
+ Don't softbreak lines inside list items (#3531).

* Org writer:

+ Reduce to two spaces after bullets (#3417, Albert Krewinkel).
+ Add unit tests (Alexander Krotov).
+ Stop using raw HTML to wrap divs (Albert Krewinkel, #3771).
+ Do not strip `#` from Org anchor links (Alexander Krotov).

* CommonMark writer:

+ Avoid excess blank lines at end of output.
+ Prefer pipe tables to HTML tables even if it means losing relative
column width information (#3734).
+ Support table, strikethrough extensions, when enabled (as with gfm).
Note that we bypass the commonmark writer from cmark and construct our
own pipe tables, with better results.
+ Properly support `--wrap=none`.
+ Use smallcaps class for `SmallCaps` (#1592).
+ Omit "fig:" prefix in image titles. This is used internally to
indicate internal figures.

* RST writer:

+ Properly handle table captions.
+ Don't wrap lines in in definition list terms. Wrapping is not allowed.
+ Implemented `+/-smart` and improved escaping with `+smart`.
+ Add empty comments when needed to avoid including a blockquote
in the indented content of a preceding block (#3675).
+ Improve grid table output, fix bug with empty rows (#3516).
Uses the new `gridTable` in Writers.Shared, which is here
improved to better handle 0-width cells.
+ Remove space at beginning/end of RST code span (#3496). Otherwise
we get invalid RST. There seems to be no way to escape the space.
+ Add header anchors when header has non-standard id (#3937).
+ Correctly handle inline code containing backticks, using a `:literal:`
role (#3974).
+ Don't backslash-escape word-internal punctuation (#3978).

* Markdown writer:

+ Don't include variables in metadata blocks. Previously variables set
on the command line were included in e.g. YAML metadata, contrary to
documentation and intentions.
+ Improved escaping with `+smart`.
+ Fixed grid tables embedded in grid tables (#2834).
+ Use span with class 'smallcaps' for SmallCaps, instead of a style
attribute as before (#1592).
+ Escape initial `%` in a paragraph if the `pandoc_title_blocks`
extension is enabled (#3454). Otherwise in a document starting with
a literal `%` the first line is wrongly interpreted as a title.
+ Fixed false ordered lists in YAML metadata (#3492, #1685). Now we
properly escape things that would otherwise start ordered lists,
such as

title: 1. inline
+ Better handling of tables with empty columns (#3337). We now
calculate the number of columns based on the longest row (or the
length of aligns or widths).
+ Escape unordered list markers at beginning of paragraph (#3497), to
avoid false interpretation as a list.
+ Escape `|` appropriately.
+ Ensure space before list at top level (#3487).
+ Avoid spurious blanklines at end of document after tables and list,
for example.
+ Fixed bugs in simple/multiline list output (#3384).
Previously we got overlong lists with `--wrap=none`. This is fixed.
Previously a multiline list could become a simple list (and would
always become one with `--wrap=none`).
+ Don't emit a simple table if `simple_tables` disabled (#3529).
+ Case-insensitive reference links (David A Roberts, #3616).
Ensure that we do not generate reference links whose labels differ only
by case. Also allow implicit reference links when the link
text and label are identical up to case.
+ Put space before reference link definitions (Mauro Bieg, #3630).
+ Better escaping for links (David A. Roberts, #3619). Previously the
Markdown writer would sometimes create links where there were none
in the source. This is now avoided by selectively escaping bracket
characters when they occur in a place where a link might be created.
+ Added missing `\n` (David A. Roberts, #3647).
+ Fixed duplicated reference links with `--reference-links`
and `--reference-location=section` (#3674). Also ensure that there
are no empty link references `[]`.
+ Avoid inline surround-marking with empty content (#3715).
E.g. we don't want `<strong></strong>` to become `****`.
Similarly for emphasis, super/subscript, strikeout.
+ Don't allow soft break in header (#3736).
+ Make sure `plain`, `markdown_github`, etc. work for raw.
Previously only `markdown` worked. Note: currently a raw block labeled
`markdown_github` will be printed for any `markdown` format.
+ Ensure that `+` and `-` are escaped properly so they don't cause
spurious lists (#3773). Previously they were only
if succeeded by a space, not if they were at end of line.
+ Use pipe tables if `raw_html` disabled and `pipe_tables` enabled,
even if the table has relative width information (#3734).
+ Markdown writer: don't crash on `Str ""`.
+ Make `Span` with null attribute transparent. That is, we don't use
brackets or `<span>` tags to mark spans when there are no attributes;
we simply output the contents.
+ Escape pipe characters when `pipe_tables` enabled (#3887).
+ Better escaping of `<` and `>`. If `all_symbols_escapable` is set,
we backslash escape these. Otherwise we use entities as before.
+ When writing plain, don't use `&nbsp;` to separate list and indented
code. There's no need for it in this context, since this isn't to be
interpreted using Markdown rules.
+ Preserve classes in JS obfuscated links (Timm Albers, #2989).
HTML links containing classes originally now preserve them when using
JavaScript email obfuscation.
+ Render `SmallCaps` as a native span when `native_spans` are enabled.
+ Always write attributes with `bracketed_spans` (d-dorazio).

* Man writer:

+ Fix handling of nested font commands (#3568). Previously pandoc emitted
incorrect markup for bold + italic, for example, or bold + code.
+ Avoid error for definition lists with no definitions (#3832).

* DocBook writer:

+ Fix internal links with `writerIdentifierPrefix opt`
(#3397, Mauro Bieg).

* Docx writer:

+ Don't include bookmarks on headers unless non-null id (#3476).
+ Support 9 levels of headers (#1642).
+ Allow 9 list levels (#3519).
+ Don't take `distArchive` from datadir (#3322). The docx writer takes
components from the distribution's version of `reference.docx` when it
can't find them in a user's custom `reference.docx`. Previously, we
allowed a `reference.docx` in the data directory (e.g. `~/.pandoc`)
to be used as the distribution's reference.docx. This led to a
bizarre situation where pandoc would produce a good docx using
`--template ~/.pandoc/ref.docx`, but if `ref.docx` were moved to
`~/.pandoc/reference.docx`, it would then produce a corrupted docx.
+ Fixed handling of soft hyphen (0173) (#3691).
+ Better handling of keywords (#3719).
+ Cleaner code for handling dir and style attributes for `Div`.
+ Use `Set` for dynamic styles to avoid duplicates.
+ Removed redundant element from data/docx/word/numbering.xml.
The elements we need are generated when the document is
compiled; this didn't do anything.
+ Activate `evenAndOddHeaders` from reference docx (#3901,
Augustín Martín Barbero).

* ODT/OpenDocument writer:

+ Calculate aspect ratio for percentage-sized images (Mauro Bieg, #3239).
+ Use more widely available bullet characters (#1400). The old
characters weren't available in some font sets. These seem to work
well on Windows and Linux versions of LibreOffice.
+ Wider labels for lists (#2421). This avoids overly narrow labels for
ordered lists with `()` delimiters. However, arguably it creates
overly wide labels for bullets. Also, lists now start flush with
the margin, rather than indented.
+ Fixed dropped elements in some ordered lists (#2434).

* FB2 writer:

+ Don't render `RawBlock` as code.
+ Don't fail with an error on interior headers (e.g. in list) (#3750).
Instead, omit them with an INFO message.
+ Add support for "lang" metadata (Alexander Krotov, #3625).
+ Format `LineBlock` as poem (Alexander Krotov). Previously writer
produced one paragraph with `<empty-line/>` elements, which are not
allowed inside `<p>` according to FB2 schema.
+ Replace `concatMap` with `cMap` (Alexander Krotov).
+ Write FB2 lists without nesting blocks inside `<p>` (Alexander
Krotov, #4004)

* HTML writer:

+ Make sure `html4`, `html5` formats work for raw blocks/inlines.
+ Render raw inline environments when `--mathjax` used (#3816).
We previously did this only with raw blocks, on the assumption
that math environments would always be raw blocks. This has changed
since we now parse them as inline environments.
+ Ensure we don't get two style attributes for width and height.
+ Report when not rendering raw inline/block.
+ Issue warning if no title specified and template used (#3473).
+ Info message if `lang` is unspecified (#3486).
+ Removed unused parameter in `dimensionsToAttributeList`.
+ Avoid two class attributes when adding `uri` class (#3716).
+ Fix internal links with `writerIdentifierPrefix opt` (#3397, Mauro
+ Use revealjs's math plugin for mathjax (#3743). This is a thin
wrapper around mathjax that makes math look better on revealjs.
+ Slidy: use h1 for all slides, even if they were originally
level 2 headers (#3566). Otherwise the built-in table of contents
in Slidy breaks.

* LaTeX writer:

+ Don't render LaTeX images with data: URIs (#3636). Note that
`--extract-media` can be used when the input contains data: URIs.
+ Make highlighted code blocks work in footnotes (Timm Albers).
+ Don't use figure inside table cell (#3836).
+ Use proper code for list enumerators (#3891). This should fix problems
with lists that don't use arabic numerals.
+ Always add hypertarget when there's a non-empty identifier (#2719).
Previously the hypertargets were only added when there was actually
a link to that identifier.
+ Use `%` after hypertarget before code block.
+ Add `\leavevmode` before hypertarget at start of paragraph (#2704,
fixes formatting problems in beamer citations).
+ Don't use `lstinline` in \item[..] (#645). If you do, the contents
of item disappear or are misplaced. Use `\texttt` instead.
+ Fix problem with escaping in `lstinline` (#1629). Previously the
LaTeX writer created invalid LaTeX when `--listings` was specified and
a code span occurred inside emphasis or another construction.
+ Fix error with line breaks after empty content (#2874). LaTeX
requires something before a line break, so we insert a `~` if no
printable content has yet been emitted.
+ Use BCP47 parser.
+ Fixed detection of otherlangs (#3770). We weren't recursing into
inline contexts.
+ Handle language in inline code with `--listings` (#3422).
+ Write euro symbol directly in LaTeX (Andrew Dunning, #3801).
The textcomp package allows pdfLaTeX to parse `€` directly, making the
`\euro` command unneeded.
+ Fixed footnotes in table captions (#2378). Note that if the table has
a first page header and a continuation page header, the notes will
appear only on the first occurrence of the header.
+ In `writeBeamer` output, allow hyperlinks to frames (#3220).
Previously you could link to a header above or below slide level but
not *to* slide level. This commit changes that. Hypertargets are
inserted inside frame titles; technically the reference is to just
after the title, but in normal use (where slides are viewed full
screen in a slide show), this does not matter.
+ Remove `\strut` at beginning of table cells (#3436). This fixes a
problem with alignment of lists in table cells. The `\strut` at the
end seems to be enough to avoid the too-close spacing that motivated
addition of the strut in #1573.
+ Add partial siunitx Support (Marc Schreiber, #3588).

* ConTeXt writer:

+ Refactored to use BCP47 module.
+ Remove unnecessary `$` (Alexander Krotov, #3482).
+ Use header identifiers for chapters (#3968).

* EPUB writer:

+ `title_page.xhtml` is now put in `text/`.
+ Don't strip formatting in TOC (#1611).

* Textile reader:

+ Fix bug for certain links in table cells (#3667).
+ Allow 'pre' code in list item (#3916).

* HTML reader:

+ Added warnings for ignored material (#3392).
+ Better sanity checks to avoid parsing unintended things as
raw HTML in the Markdown reader (#3257).
+ Revise treatment of `li` with `id` attribute (#3596). Previously we
always added an empty div before the list item, but this created
problems with spacing in tight lists. Now we do this: If the list
item contents begin with a `Plain` block, we modify the `Plain`
block by adding a `Span` around its contents. Otherwise, we add a
`Div` around the contents of the list item (instead of adding an
empty `Div` to the beginning, as before).
+ Add `details` tag to list of block tags (#3694).
+ Removed `button` from block tag list (#3717). It is already in the
`eitherBlockOrInlineTag` list, and should be both places.
+ Use `Set`s instead of lists for block tag lookup.
+ Rewrote to use `Text` throughout. Effect on memory usage is modest
(< 10%).
+ Use the lang value of `<html>` to set the lang meta value (bucklereed,
+ Ensure that paragraphs are closed properly when the parent block
element closes, even without `</p>` (#3794).
+ Parse `<figure>` and `<figcaption>` (Mauro Bieg, #3813).
+ Parse `<main>` like `<div role=main>` (bucklereed, #3791).
`<main>` closes `<p>` and behaves like a block element generally
+ Support column alignments (#1881). These can be set either
with a `width` attribute or with `text-width` in a `style` attribute.
+ Modified state type to be an instance of `HasLogMessages`, so
`registerHeader` can issue warnings.
+ `</td>` or `</th>` should close any open block tag (#3991).
+ `<td>` should close an open `<th>` or `<td>`.
+ `htmlTag` improvements (#3989). We previously failed on cases
where an attribute contained a `>` character. This patch fixes the
bug, which especially affects raw HTML in Markdown.

* Txt2Tags reader:

+ Newline is not indentation (Alexander Krotov).

* MediaWiki reader:

+ Allow extra hyphens after `|-` in tables (#2649).
+ Allow blank line after table start (#2649).
+ Fixed more table issues (#2649).
+ Ensure that list starts begin at left margin (#2606). Including when
they're in tables or other list items.
+ Make smart double quotes depend on `smart` extension (#3585).
+ Don't do curly quotes inside `<tt>` contexts (#3585). Even if `+smart`.
+ Modified state type to be an instance of `HasLogMessages`, so
`registerHeader` can issue warnings.

* TWiki reader (Alexander Krotov):

+ Remove unnecessary `$` (#3597).
+ Simplify `linkText` (#3605).

* EPUB reader:

+ Minor refactoring, avoiding explicit MediaBag handling.
This all works behind the scenes in CommonState plumbing.

* Docx reader:

+ Don't drop smartTag contents (#2242).
+ Handle local namespace declarations (#3365). Previously we didn't
recognize math, for example, when the xmlns declaration occurred on
the element and not the root.
+ More efficient trimSps (#1530). Replacing `trimLineBreaks`. This
does the work of `normalizeSpaces` as well, so we avoid the need for
that function here.
+ Avoid 0-level headers (Jesse Rosenthal, #3830). We used to parse
paragraphs styled with "HeadingN" as "nth-level header." But if a
document has a custom style named "Heading0", this will produce a
0-level header, which shouldn't exist. We only parse this style
if N>0. Otherwise we treat it as a normal style name, and
follow its dependencies, if any.
+ Add tests for avoiding zero-level header (Jesse Rosenthal).

* ODT reader:

+ Replaced `collectRights` with Rights from `Data.Either`.
+ Remove dead code (Albert Krewinkel).

* Org reader (Albert Krewinkel, unless noted).

+ Don't allow tables inside list items (John MacFarlane, #3499).
+ Disallow tables on list marker lines (#3499).
+ Convert markup at beginning of footnotes (John MacFarlane, #3576).
+ Allow emphasized text to be followed by `[` (#3577).
+ Handle line numbering switch for src blocks.
The line-numbering switch that can be given to source blocks (`-n` with
an start number as an optional parameter) is parsed and translated to a
class/key-value combination used by highlighting and other readers and
+ Stop adding rundoc prefix to src params. Source block parameter names
are no longer prefixed with `rundoc`. This was intended to simplify
working with the rundoc project, a babel runner. However, the rundoc
project is unmaintained, and adding those markers is not the reader's
job anyway. The original language that is specified for a source
element is now retained as the `data-org-language` attribute and only
added if it differs from the translated language.
+ Allow multi-word arguments to src block params (#3477). The reader now
correctly parses src block parameter list even if parameter arguments
contain multiple words.
+ Avoid creating `nullMeta` by applying `setMeta` directly
(Alexander Krotov).
+ Replace `sequence . map` with `mapM`.
+ Fix smart parsing behavior. Parsing of smart quotes and special
characters can either be enabled via the `smart` language extension or
the `'` and `-` export options. Smart parsing is active if either the
extension or export option is enabled. Only smart parsing of special
characters (like ellipses and en and em dashes) is enabled by default,
while smart quotes are disabled. Previously, all smart parsing was
disabled unless the language extension was enabled.
+ Subject full doc tree to headline transformations (Albert Krewinkel,
#3695). Emacs parses org documents into a tree structure, which is
then post-processed during exporting. The reader is changed to do the
same, turning the document into a single tree of headlines starting
at level 0.
+ Fix cite parsing behaviour (Herwig Stuetz). Until now, `org-ref`
cite keys included special characters also at the end. This caused
problems when citations occur right before colons or at the end of
a sentence. With this change, all non alphanumeric characters at
the end of a cite key are ignored. This also adds `,` to the list
of special characters that are legal in cite keys to better mirror
the behaviour of org-export.
+ Fix module names in haddock comments. Copy-pasting had lead to
haddock module descriptions containing the wrong module names.
+ Recognize babel result blocks with attributes (#3706). Babel
result blocks can have block attributes like captions and names.
Result blocks with attributes were not recognized and were parsed
as normal blocks without attributes.
+ Include tags in headlines. The Emacs default is to include tags in the
headline when exporting. Instead of just empty spans, which contain the
tag name as attribute, tags are rendered as small caps and wrapped in
those spans. Non-breaking spaces serve as separators for multiple tags.
+ Respect export option for tags (#3713). Tags are appended to
headlines by default, but will be omitted when the `tags` export option
is set to nil.
+ Use `tag-name` attribute instead of `data-tag-name`.
+ Use `org-language` attribute rather than `data-org-language`.
+ Modified state type to be an instance of `HasLogMessages`, so
`registerHeader` can issue warnings.
+ End footnotes after two blank lines. Footnotes can not only be
terminated by the start of a new footnote or a header, but also by two
consecutive blank lines.
+ Update emphasis border chars (#3933). The org reader was updated to
match current org-mode behavior: the set of characters which are
acceptable to occur as the first or last character in an org emphasis
have been changed and now allows all non-whitespace chars at the
inner border of emphasized text (see `org-emphasis-regexp-components`).

* RST reader:

+ Fixed small bug in list parsing (#3432). Previously the parser didn't
handle properly this case:

* - a
- b
* - c
- d
+ Handle multiline cells in simple tables (#1166).
+ Parse list table directive (Keiichiro Shikano, #3432).
+ Make use of `anyLineNewline` (Alexander Krotov, #3686).
+ Use `anyLineNewline` in `rawListItem` (Alexander Krotov, #3702).
+ Reorganize block parsers for ~20% faster parsing.
+ Fixed `..include::` directive (#3880).
+ Handle blank lines correctly in line blocks (Alexander Krotov, #3881).
Previously pandoc would sometimes combine two line blocks separated
by blanks, and ignore trailing blank lines within the line block.
+ Fix indirect hyperlink targets (#512).

* Markdown reader:

+ Allow attributes in reference links to start on next line (#3674).
+ Parse YAML metadata in a context that sees footnotes defined in
the body of the document (#1279).
+ When splitting pipe table cells, skip tex math (#3481).
You might have a `|` character inside math. (Or for that matter
something that the parser might mistake for raw HTML.)
+ Treat span with class `smallcaps` as SmallCaps.
This allows users to specify small caps in Markdown this way:
`[my text]{.smallcaps}` (#1592).
+ Fixed internal header links (#2397).
This patch also adds `shortcut_reference_links` to the list
of mmd extensions.
+ Treat certain environments as inline
when they occur without space surrounding them (#3309, #2171).
E.g. equation, math. This avoids incorrect vertical space
around equations.
+ Optimized `nonindentSpaces`. Makes the benchmark go from 40 to 36 ms.
+ Allow latex macro definitions indented 1-3 spaces.
Previously they only worked if nonindented.
+ Improved parsing of indented raw HTML blocks (#1841).
Previously we inadvertently interpreted indented HTML as
code blocks. This was a regression. We now seek to determine the
indentation level of the contents of an HTML block, and (optionally)
skip that much indentation. As a side effect, indentation may be
stripped off of raw HTML blocks, if `markdown_in_html_blocks` is
used. This is better than having things interpreted as indented
code blocks.
+ Fixed smart quotes after emphasis (#2228). E.g. in `*foo*'s 'foo'`.
+ Warn for notes defined but not used (#1718).
+ Use `anyLineNewline` (Alexander Krotov).
+ Interpret YAML metadata as Inlines when possible (#3755). If
the metadata field is all on one line, we try to interpret it as
Inlines, and only try parsing as Blocks if that fails. If it
extends over one line (including possibly the `|` or `>` character
signaling an indented block), then we parse as Blocks. This was
motivated by some German users finding that `date: '22. Juin 2017'`
got parsed as an ordered list.
+ Fixed spurious parsing as citation as reference def (#3840).
We now disallow reference keys starting with `@` if the
`citations` extension is enabled.
+ Parse `-@roe` as suppress-author citation (pandoc-citeproc#237).
Previously only `[-@roe]` (with brackets) was recognized as
suppress-author, and `-@roe` was treated the same as `@roe`.
+ Fixed parsing of fenced code after list when there is no intervening
blank line (#3733).
+ Allow raw latex commands starting with `\start` (#3558). Previously
these weren't allowed because they were interpreted as starting
ConTeXt environments, even without a corresponding `\stop`...
+ Added `inlines`, `inlines1`.
+ Require nonempty alt text for `implicit_figures` (#2844).
A figure with an empty caption doesn't make sense.
+ Removed texmath macro material; now all this is handled
in the LaTeX reader functions.
+ Fixed bug with indented code following raw LaTeX (#3947).

* LaTeX reader:

+ Rewrote LaTeX reader with proper tokenization (#1390,
#2118, #3236, #3779, #934, #982). This rewrite is primarily
motivated by the need to get macros working properly. A side benefit
is that the reader is significantly faster. We now tokenize the
input text, then parse the token stream. Macros modify the token
stream, so they should now be effective in any context, including
math. Thus, we no longer need the clunky macro processing
capacities of texmath.
+ Parse `\,` to `\8198` (six-per-em space) (Henri Werth).
+ Allow `\newcommand\foo{blah}` without braces.
+ Support `\lstinputlisting` (#2116).
+ Issue warnings when skipping unknown latex commands (#3392).
+ Include contents of `\parbox`.
+ Allow `\hspace` and `\vspace` to count as raw block or inline.
Previously we would refuse to parse anything as raw inline if
it was in the `blockCommands` list. Now we allow exceptions
if they're listed under ignoreInlines in inlineCommands.
This should make it easier e.g. to include an `\hspace`
between two side-by-side raw LaTeX tables.
+ Don't drop contents of `\hypertarget`.
+ Handle spaces before `\cite` arguments.
+ Allow newpage, clearpage, pagebreak in inline contexts as well as
block contexts (#3494).
+ Treat `{{xxx}}` the same as `{xxx}` (#2115).
+ Use `pMacroDefinition` in macro (for more direct parsing).
Note that this means that `macro` will now parse one
macro at a time, rather than parsing a whole group together.
+ Fixed failures on \ref{}, \label{} with `+raw_tex`. Now these
commands are parsed as raw if `+raw_tex`; otherwise, their argument
is parsed as a bracketed string.
+ Don't crash on empty `enumerate` environment (#3707).
+ Handle escaped `&` inside table cell (#3708).
+ Handle block structure inside table cells (#3709). `minipage` is no
longer required.
+ Handle some width specifiers on table columns (#3709). Currently
we only handle the form `0.9\linewidth`. Anything else would have
to be converted to a percentage, using some kind arbitrary assumptions
about line widths.
+ Make sure `\write18` is parsed as raw LaTeX. The change is in the
LaTeX reader's treatment of raw commands, but it also affects the
Markdown reader.
+ Fixed regression with starred environment names (#3803).
+ Handle optional args in raw `\titleformat` (#3804).
+ Improved heuristic for raw block/inline. An unknown command at the
beginning of the line that could be either block or inline is
treated as block if we have a sequence of block commands followed by
a newline or a `\startXXX` command (which might start a raw ConTeXt
+ Don't remove macro definitions from the output, even if
`Ext_latex_macros` is set, so that macros will be applied.
Since they're only applied to math in Markdown, removing the macros
can have bad effects. Even for math macros, keeping them should be
+ Removed `macro`. It is no longer necessary, since the
`rawLaTeXBlock` parser will parse macro definitions. This also avoids
the need for a separate `latexMacro` parser in the Markdown reader.
+ Use `label` instead of `data-label` for label in caption (#3639).
+ Fixed space after \figurename etc.
+ Resolve references to section numbers.
+ Fix `\let\a=0` case, with single character token.
+ Allow `@` as a letter in control sequences. `@` is commonly used
in macros using `\makeatletter`. Ideally we'd make the tokenizer
sensitive to `\makeatletter` and `\makeatother`, but until then this
seems a good change.
+ Track header numbers and correlate with labels.
+ Allow `]` inside group in option brackets (#3857).
+ lstinline with braces can be used (verb cannot be used with braces)
(Marc Schreiber, #3535).
+ Fix keyval function: pandoc did not parse options in braces correctly
(Marc Schreiber, #3642).
+ When parsing raw LaTeX commands, include trailing space (#1773).
Otherwise things like `\noindent foo` break and turn into
`\noindentfoo`. Affects `-f latex+raw_tex` and `-f markdown` (and other
formats that allow `raw_tex`).
+ Don't treat "..." as Quoted (#3958). This caused quotes to be omitted in
`\texttt` contexts.
+ Add tests for existing `\includegraphics` behaviour (Ben Firshman).
+ Allow space before `=` in bracketd options (Ben Firshman).
+ Be more forgiving in parsing command options. This was needed, for
example, to make some minted options work.
+ Strip off quotes in `\include` filenames.

* Added `Text.Pandoc.CSV`, simple (unexported) CSV parser.

* `Text.Pandoc.PDF`:

+ Got `--resource-path` working with PDF output (#852).
+ Fetch images when generating PDF via context (#3380).
To do this, we create the temp directory as a subdirectory
of the working directory. Since context mk IV by default looks
for images in the parent directory, this works.
+ Use `report` instead of `warn`, make it sensitive to verbosity settings.
+ Use `fillMediaBag` and `extractMedia` to extract media to temp dir.
This reduces code duplication.
+ `html2pdf`: use stdin instead of intermediate HTML file
+ Removed useless `TEXINPUTS` stuff for `context2pdf`. mkiv context
doesn't use `TEXINPUTS`.

* `Text.Pandoc.Pretty`:

+ Simplified definition of `realLength`.
+ Don't error for blocks of size < 1. Instead, resize to 1 (see #1785).

* `Text.Pandoc.MIME`:

+ Use `application/javascript` (not `application/x-javascript`).
+ Added `emf` to mimeTypes with type `application/x-msmetafile` (#1713).

* `Text.Pandoc.ImageSize`:

+ Improve SVG image size code (Marc Schreiber, #3580).
+ Make `imageSize` recognize basic SVG dimensions (Mauro Bieg, #3462).

* Use `Control.Monad.State.Strict` throughout. This gives 20-30% speedup
and reduction of memory usage in most of the writers.

* Use `foldrWithKey` instead of deprecated `foldWithKey`.

* `Text.Pandoc.SelfContained`:

+ Fixed problem with embedded fonts (#3629).
+ Refactored getData from `getDataURI` in `SelfContained`.
+ Don't use data URIs for script or style (#3423). Instead, just use
script or style tags with the content inside. The old method with
data URIs prevents certain optimizations outside pandoc. Exception:
data URIs are still used when a script contains `</script>` or a
style contains `</`.
+ SelfContained: Handle URL inside material retrieved from a URL
(#3629). This can happen e.g. with an @import of a google web font.
(What is imported is some CSS which contains an url reference
to the font itself.) Also, allow unescaped pipe (|) in URL.
+ Load resources from `data-src` (needed for lazy loading in
reveal.js slide shows).
+ Handle `data-background-image` attribute on section (#3979).

* `Text.Pandoc.Parsing`:

+ Added `indentWith` (Alexander Krotov, #3687).
+ Added `stateCitations` to `ParserState`.
+ Removed `stateChapters` from `ParserState`.
+ In `ParserState`, make `stateNotes'` a Map, add `stateNoteRefs`.
+ Added `gobbleSpaces` and `gobbleAtMostSpaces`.
+ Adjusted type of `insertIncludedFile` so it can be used with token
+ Replace old texmath macro stuff from Parsing. Use Macro from
Text.Pandoc.Readers.LaTeX.Types instead.
+ Export `insertIncludedFile`.
+ Added `HasLogMessages`, `logMessage`, `reportLogMessages` (#3447).
+ Replace partial with total function (Albert Krewinkel).
+ Introduce `HasIncludeFiles` type class (Albert Krewinkel). The
`insertIncludeFile` function is generalized to work with all parser
states which are instances of that class.
+ Add `insertIncludedFilesF` which returns F blocks (Albert Krewinkel).
The `insertIncludeFiles` function was generalized and renamed
to `insertIncludedFiles'`; the specialized versions are based on that.
+ `many1Till`: Check for the end condition before parsing (Herwig
Stuetz). By not checking for the end condition before the first
parse, the parser was applied too often, consuming too much of the
input. This only affects `many1Till p end` where `p` matches on a
prefix of `end`.
+ Provide `parseFromString` (#3690). This is a version of
`parseFromString` specialied to ParserState, which resets
`stateLastStrPos` at the end. This is almost always what we want.
This fixes a bug where `_hi_` wasn't treated as emphasis in the
following, because pandoc got confused about the position of the
last word: `- [o] _hi_`.
+ Added `takeP`, `takeWhileP` for efficient parsing of `[Char]`.
+ Fix `blanklines` documentation (Alexander Krotov, #3843).
+ Give less misleading line information with `parseWithString`.
Previously positions would be reported past the end of the chunk.
We now reset the source position within the chunk and report
positions "in chunk."
+ Add `anyLineNewline` (Alexander Krotov).
+ Provide shared F monad functions for Markdown and Org readers
(Albert Krewinkel). The `F` monads used for delayed evaluation
of certain values in the Markdown and Org readers are based on a
shared data type capturing the common pattern of both `F` types.
+ Add `returnF` (Alexander Krotov).
+ Avoid parsing `Notes:**` as a bare URI (#3570). This avoids parsing
bare URIs that start with a scheme + colon + `*`, `_`, or `]`.
+ Added `readerAbbreviations` to `ParserState`. Markdown reader
now consults this to determine what is an abbreviation.
+ Combine grid table parsers (Albert Krewinkel, #3638). The grid table
parsers for markdown and rst was combined into one single
parser `gridTable`, slightly changing parsing behavior of both
parsers: (1) The markdown parser now compactifies block content
cell-wise: pure text blocks in cells are now treated as paragraphs
only if the cell contains multiple paragraphs, and as plain blocks
otherwise. Before, this was true only for single-column tables. (2)
The rst parser now accepts newlines and multiple blocks in header
+ Generalize tableWith, gridTableWith (Albert Krewinkel).
The parsing functions `tableWith` and `gridTableWith` are generalized
to work with more parsers. The parser state only has to be an
instance of the `HasOptions` class instead of requiring a concrete
type. Block parsers are required to return blocks wrapped into a
monad, as this makes it possible to use parsers returning results
wrapped in `Future`s.

* `Text.Pandoc.Shared`:

+ Simplify `toRomanNumeral` using guards (Alexander Krotov, #3445)
+ `stringify`: handle Quoted better (#3958). Previously we were losing
the quotation marks in Quoted elements.

* `Text.Pandoc.Writers.Shared`:

+ Export `metaToJSON'`, `addVariablesToJSON` (#3439).
This allows us to add the variables AFTER using the metadata
to generate a YAML header (in the Markdown writer).
+ Added `unsmartify` (previously in RST writer).
Undo literal double curly quotes. Previously we left these.
+ Generalize type of `metaToJSON` so it can take a Text. Previously a
String was needed as argument; now any ToJSON instance will do.
+ Added `gridTable` (previously in Markdown writer).
+ `gridTable`: Refactored to use widths in chars.
+ `gridTable`: remove unnecessary extra space in cells.
+ Fixed `addVariablesToJSON`. It was previously not allowing multiple
values to become lists.
+ Pipe tables: impose minimum cell size (see #3526).

[default template changes]

* HTML templates (including EPUB and HTML slide show templates):

+ Make default.html5 polyglot markup conformant (John Luke Bentley,
#3473). Polyglot markup is HTML5 that is also valid XHTML. See
<>. With this change, pandoc's
html5 writer creates HTML that is both valid HTML5 and valid XHTML.
+ Regularized CSS in html/epub/html slide templates (#3485).
All templates now include `code{white-space: pre-wrap}`
and CSS for `q` if `--html-q-tags` is used. Previously some templates
had `pre` and others `pre-wrap`; the `q` styles were only sometimes
+ CSS for `.smallcaps`, (Mauro Bieg, #1592)
+ `default.revealjs`: make `history` default to true.
+ `default.revealjs`: use lazy loading (#2283).
+ `default.revealjs`: add `mathjax` variable and some conditional code
to use the MathJaX plugin.
+ `default.slidy` uses `https` instead of `http` (ickc, #3848).
+ `default.dzslides`: Load Google Font using HTTPS by default
(Yoan Blanc).

* DocBook5 template: Use `lang` and `subtitle` variables (Jens Getreu,

* LaTeX/Beamer template:

+ Combine LaTeX/Beamer templates (Andrew Dunning, #3878).
`default.beamer` has been removed; beamer now uses the
`default.latex` template. Beamer-specific parts are conditional
on the `beamer` variable set by the writer. Note that
`pandoc -D beamer` will return this (combined) template.
+ Use `xcolor` for `colorlinks` option (Andrew Dunning, #3877).
Beamer loads `xcolor` rather than `color`, and thus the
`dvipsnames` option doesn't take effect. This also provides a wider
range of colour selections with the `svgnames` option.
+ Use starred versions of `xcolor` names (Andrew Dunning).
Prevents changes to documents defined using the `dvipsnames` list (e.g.
`Blue` gives a different result with svgnames enabled).
+ Load `polyglossia` after header-includes (#3898). It needs to be
loaded as late as possible.
+ Use `unicode-math` (Vaclav Haisman). Use `mathspec` with only
XeLaTeX on request.
+ Don't load `fontspec` before `unicode-math` (over there).
The `unicode-math` package loads `fontspec` so explicit loading of
`fontspec` before `unicode-math` is not necessary.
+ Use `unicode-math` by default in default.latex template. mathspec will
be used in xelatex if the `mathspec` variable is set; otherwise
unicode-math will be used (Václav Haisman).
+ Use `dvipsnames` options when `colorlinks` specified (otherwise
we get an error for `maroon`) (Thomas Hodgson).
+ Added beamer `titlegraphic` and `logo` variables (Thomas Hodgson).
+ Fix typo in fix for notes in tables (#2378, zeeMonkeez).
+ Fix `hyperref` options clash (Andrew Dunning, #3847) Avoids an options
clash when loading a package (e.g. `tufte-latex`) that uses
`hyperref` settings different from those in the template.
+ Add `natbiboptions` variable (#3768).
+ Fix links inside captions in LaTeX output with links-as-notes
(Václav Haisman, #3651). Declare our redefined `\href` robust.
+ Load `parskip` before `hyperref` (Václav Haisman, #3654).
+ Allow setting Japanese fonts when using LuaLaTeX (Václav Haisman,
#3873). by using the `luatexja-fontspec` and `luatexja-preset`
packages. Use existing `CJKmainfont` and `CJKoptions` template
variables. Add `luatexjafontspecoptions` for `luatexja-fontspec`
and `luatexjapresetoptions` for `luatexja-preset`.
+ Added `aspectratio` variable to beamer template (Václav Haisman,
+ Modified template.latex to fix XeLaTex being used with tables
(lwolfsonkin, #3661). Reordered `lang` variable handling to
immediately before `bidi`.

* ConTeXt template: Improved font handling: `simplefonts` is now
obsolete in ConTeXt (Pablo Rodríguez).

[documentation improvements]

* MANUAL.txt:

+ Add URL for Prince HTML > PDF engine (Ian, #3919).
+ Document that content above slide-level will be omitted in
slide shows. See #3460, #2265.
+ Explain `--webtex` SVG url (Mauro Bieg, #3471)
+ Small clarification in YAML metadata section.
+ Document that html4 is technically XHTML 1.0 transitional.
+ Remove refs to highlighting-kate (#3672).
+ Document ibooks specific epub metadata.
+ Clarify that mathml is used for ODT math.
+ Mention limitations of Literate Haskell Support (#3410,
Joachim Breitner).
+ Add documentation of limitations of grid tables (Stephen
McDowell, #3864).
+ Clarify that meta-json contains transformed values (Jakob Voß,
#3491) Make clear that template variable `meta-json` does not
contain plain text values or JSON output format but field values
transformed to the selected output format.


+ Clarify that templates are dual-licensed.
+ Clarify that pandoc-types is BSD3 licensed.
+ List new files not written by jgm (Albert Krewinkel).
+ Update dates in copyright notices (Albert Krewinkel). This follows
the suggestions given by the FSF for GPL licensed software.


+ Improved instructions for tests with patterns.
+ Put RPM-based distros on separate point (Mauro Bieg, #3449)


+ Fixed typos (Wandmalfarbe, #3479).
+ Add "ask on pandoc-discuss" (Mauro Bieg).

* Add lua filter documentation in `doc/`. Note that the
end of this document is autogenerated from `data/pandoc.lua`
using `make doc/`, which uses `tools/ldoc.ltp`
(Albert Krewinkel).

* Add `doc/`. This is the old scripting tutorial from
the website.

* Add `doc/` (#3289). This gives an introduction
to using pandoc as a Haskell library.

[build infrastructure improvements]

* Removed `data/templates` submodule. Templates are now a subtree
in `data/templates`. This removes the need to do `git submodule

* Renamed `tests` -> `test`.

* Remove `https` flag. Always build with HTTPS support.

* Use `file-embed` instead of `hsb2hs` to embed data files when
`embed_data_files` flag is set. `file-embed` gives us better dependency
tracking: if a data file changes, ghc/stack/cabal know to recompile
the Data module. This also removes `hsb2hs` as a build dependency.

* Add `custom-setup` stanza to pandoc, lowercase field names.

* Add `static` Cabal flag.

* Name change OSX -> MacOS. Add a -MacOS suffix to mac package rather
than -OSX. Changed local names from osx to macos.

* - Use strip to reduce executable size.

* Revised binary linux package. Now a completely static executable
is created, using Docker and alpine. We create both a deb and a
tarball. The old `deb` directory has been replaced with a `linux`
directory. Running `make` in the `linux` directory should
perform the build, putting the binary packages in `artifacts/`.

* `linux/`: add `Replaces:`, so existing pandoc-citeproc and
pandoc-data packages will be uninstalled; this package provides
both (#3822). Add latex packages as 'suggested', update

* Remove cpphs build requirement -- it is no longer needed.

* Replaced `{deb,macos,windows}/stack.yaml` with `stack.pkg.yaml`.

* Name change OSX -> macOS (ickc, #3869).

* Fix casing of Linux, UNIX, and Windows (ickc).

* `.travis.yml`: create a source dist and do cabal build and test there.
That way we catch errors due to files missing from the data
section of pandoc.cabal.

* Makefile:

+ Split `make haddock` from `make full`.
+ Add BRANCH variable for winpkg.
+ Add `lint` target.
+ Improve `make full`. Disable optimizations.
Build everything, inc. trypandoc and benchmarks. Use parallel build.
+ Allow `make test` to take `TESTARGS`.

* Added new command tests (`Tests.Command`), using small text files
in `test/command/`. Any files added in this directory will be treated
as shell tests (see for an example). This makes it very easy
to add regression tests etc.

* Test fixes so we can find data files. In old tests & command tests,
we now set the environment variable `pandoc_datadir`. In lua tests,
we set the datadir explicitly.

* Refactored `compareOutput` in docx writer test.

* Consolidated some common functions in `Tests.Helper`.

* Small change to unbalanced bracket test to speed up test suite.

* Speed up Native writer quickcheck tests.

* Use tasty for tests rather than test-framework.

* Add simple Emacs mode to help with Pandoc templates editing.
(Václav Haisman, #3889). `tools/pandoc-template-mode.el`

pandoc (1.19.2)

* Use skylighting library instead of highlighting-kate for syntax
highlighting. Skylighting is faster and more accurate (#3363).
Later we'll be able to add features like warning messages, dynamic
loading of xml syntax definitions, and dynamic loading of themes.

* Added a new highlight style, `breezeDark`.

* Text.Pandoc.Highlighting: Update list of `listings` languages (#3374).
This allows more languages to be used when using the `--listings`

* OpenDocument writer:

+ Small refactoring. Removed separate 'parent' parameter in paraStyle.
+ Don't profilerate text styles unnecessarily (#3371).
This change makes the writer create only as many temporary
text styles as are absolutely necessary. It also consolidates
adjacent nodes with the same style.

* Org reader (Albert Krewinkel):

+ Allow short hand for single-line raw blocks (Albert Krewinkel,
#3366). Single-line raw blocks can be given via `#+FORMAT: raw line`,
where `FORMAT` must be one of `latex`, `beamer`, `html`, or `texinfo`.
+ Accept org-ref citations followed by commas (Albert Krewinkel).
Bugfix for an issue which, whenever the citation was immediately
followed by a comma, prevented correct parsing of org-ref citations.
+ Ensure emphasis markup can be nested. Nested emphasis markup (e.g.
`/*strong and emphasized*/`) was interpreted incorrectly in that the
inner markup was not recognized.
+ Remove pipe char irking the haddock coverage tool (Albert Krewinkel).

* Docx reader: Empty header should be list of lists (Jesse Rosenthal).
In the past, the docx reader wrote an empty header as an empty list. It
should have the same width as a row (and be filled with empty cells).

* MediaWiki reader:

+ Improved handling of display math (#3362). Sometimes display math is
indented with more than one colon. Previously we handled these cases
badly, generating definition lists and missing the math.
+ Fix quotation mark parsing (#3336, tgkokk). Change MediaWiki reader's
behavior when the smart option is parsed to match other readers'

* Markdown reader:

+ Fixed `-f markdown_github-hard_line_breaks+escaped_line_breaks`
(#3341). Previously this did not properly enable escaped line breaks.
+ Disallow space between inline code and attributes (#3326, #3323,
Mauro Bieg).

* DocBook5 writer: make id attribute xml:id, fixes #3329 (#3330, Mauro Bieg).

* Added some test cases for ODT reader (#3306, #3308, Hubert Plociniczak).

* LaTeX writer: allow tables with empty cells to count as "plain."
This addresses a problem of too-wide tables when empty cells
are used. Thanks to Joost Kremers for reporting the issue.

* Org writer: prefix footnote numbers with `fn:` (Albert Krewinkel).
Unprefixed numbers where used by older org-mode versions, but are no
longer supported.

* HTML writer: don't process pars with empty RawInline, (#1040, #3327,
Mauro Bieg).

* Markdown writer: Fix display math with `--webtex` (#3298).

* Fix sample.lua so it properly handles raw blocks/inlines (#3358,

* Templates:

+ default.latex: Moved geometry after hyperref (Václav Haisman).
Otherwise PDF sizes can be wrong in some circumstances.
+ Copied a few changes from default.latex to default.beamer
+ default.latex, default.beamer: Changed position of `\VerbatimNotes`
and `fancyvrb`. This fixes hyperlinks on footnotes in documents
that contain verbatim in notes (#3361). (Note: the beamer template
was updated to match the LaTeX template, but at this point verbatim
in notes seems not to work in beamer.)
+ default.latex: Allow passing `microtypeoptions` to microtype
(Václav Haisman).
+ default.latex: Add hyphen option to url package.
+ default.docbook5: Fix namespace declarations (Mauro Bieg).

* Moved `` to `osx/` directory.

* Travis continuous integration:

+ Fix false positives with dist build.
+ Speed improvements (Kolen Cheung, #3304, #3357).

* MANUAL.txt:

+ Clarify that blank space is needed around footnotes (#3352).
+ Fixed typo (#3351, Alexey Rogechev).
+ Note that `--wrap=auto` does not work in HTML output.
+ Default `--columns` width is 72, not 80.
+ Fixed broken links (#3316, Kolen Cheung).
+ Document usage of `@*` in nocite section (#3333, John Muccigrosso).


+ Indent code so it's properly formatted (#3335, Bheesham Persaud).
+ Added instructions for extracting binary from OSX, Windows packages.

* Describe labels currently used in issue tracker
(Albert Krewinkel). The labels have changed over time, the list of
labels is updated to reflect the current set of labels used in the
issue tracker.

* Rearrange and extend badges in README (Albert Krewinkel, #3354)

* Bumped version bounds for dependencies.

pandoc (1.19.1)

* Set `PANDOC_VERSION` environment variable for filters (#2640).
This allows filters to check the pandoc version that produced
the JSON they are receiving.

* Docx reader: Ensure one-row tables don't have header (#3285,
Jesse Rosenthal). Tables in MS Word are set by default to have
special first-row formatting, which pandoc uses to determine whether
or not they have a header. This means that one-row tables will, by
default, have only a header -- which we imagine is not what people
want. This change ensures that a one-row table is not understood to
be a header only. Note that this means that it is impossible to
produce a header-only table from docx, even though it is legal
pandoc. But we believe that in nearly all cases, it will be an
accidental (and unwelcome) result

* HTML reader:

+ Fixed some bad regressions in HTML table parser (#3280).
This regression leads to the introduction of empty rows
in some circumstances.
+ Understand `style=width:` as well as `width` in `col` (#3286).

* RST reader:

+ Print warnings when keys, substitition, notes not found.
Previously the parsers failed and we got raw text. Now we get a
link with an empty URL, or empty inlines in the case of a note or

+ Fix hyperlink aliases (#3283).

* Man writer: Ensure that periods are escaped at beginning of line

* LaTeX writer: Fix unnumbered headers when used with `--top-level`
(#3272, Albert Krewinkel). Fix interaction of top-level divisions
`part` or `chapter` with unnumbered headers when emitting LaTeX. Headers
are ensured to be written using stared commands (like `\subsection*{}`).

* LaTeX template: use comma not semicolon to separate keywords for
`pdfkeywords`. Thanks to Wandmalfarbe.

* Markdown writer: Fixed incorrect word wrapping (#3277). Previously pandoc
would sometimes wrap lines too early due to this bug.

* Text.Pandoc.Pretty: Added `afterBreak` [API change]. This makes it
possible to insert escape codes for content that needs escaping at the
beginning of a line.

* Removed old MathMLInHTML.js from 2004, which should no longer
be needed for MathML with modern browsers.

* Fixed tests with dynamic linking (#2709).

* Makefile: Use stack instead of cabal for targets. This is just
a convenience for developers.

* Fixed bash completion of filenames with space (#2749).

* MANUAL: improved documentation on how to create a custom `reference.docx`.

* Fix minor spelling typos in the manual (#3273, Anthony Geoghegan)

pandoc (1.19)

* Changed resolution of filter paths.

+ We now first treat the argument of `--filter` as a full (absolute
or relative) path, looking for a program there. If it's found, we
run it.
+ If not, and if it is a simple program name or a relative path, we
try resolving it relative to `$DATADIR/filters`.
+ If this fails, then we treat it as a program name and look in the
user's PATH.
+ Removed a hardcoded '/' that may have caused problems with
Windows paths.

Previously if you did `--filter foo` and you had `foo` in your path and
also an executable `foo` in your working directory, the one in the path
would be used. Now the one in the working directory is used.

In addition, when you do `--filter foo/bar.hs`, pandoc will now find a
filter `$DATADIR/filters/foo/bar.hs` -- assuming there isn't a
`foo/bar.hs` relative to the working directory.

* Allow `file://` URIs as arguments (#3196). Also improved default reader
format detection. Previously with a URI ending in .md or .markdown,
pandoc would assume HTML input. Now it treats these as markdown.

* Allow to overwrite top-level division type heuristics (#3258,
Albert Krewinkel). Pandoc uses heuristics to determine the most
reasonable top-level division type when emitting LaTeX or
Docbook markup. It is now possible to overwrite this implicitly set
top-level division via the `top-level-division` command line parameter.

* Text.Pandoc.Options \[API changes\]:

+ Removed `writerStandalone` field in `WriterOptions`, made
`writerTemplate` a `Maybe` value. Previously setting
`writerStandalone = True` did nothing unless a template was provided
in writerTemplate. Now a fragment will be generated if
`writerTemplate` is `Nothing`; otherwise, the specified template
will be used and standalone output generated.
+ `Division` has been renamed `TopLevelDivision` (#3197). The
`Section`, `Chapter`, and `Part` constructors were renamed to
`TopLevelSection`, `TopLevelChapter`, and
`TopLevelPart`, respectively. An additional `TopLevelDefault`
constructor was added, which is now also the new default value of
the `writerTopLevelDivision` field in `WriterOptions`.

* Improved error if they give wrong arg to `--top-level-division`.

* Use new module from texmath to lookup MS font codepoints in Docx reader.
Removed unexported module Text.Pandoc.Readers.Docx.Fonts. Its code now
lives in texmath (0.9).

* DocBook reader: Fixed xref lookup (#3243). It previously only worked
when the qnames lacked the docbook namespace URI.

* HTML reader:

+ Improved table parsing (#3027). We now check explicitly for non-1
rowspan or colspan attributes, and fail when we encounter them.
Previously we checked that each row had the same number of cells,
but that could be true even with rowspans/colspans. And there are
cases where it isn't true in tables that we can handle fine -- e.g.
when a tr element is empty. So now we just pad rows with empty cells
when needed.
+ Treat `<math>` as MathML by default unless something else is
explicitly specified in xmlns. Provided it parses as MathML,
of course. Also fixed default which should be to inline math if no
display attribute is used.
+ Only treat "a" element as link if it has href (#3226). Otherwise
treat as span.

* Docx reader (Jesse Rosenthal):

+ Add a placeholder value for CHART. We wrap `[CHART]` in a
`<span class="chart">`. Note that it maps to inlines because, in
docx, anything in a drawing tag can be part of a larger paragraph.
+ Be more specific in parsing images We not only want `w:drawing`,
because that could also include charts. Now we specify
`w:drawing/pic:pic`. This shouldn't change behavior at all, but it's
a first step toward allowing other sorts of drawing data as well.
+ Abstract out function to avoid code repetition.
+ Update tests for img title and alt (#3204).
+ Handle Alt text and titles in images. We use the "description" field
as alt text and the "title" field as title. These can be accessed
through the "Format Picture" dialog in Word.
+ Docx reader utils: handle empty namespace in `elemName`. Previously,
if given an empty namespace `(elemName ns "" "foo")` `elemName`
would output a QName with a `Just ""` namespace. This is never what
we want. Now we output a `Nothing`. If someone *does* want a
`Just ""` in the namespace, they can enter the QName
value explicitly.

* ODT reader/writer:

+ Inline code when text has a special style (Hubert Plociniczak). When
a piece of text has a text `Source_Text` then we assume that this is
a piece of the document that represents a code that needs to
be inlined. Adapted the writer to also reflect that change.
Previously it was just writing a 'preformatted' text using a
non-distinguishable font style. Code blocks are still not recognized
by the ODT reader. That's a separate issue.
+ Infer table's caption from the paragraph (#3224,
Hubert Plociniczak). ODT's reader always put empty captions for the
parsed tables. This commit

1. checks paragraphs that follow the table definition
2. treats specially a paragraph with a style named 'Table'
3. does some postprocessing of the paragraphs that combines tables
followed immediately by captions

The ODT writer used the `TableCaption` style for the caption
paragraph. This commit follows the OpenOffice approach which allows
for appending captions to table but uses a built-in style named
`Table` instead of `TableCaption`. Users of a custom `reference.odt`
should change the style's name from `TableCaption` to `Table`.

* ODT reader: Infer tables' header props from rows (#3199,
Hubert Plociniczak). ODT reader simply provided an empty header list
which meant that the contents of the whole table, even if not empty, was
simply ignored. While we still do not infer headers we at least have to
provide default properties of columns.

* Markdown reader:

+ Allow reference link labels starting with `@...` if `citations`
extension disabled (#3209). Example: in

\[link text\]\[@a\]

`link text` isn't hyperlinked because `[@a]` is parsed as
a citation. Previously this happened whether or not the `citations`
extension was enabled. Now it happens only if the `citations`
extension is enabled.
+ Allow alignments to be specified in Markdown grid tables. For

| Right | Left | Centered |
| Bananas | $1.34 | built-in wrapper |

+ Allow Small Caps elements to be created using bracketed spans (as
they already can be using HTML-syntax spans) (#3191, Kolen Cheung).

* LaTeX reader:

+ Don't treat `\vspace` and `\hspace` as block commands (#3256).
Fixed an error which came up, for example, with `\vspace` inside
a caption. (Captions expect inlines.)
+ Improved table handling. We can now parse all of the tables emitted
by pandoc in our tests. The only thing we don't get yet are
alignments and column widths in more complex tables. See #2669.
+ Limited support for minipage.
+ Allow for `[]`s inside LaTeX optional args. Fixes cases like:
+ Handle BVerbatim from fancyvrb (#3203).
+ Handle hungarumlaut (#3201).
+ Allow beamer-style `<...>` options in raw LaTeX (also in Markdown)
(#3184). This allows use of things like `\only<2,3>{my content}` in
Markdown that is going to be converted to beamer.

* Use pre-wrap for code in dzslides template (Nicolas Porcel). Otherwise
overly long code will appear on every slide.

* Org reader (Albert Krewinkel):

+ Respect column width settings (#3246). Table column properties can
optionally specify a column's width with which it is displayed in
the buffer. Some exporters, notably the ODT exporter in org-mode
v9.0, use these values to calculate relative column widths. The org
reader now implements the same behavior. Note that the org-mode
LaTeX and HTML exporters in Emacs don't support this feature yet,
which should be kept in mind by users who use the column
widths parameters.
+ Allow HTML attribs on non-figure images (#3222). Images which are
the only element in a paragraph can still be given HTML attributes,
even if the image does not have a caption and is hence not a figure.
The following will add set the `width` attribute of the image to

+ATTR\_HTML: :width 50%


+ Support `ATTR_HTML` for special blocks (#3182). Special
blocks (i.e. blocks with unrecognized names) can be prefixed with an
`ATTR_HTML` block attribute. The attributes defined in that
meta-directive are added to the `Div` which is used to represent the
special block.
+ Support the `todo` export option. The `todo` export option allows to
toggle the inclusion of TODO keywords in the output. Setting this to
`nil` causes TODO keywords to be dropped from headlines. The default
is to include the keywords.
+ Add support for todo-markers. Headlines can have optional
todo-markers which can be controlled via the `#+TODO`, `#+SEQ_TODO`,
or `#+TYP_TODO` meta directive. Multiple such directives can be
given, each adding a new set of recognized todo-markers. If no
custom todo-markers are defined, the default `TODO` and `DONE`
markers are used. Todo-markers are conceptually separate from
headline text and are hence excluded when autogenerating
headline IDs. The markers are rendered as spans and labelled with
two classes: One class is the markers name, the other signals the
todo-state of the marker (either `todo` or `done`).

* LaTeX writer:

+ Use `\autocites*` when "suppress-author" citation used.
+ Ensure that simple tables have simple cells (#2666). If cells
contain more than a single Plain or Para, then we need to set
nonzero widths and put contents into minipages.
+ Remove invalid inlines in sections (#3218, Hubert Plociniczak).

* Markdown writer:

+ Fix calculation of column widths for aligned multiline tables
(#1911, Björn Peemöller). This also fixes excessive CPU and memory
usage for tables when `--columns` is set in such a way that cells
must be very tiny. Now cells are guaranteed to be big enough so that
single words don't need to line break, even if this pushes the line
length above the column width.
+ Use bracketed form for native spans when `bracketed_spans`
enabled (#3229).
+ Fixed inconsistent spacing issue (#3232). Previously a tight bullet
sublist got rendered with a blank line after, while a tight ordered
sublist did not. Now we don't get the blank line in either case.
+ Fix escaping of spaces in super/subscript (#3225). Previously two
backslashes were inserted, which gave a literal backslash.
+ Adjust widths in Markdown grid tables so that they match
on round-trip.

* Docx writer:

+ Give full detail when there are errors converting tex math.
+ Handle title text in images (Jesse Rosenthal). We already handled
alt text. This just puts the image "title" into the docx
"title" attr.
+ Fixed XML markup for empty cells (#3238). Previously the Compact
style wasn't being applied properly to empty cells.

* HTML writer:

+ Updated `renderHtml` import from blaze-html.

* Text.Pandoc.Pretty:

+ Fixed some bugs that caused blank lines in tables (#3251). The bugs
caused spurious blank lines in grid tables when we had things like
`blankline $$ blankline`.
+ Add exported function `minOffet` \[API change\] (Björn Peemöller).
+ Added error message for illegal call to `block` (Björn Peemöller).

* Text.Pandoc.Shared:

+ Put `warn` in MonadIO.
+ `fetchItem`: Better handling of protocol-relative URL (#2635). If
URL starts with `//` and there is no "base URL" (as there would be
if a URL were used on the command line), then default to http:.

* Export Text.Pandoc.getDefaultExtensions \[API change\] (#3178).

* In --version, trap error in `getAppUserDataDirectory` (#3241). This
fixes a crash with `pandoc --version` on unusual systems with no real
user (e.g. SQL Server 2016).

* Added weigh-pandoc for memory usage diagnostics (#3169).

* Use correct mime types for woff and woff2 (#3228).

* Remove make\_travis\_yml.hs (#3235, Kolen Cheung).

* changelog: Moved an item that was misplaced in the 1.17.2 section to the
1.18 section where it belongs.

* minor change in wording and punctuation (#3252,
Kolen Cheung).

* Further revisions to manual for `--version` changes (#3244).

pandoc (1.18)

* Added `--list-input-formats`, `--list-output-formats`,
`--list-extensions`, `--list-highlight-languages`, and
`--list-highlight-styles` (#3173). Removed list of highlighting
languages from `--version` output. Removed list of input and output
formats from default `--help` output.

* Added `--reference-location=block|section|document` option
(Jesse Rosenthal). This determines whether Markdown link references
and footnotes are placed at the end of the document, the end of the
section, or the end of the top-level block.

* Added `--top-level-division=section|chapter|part` (Albert Krewinkel).
This determines what a level-1 header corresponds to in LaTeX,
ConTeXt, DocBook, and TEI output. The default is `section`.
The `--chapters` option has been deprecated in favor of

* Added `LineBlock` constructor for `Block` (Albert Krewinkel). This
is now used in parsing RST and Markdown line blocks, DocBook
`linegroup`/`line` combinations, and Org-mode `VERSE` blocks.
Previously `Para` blocks with hard linebreaks were used. `LineBlock`s
are handled specially in the following output formats: AsciiDoc
(as `[verse]` blocks), ConTeXt (`\startlines`/`\endlines`),
HTML (`div` with a style), Markdown (line blocks if `line_blocks`
is enabled), Org-mode (`VERSE` blocks), RST (line blocks). In
other output formats, a paragraph with hard linebreaks is emitted.

* Allow binary formats to be written to stdout (but not to tty) (#2677).
Only works on posix, since we use the unix library to check whether
output is to tty. On Windows, pandoc works as before and always requires
an output file parameter for binary formats.

* Changed JSON output format (Jesse Rosenthal). Previously we used
generically generated JSON, but this was subject to change depending
on the version of aeson pandoc was compiled with. To ensure stability,
we switched to using manually written ToJSON and FromJSON
instances, and encoding the API version. **Note:** pandoc filter
libraries will need to be revised to handle the format change.
Here is a summary of the essential changes:

+ The toplevel JSON format is now `{"pandoc-api-version" :
[MAJ, MIN, REV], "meta" : META, "blocks": BLOCKS}`
instead of `[{"unMeta": META}, [BLOCKS]]`.
Decoding fails if the major and minor version numbers don't
+ Leaf nodes no longer have an empty array for their "c" value.
Thus, for example, a `Space` is encoded as `{"t":"Space"}`
rather than `{"t":"Space","c":[]}` as before.

* Removed `tests/Tests/Arbitrary.hs` and added a `Text.Pandoc.Arbitrary`
module to pandoc-types (Jesse Rosenthal). This makes it easier
to use QuickCheck with pandoc types outside of pandoc itself.

* Add `bracketed_spans` Markdown extension, enabled by default
in pandoc `markdown`. This allows you to create a native span
using this syntax: `[Here is my span]{#id .class key="val"}`.

* Added `angle_brackets_escapable` Markdown extension (#2846).
This is needed because github flavored Markdown has a slightly
different set of escapable symbols than original Markdown;
it includes angle brackets.

* Export `Text.Pandoc.Error` in `Text.Pandoc` [API change].

* Print highlighting-kate version in `--version`.

* `Text.Pandoc.Options`:

+ `Extension` has new constructors `Ext_brackted_spans` and
`Ext_angle_brackets_escapable` [API change].
+ Added `ReferenceLocation` type [API change] (Jesse Rosenthal).
+ Added `writerReferenceLocation` field to `WriterOptions` (Jesse

* `--filter`: we now check `$DATADIR/filters` for filters before
looking in the path (#3127, Jesse Rosenthal, thanks to Jakob
Voß for the idea). Filters placed in this directory need not
be executable; if the extension is `.hs`, `.php`, `.pl`, `.js`,
or `.rb`, pandoc will run the right interpreter.

* For `--webtex`, replace deprecated Google Chart API by CodeCogs as
default (Kolen Cheung).

* Removed `raw_tex` extension from `markdown_mmd` defaults (Kolen Cheung).

* Execute .js filters with node (Jakob Voß).

* Textile reader:

+ Support `bc..` extended code blocks (#3037). Also, remove trailing
newline in code blocks (consistently with Markdown reader).
+ Improve table parsing. We now handle cell and row attributes, mostly
by skipping them. However, alignments are now handled properly.
Since in pandoc alignment is per-column, not per-cell, we
try to devine column alignments from cell alignments.
Table captions are also now parsed, and textile indicators
for thead and tfoot no longer cause parse failure. (However,
a row designated as tfoot will just be a regular row in pandoc.)
+ Improve definition list parsing. We now allow multiple terms
(which we concatenate with linebreaks). An exponential parsing
bug (#3020) is also fixed.
+ Disallow empty URL in explicit link (#3036).

* RST reader:

+ Use Div instead of BlockQuote for admonitions (#3031).
The Div has class `admonition` and (if relevant) one of the
following: `attention`, `caution`, `danger`, `error`, `hint`,
`important`, `note`, `tip`, `warning`. **Note:** This will change
the rendering of some RST documents! The word ("Warning", "Attention",
etc.) is no longer added; that must be done with CSS or a filter.
+ A Div is now used for `sidebar` as well.
+ Skip whitespace before note (Jesse Rosenthal, #3163). RST requires a
space before a footnote marker. We discard those spaces so that footnotes
will be adjacent to the text that comes before it. This is in line with
what rst2latex does.
+ Allow empty lines when parsing line blocks (Albert Krewinkel).

* Markdown reader:

+ Allow empty lines when parsing line blocks (Albert Krewinkel).
+ Allow attributes on autolinks (#3183, Daniele D'Orazio).

* LaTeX reader:

+ More robust parsing of unknown environments (#3026).
We no longer fail on things like `^` inside options for tikz.
+ Be more forgiving of non-standard characters, e.g. `^` outside of math.
Some custom environments give these a meaning, so we should try not to
fall over when we encounter them.
+ Drop duplicate `*` in bibtexKeyChars (Albert Krewinkel)

* MediaWiki reader:

+ Fix for unquoted attribute values in mediawiki tables (#3053).
Previously an unquoted attribute value in a table row
could cause parsing problems.
+ Improved treatment of verbatim constructions (#3055).
Previously these yielded strings of alternating Code and Space
elements; we now incorporate the spaces into the Code. Emphasis
etc. is still possible inside these.
+ Properly interpret XML tags in pre environments (#3042). They are meant
to be interpreted as literal text.

* EPUB reader: don't add root path to data: URIs (#3150).
Thanks to @lep for the bug report and patch.

* Org reader (Albert Krewinkel):

+ Preserve indentation of verse lines (#3064). Leading spaces in verse
lines are converted to non-breaking spaces, so indentation is preserved.
+ Ensure image sources are proper links. Image sources as those in plain
images, image links, or figures, must be proper URIs or relative file
paths to be recognized as images. This restriction is now enforced
for all image sources. This also fixes the reader's usage of uncleaned
image sources, leading to `file:` prefixes not being deleted from
figure images. Thanks to @bsag for noticing this bug.
+ Trim verse lines properly (Albert Krewinkel).
+ Extract meta parsing code to module. Parsing of meta-data is well
separable from other block parsing tasks. Moving into new module to
get small files and clearly arranged code.
+ Read markup only for special meta keys. Most meta-keys should be read
as normal string values, only a few are interpreted as marked-up text.
+ Allow multiple, comma-separated authors. Multiple authors can be
specified in the `#+AUTHOR` meta line if they are given as a
comma-separated list.
+ Give precedence to later meta lines. The last meta-line of any given
type is the significant line. Previously the value of the first line
was kept, even if more lines of the same type were encountered.
+ Read LaTeX_header as header-includes. LaTeX-specific header commands
can be defined in `#+LaTeX_header` lines. They are parsed as
format-specific inlines to ensure that they will only show up in LaTeX
+ Set documentclass meta from LaTeX_class.
+ Set classoption meta from LaTeX_class_options.
+ Read HTML_head as header-includes. HTML-specific head content can be
defined in `#+HTML_head` lines. They are parsed as format-specific
inlines to ensure that they will only show up in HTML output.
+ Respect `author` export option. The `author` option controls whether
the author should be included in the final markup. Setting
`#+OPTIONS: author:nil` will drop the author from the final meta-data
+ Respect `email` export option. The `email` option controls whether the
email meta-field should be included in the final markup. Setting
`#+OPTIONS: email:nil` will drop the email field from the final
meta-data output.
+ Respect `creator` export option. The `creator` option controls whether
the creator meta-field should be included in the final markup. Setting
`#+OPTIONS: creator:nil` will drop the creator field from the final
meta-data output. Org-mode recognizes the special value `comment` for
this field, causing the creator to be included in a comment. This is
difficult to translate to Pandoc internals and is hence interpreted the
same as other truish values (i.e. the meta field is kept if it's
+ Respect unnumbered header property (#3095). Sections the `unnumbered`
property should, as the name implies, be excluded from the automatic
numbering of section provided by some output formats. The Pandoc
convention for this is to add an "unnumbered" class to the header. The
reader treats properties as key-value pairs per default, so a special
case is added to translate the above property to a class instead.
+ Allow figure with empty caption (Albert Krewinkel, #3161).
A `#+CAPTION` attribute before an image is enough to turn an image into
a figure. This wasn't the case because the `parseFromString` function,
which processes the caption value, would fail on empty values. Adding
a newline character to the caption value fixes this.

* Docx reader:

+ Use XML convenience functions (Jesse Rosenthal).
The functions `isElem` and `elemName` (defined in Docx/Util.hs) make
the code a lot cleaner than the original XML.Light functions, but they
had been used inconsistently. This puts them in wherever applicable.
+ Handle anchor spans with content in headers. Previously, we would only
be able to figure out internal links to a header in a docx if the
anchor span was empty. We change that to read the inlines out of the
first anchor span in a header.
+ Let headers use existing id. Previously we always generated an id for
headers (since they wouldn't bring one from Docx). Now we let it use an
existing one if possible. This should allow us to recurs through anchor
+ Use all anchor spans for header ids. Previously we only used the first
anchor span to affect header ids. This allows us to use all the anchor
spans in a header, whether they're nested or not (#3088).
+ Test for nested anchor spans in header. This ensures that anchor spans
in header with content (or with other anchor spans inside) will resolve
to links to a header id properly.

* ODT reader (Hubert Plociniczak)

+ Include list's starting value. Previously the starting value of
the lists' items has been hardcoded to 1. In reality ODT's list
style definition can provide a new starting value in one of its
+ Infer caption from the text following the image.
Frame can contain other frames with the text boxes.
+ Add `fig:` to title for Image with a caption (as expected
by pandoc's writers).
+ Basic support for images in ODT documents.
+ Don't duplicate text for anchors (#3143). When creating an anchor
element we were adding its representation as well as the original
content, leading to text duplication.

* DocBook writer:

+ Include an anchor element when a div or span has an id (#3102).
Note that DocBook does not have a class attribute, but at least this
provides an anchor for internal links.

* LaTeX writer:

+ Don't use * for unnumbered paragraph, subparagraph. The starred
variants don't exist. This helps with part of gets rid of
the spurious `*`s. But we still have numbers on the 4th and 5th level
+ Properly escape backticks in verbatim (#3121, Jesse Rosenthal).
Otherwise they can cause unintended ligatures like `` ?` ``.
+ Handle NARRAOW NO-BREAK SPACE into LaTeX (Vaclav Zeman) as `\,`.
+ Don't include `[htbp]` placement for figures (#3103, Václav Haisman).
This allows figure placement defaults to be changed by the user
in the template.

* HTML writer (slide show formats): In slide shows, don't change slide title
to level 1 header (#2221).

* TEI writer: remove heuristic to detect book template (Albert Krewinkel).
TEI doesn't have `<book>` elements but only generic `<divN>` division
elements. Checking the template for a trailing `</book>` is nonsensical.

* MediaWiki writer: transform filename with underscores in images (#3052).
`foo bar.jpg` becomes `foo_bar.jpg`. This was already done
for internal links, but it also needs to happen for images.

* ICML writer: replace partial function (!!) in table handling (#3175,
Mauro Bieg).

* Man writer: allow section numbers that are not a single digit (#3089).

* AsciiDoc writer: avoid unnecessary use of "unconstrained" emphasis
(#3068). In AsciiDoc, you must use a special form of emphasis
(double `__`) for intraword emphasis. Pandoc was previously using
this more than necessary.

* EPUB writer: use stringify instead of plain writer for metadata
(#3066). This means that underscores won't be used for emphasis,
or CAPS for bold. The metadata fields will just have unadorned

* Docx Writer:

+ Implement user-defined styles (Jesse Rosenthal). Divs and Spans
with a `custom-style` key in the attributes will apply the corresponding
key to the contained blocks or inlines.
+ Add ReaderT env to the docx writer (Jesse Rosenthal).
+ Clean up and streamline RTL behavior (Jesse Rosenthal, #3140).
You can set `dir: rtl` in YAML metadata, or use `-M dir=rtl`
on the command line. For finer-grained control, you can set
the `dir` attribute in Div or Span elements.

* Org writer (Albert Krewinkel):

+ Remove blank line after figure caption. Org-mode only treats an image
as a figure if it is directly preceded by a caption.
+ Ensure blank line after figure. An Org-mode figure should be surrounded
by blank lines. The figure would be recognized regardless, but images
in the following line would unintentionally be treated as figures as
+ Ensure link targets are paths or URLs. Org-mode treats links as
document internal searches unless the link target looks like a URL or
file path, either relative or absolute. This change ensures that this
is always the case.
+ Translate language identifiers. Pandoc and Org-mode use different
programming language identifiers. An additional translation between
those identifiers is added to avoid unexpected behavior. This fixes a
problem where language specific source code would sometimes be output
as example code.
+ Drop space before footnote markers (Albert Krewinkel, #3162).
The writer no longer adds an extra space before footnote markers.

* Markdown writer:

+ Don't emit HTML for tables unless `raw_html` extension is set (#3154).
Emit `[TABLE]` if no suitable table formats are enabled and raw HTML
is disabled.
+ Check for the `raw_html` extension before emitting a raw HTML block.
+ Abstract out note/ref function (Jesse Rosenthal).
+ Add ReaderT monad for environment variables (Jesse Rosenthal).

* HTML, EPUB, slidy, revealjs templates: Use `<p>` instead of `<h1>` for
subtitle, author, date (#3119). Note that, as a result of this change,
authors may need to update CSS.

* revealjs template: Added `notes-server` option
(jgm/pandoc-templates#212, Yoan Blanc).

* Beamer template:

+ Restore whitespace between paragraphs. This was
a regression in the last release (jgm/pandoc-templates#207).
+ Added `themeoptions` variable (Carsten Gips).
+ Added `beamerarticle` variable. This causes the `beamerarticle`
package to be loaded in beamer, to produce an article from beamer
slides. (Carsten Gips)
+ Added support for `fontfamilies` structured variable
(Artem Klevtsov).
+ Added hypersetup options (Jake Zimmerman).

* LaTeX template:

+ Added dummy definition for `\institute`.
This isn't a standard command, and we want to avoid a crash when
`institute` is used with the default template.
+ Define default figure placement (Václav Haisman), since pandoc
no longer includes `[htbp]` for figures. Users with custom templates
will want to add this. See #3103.
+ Use footnote package to fix notes in tables (jgm/pandoc-templates#208,
Václav Haisman).

* Moved template compiling/rendering code to a separate library.
`doctemplates`. This allows the pandoc templating system to be
used independently.

* Text.Pandoc.Error: Fix out of index error in `handleError`
(Matthew Pickering). The fix is to not try to show the exact line when
it would cause an out-of-bounds error as a result of included files.

* Text.Pandoc.Shared: Add `linesToBlock` function (Albert Krewinkel).

* Text.Pandoc.Parsing.emailAddress: tighten up parsing of email
addresses. Technically `**@user` is a valid email address, but if we
allow things like this, we get bad results in markdown flavors
that autolink raw email addresses (see #2940). So we exclude a few
valid email addresses in order to avoid these more common bad cases.

* Text.Pandoc.PDF: Don't crash with nonexistent image (#3100). Instead,
emit the alt text, emphasized. This accords with what the ODT writer
currently does. The user will still get a warning about a nonexistent

* Fix example in API documentation (#3176, Thomas Weißschuh).

* Tell where to get tarball in INSTALL (#3062).

* Rename README to MANUAL.txt and add GitHub-friendly
(Albert Krewinkel, Kolen Cheung).

* Replace COPYING with Markdown version from GNU (Kolen Cheung).

* MANUAL.txt:

+ Put note on structured vars in separate paragraph (#2148, Albert
Krewinkel). Make it clearer that structured author variables require a
custom template
+ Note that `--katex` works best with `html5` (#3077).
+ Fix the LaTeX and EPUB links in manual (Morton Fox).
+ Document `biblio-title` variable.

* Improve spacing of footnotes in `--help` output (Waldir Pimenta).

* Update KaTeX to v0.6.0 (Kolen Cheung).

* Allow latest dependencies.

* Use texmath (#3040).

* Allow http-client 0.4.30, which is the version in stackage lts.
Previously we required 0.5.
Remove CPP conditionals for earlier versions.

* Remove support for GHC < 7.8 (Jesse Rosenthal).

+ Remove Compat.Monoid.
+ Remove an inline monad compatibility macro.
+ Remove Text.Pandoc.Compat.Except.
+ Remove directory compat.
+ Change constraint on mtl.
+ Remove unnecessary CPP condition in UTF8.
+ Bump base lower bound to 4.7.
+ Remove 7.6 build from .travis.yaml.
+ Bump supported ghc version in
+ Add note about GHC version support to INSTALL.
+ Remove GHC 7.6 from list of tested versions (Albert Krewinkel).
+ Remove TagSoup compat.
+ Add EOL note to time compat module. Because time 1.4 is a boot library
for GHC 7.8, we will support the compatibility module as long as we
support 7.8. But we should be clear about when we will no longer need
+ Remove blaze-html CPP conditional.
+ Remove unnecessary CPP in custom Prelude.

pandoc (1.17.2)

* Added Zim Wiki writer, template and tests. `zimwiki` is now
a valid output format. (Alex Ivkin)

* Changed email-obfuscation default to no obfuscation (#2988).
+ `writerEmailObfuscation` in `defaultWriterOptions` is now
+ the default for the command-line `--email-obfuscation` option is
now `none`.

* Docbook writer: Declare xlink namespace in Docbook5 output (Ivo Clarysse).

* Org writer:

+ Support arbitrary raw inlines (Albert Krewinkel).
Org mode allows arbitrary raw inlines ("export snippets" in Emacs
parlance) to be included as `@@format:raw foreign format text@@`.
+ Improve Div handling (Albert Krewinkel). Div blocks handling is
changed to make the output look more like idiomatic org mode:
- Div-wrapped content is output as-is if the div's attribute is the
null attribute.
- Div containers with an id but neither classes nor key-value pairs
are unwrapped and the id is added as an anchor.
- Divs with classes associated with greater block elements are
wrapped in a `#+BEGIN`...`#+END` block.
- The old behavior for Divs with more complex attributes is kept.

* HTML writer:

+ Better support for raw LaTeX environments (#2758).
Previously we just passed all raw TeX through when MathJax was used for
HTML math. This passed through too much. With this patch, only raw
LaTeX environments that MathJax can handle get passed through.
This patch also causes raw LaTeX environments to be treated
as math, when possible, with MathML and WebTeX output.

* Markdown writer: use raw HTML for simple, pipe tables with linebreaks
(#2993). Markdown line breaks involve a newline, and simple and pipe
tables can't contain one.

* Make --webtex work with the Markdown writer (#1177).
This is a convenient option for people using
websites whose Markdown flavors don't provide for math.

* Docx writer:

+ Set paragraph to FirstPara after display math (Jesse Rosenthal).
We treat display math like block quotes, and apply FirstParagraph style
to paragraphs that follow them. These can be styled as the user
wishes. (But, when the user is using indentation, this allows for
paragraphs to continue after display math without indentation.)
+ Use actual creation time as doc prop (Jesse Rosenthal).
Previously, we had used the user-supplied date, if available, for Word's
document creation metadata. This could lead to weird results, as in
cases where the user post-dates a document (so the modification might be
prior to the creation). Here we use the actual computer time to set the
document creation.

* LaTeX writer:

+ Don't URI-escape image source (#2825). Usually this is a local file,
and replacing spaces with `%20` ruins things.
+ Allow 'standout' as a beamer frame option (#3007).
`## Slide title {.standout}`.

* RST reader: Fixed links with no explicit link text. The link
`` `<foo>`_ `` should have `foo` as both its link text and its URL.
See RST spec at <>
Closes Debian #828167 -- reported by Christian Heller.

* Textile reader:

+ Fixed attributes (#2984). Attributes can't be followed by
a space. So, `_(class)emph_` but `_(noclass) emph_`.
+ Fixed exponential parsing bug (#3020).
+ Fix overly aggressive interpretation as images (#2998).
Spaces are not allowed in the image URL in textile.

* LaTeX reader:

+ Fix `\cite` so it is a NormalCitation not AuthorInText.
+ Strip off double quotes around image source if present (#2825).
Avoids interpreting these as part of the literal filename.

* Org reader:

+ Add semicolon to list of special chars (Albert Krewinkel)
Semicolons are used as special characters in citations syntax. This
ensures the correct parsing of Pandoc-style citations: `[prefix; @key;
suffix]`. Previously, parsing would have failed unless there was a space
or other special character as the last <prefix> character.
+ Add support for "Berkeley-style" cites (Albert Krewinkel, #1978).
A specification for an official Org-mode citation syntax was drafted by
Richard Lawrence and enhanced with the help of others on the orgmode
mailing list. Basic support for this citation style is added to the
+ Support arbitrary raw inlines (Albert Krewinkel).
Org mode allows arbitrary raw inlines ("export snippets" in Emacs
parlance) to be included as `@@format:raw foreign format text@@`.
+ Remove partial functions (Albert Krewinkel, #2991).
Partial functions like `head` lead to avoidable errors and should be
avoided. They are replaced with total functions.
+ Support figure labels (Albert Krewinkel, #2496, #2999).
Figure labels given as `#+LABEL: thelabel` are used as the ID of the
respective image. This allows e.g. the LaTeX to add proper `\label`
+ Improve tag and properties type safety (Albert Krewinkel).
Specific newtype definitions are used to replace stringly typing of tags
and properties. Type safety is increased while readability is improved.
+ Parse as headlines, convert to blocks (Albert Krewinkel).
Emacs org-mode is based on outline-mode, which treats documents as trees
with headlines are nodes. The reader is refactored to parse into a
similar tree structure. This simplifies transformations acting on
document (sub-)trees.
* Refactor comment tree handling (Albert Krewinkel).
Comment trees were handled after parsing, as pattern matching on lists
is easier than matching on sequences. The new method of reading
documents as trees allows for more elegant subtree removal.
* Support archived trees export options (Albert Krewinkel).
Handling of archived trees can be modified using the `arch` option.
Archived trees are either dropped, exported completely, or collapsed to
include just the header when the `arch` option is nil, non-nil, or
`headline`, respectively.
* Put export setting parser into module (Albert Krewinkel).
Export option parsing is distinct enough from general block parsing to
justify putting it into a separate module.
* Support headline levels export setting (Albert Krewinkel).
The depths of headlines can be modified using the `H` option. Deeper
headlines will be converted to lists.
* Replace ugly code with view pattern (Albert Krewinkel).
Some less-than-smart code required a pragma switching of overlapping
pattern warnings in order to compile seamlessly. Using view patterns
makes the code easier to read and also doesn't require overlapping
pattern checks to be disabled.
* Fix parsing of verbatim inlines (Albert Krewinkel, #3016).
Org rules for allowed characters before or after markup chars were not
checked for verbatim text. This resultet in wrong parsing outcomes of
if the verbatim text contained e.g. space enclosed markup characters as
part of the text (`=is_substr = True=`). Forcing the parser to update
the positions of allowed/forbidden markup border characters fixes this.

* LaTeX template: fix for obscure hyperref/xelatex issue.
Here's a minimal case:


Without this change, this fails on the second invocation of xelatex.
This affects inputs this like `# %á` with pdf output via xelatex.

* trypandoc: call results 'html' instead of 'result'.
This is for better compatibility with babelmark2.

* Document MultiMarkdown as input/output format (Albert Krewinkel, #2973).
MultiMarkdown was only mentioned as a supported Markdown dialect but not
as a possible input or output format. A brief mention is added
everywhere the other supported markdown dialects are mentioned.

* Document Org mode as a format containing raw HTML (Albert Krewinkel)
Raw HTML is kept when the output format is Emacs Org mode.

* Implement `RawInline` and `RawBlock` in sample lua custom writer (#2985).

* Text.Pandoc.Shared:

+ Introduce blocksToInlines function (Jesse Rosenthal).
This is a lossy function for converting `[Block] -> [Inline]`. Its main
use, at the moment, is for docx comments, which can contain arbitrary
blocks (except for footnotes), but which will be converted to spans.
This is, at the moment, pretty useless for everything but the basic
`Para` and `Plain` comments. It can be improved, but the docx reader
should probably emit a warning if the comment contains more than this.
+ Add BlockQuote to blocksToInlines (Jesse Rosenthal).
+ Add further formats for `normalizeDate` (Jesse Rosenthal).
We want to avoid illegal dates -- in particular years with greater than
four digits. We attempt to parse series of digits first as `%Y%m%d`, then
`%Y%m`, and finally `%Y`.
+ `normalizeDate` should reject illegal years (Jesse Rosenthal).
We only allow years between 1601 and 9999, inclusive. The ISO 8601
actually says that years are supposed to start with 1583, but MS Word
only allows 1601-9999. This should stop corrupted word files if the date
is out of that range, or is parsed incorrectly.
+ Improve year sanity check in normalizeDate (Jesse Rosenthal).
Previously we parsed a list of dates, took the first one, and then
tested its year range. That meant that if the first one failed, we
returned nothing, regardless of what the others did. Now we test for
sanity before running `msum` over the list of Maybe values. Anything
failing the test will be Nothing, so will not be a candidate.

* Docx reader:

+ Add simple comment functionality. (Jesse Rosenthal).
This adds simple track-changes comment parsing to the docx reader. It is
turned on with `--track-changes=all`. All comments are converted to
inlines, which can list some information. In the future a warning will be
added for comments with formatting that seems like it will be excessively
denatured. Note that comments can extend across blocks. For that reason
there are two spans: `comment-start` and `comment-end`. `comment-start`
will contain the comment. `comment-end` will always be empty. The two
will be associated by a numeric id.
+ Enable warnings in top-level reader (Jesse Rosenthal).
Previously we had only allowed for warnings in the parser. Now we allow
for them in the `Docx.hs` as well. The warnings are simply concatenated.
+ Add warning for advanced comment formatting. (Jesse Rosenthal).
We can't guarantee we'll convert every comment correctly, though we'll
do the best we can. This warns if the comment includes something other
than Para or Plain.
+ Add tests for warnings. (Jesse Rosenthal).
+ Add tests for comments (Jesse Rosenthal).
We test for comments, using all track-changes options. Note that we
should only output comments if `--track-changes=all`. We also test for
emitting warnings if there is complicated formatting.

* README: update to include track-changes comments. (Jesse Rosenthal)

* Improved Windows installer - don't ignore properties set on command-line.
See #2708. Needs testing to see if this resolves the issue.
Thanks to @nkalvi.

* Process markdown extensions on command line in L->R order (#2995).
Previously they were processed, very unintuitively, in R->L
order, so that `markdown-tex_math_dollars+tex_math_dollars`
had `tex_math_dollars` disabled.

* Added `secnumdepth` variable to LaTeX template (#2920).

* Include table of contents in README.html in Windows package.

* Writers: treat SoftBreak as space for stripping (Jesse Rosenthal)
In Writers.Shared, we strip leading and trailing spaces for display
math. Since SoftBreak's are treated as spaces, we should strip those

* beamer, latex templates: pass biblatexoptions directly in package load.
This allows runtime optinos to be used. Fixes jgm/pandoc-citeproc#201

* CPP workaround for deprecation of `parseUrl` in http-client.

* Removed some redundant class constraints.

* - use OSX env variable.

* Replaced INSTALL with, incorporating INSTALL and the
old installing page from website.

* Added `winpkg` target to Makefile. This downloads the windows package
from appveyor and signs it using the key.

* Document Org mode as a format containing raw TeX (Albert Krewinkel).
Raw TeX is kept verbatim when the output format is Emacs Org mode.

* Support math with haddock-library >= 1.4.

* Removed `-rtsopts` from library stanza. It has no effect, and Hackage
wouldn't accept the package.

* Update library dependency versions.

pandoc (1.17.1)

* New output format: `docbook5` (Ivo Clarysse).

* `Text.Pandoc.Options`: Add `writerDocBook5` to `WriterOptions`
(API change).

* Org writer:

+ Add :PROPERTIES: drawer support (Albert Krewinkel, #1962).
This allows header attributes to be added to org documents in the form
of `:PROPERTIES:` drawers. All available attributes are stored as
key/value pairs. This reflects the way the org reader handles
`:PROPERTIES:` blocks.
+ Add drawer capability (Carlos Sosa). For the implementation of the
Drawer element in the Org Writer, we make use of a generic Block
container with attributes. The presence of a `drawer` class defines
that the `Div` constructor is a drawer. The first class defines the
drawer name to use. The key-value list in the attributes defines
the keys to add inside the Drawer. Lastly, the list of Block elements
contains miscellaneous blocks elements to add inside of the Drawer.
+ Use `CUSTOM_ID` in properties (Albert Krewinkel). The `ID` property is
reserved for internal use by Org-mode and should not be used.
The `CUSTOM_ID` property is to be used instead, it is converted to the
`ID` property for certain export format.

* LaTeX writer:

+ Ignore `--incremental` unless output format is beamer (#2843).
+ Fix polyglossia to babel env mapping (Mauro Bieg, #2728).
Allow for optional argument in square brackets.
+ Recognize `la-x-classic` as Classical Latin (Andrew Dunning).
This allows one to access the hyphenation patterns in CTAN's
+ Add missing languages from hyph-utf8 (Andrew Dunning).
+ Improve use of `\strut` with `\minipage` inside tables
(Jose Luis Duran). This improves spacing in multiline
+ Use `{}` around options containing special chars (#2892).
+ Avoid lazy `foldl`.
+ Don't escape underscore in labels (#2921). Previously they were
escaped as `ux5f`.
+ brazilian -> brazil for polyglossia (#2953).

* HTML writer: Ensure mathjax link is added when math appears in footnote
(#2881). Previously if a document only had math in a footnote, the
MathJax link would not be added.

* EPUB writer: set `navpage` variable on nav page.
This allows templates to treat it differently.

* DocBook writer:

+ Use docbook5 if `writerDocbook5` is set (Ivo Clarysse).
+ Properly handle `ulink`/`link` (Ivo Clarysse).

* EPUB reader:

+ Unescape URIs in spine (#2924).
+ EPUB reader: normalise link id (Mauro Bieg).

* Docx Reader:

+ Parse `moveTo` and `moveFrom` (Jesse Rosenthal).
`moveTo` and `moveFrom` are track-changes tags that are used when a
block of text is moved in the document. We now recognize these tags and
treat them the same as `insert` and `delete`, respectively. So,
`--track-changes=accept` will show the moved version, while
`--track-changes=reject` will show the original version.
+ Tests for track-changes moving (Jesse Rosenthal).

* ODT, EPUB, Docx readers: throw `PandocError` on unzip failure
(Jesse Rosenthal) Previously, `readDocx`, `readEPUB`, and `readOdt`
would error out if zip-archive failed. We change the archive extraction
step from `toArchive` to `toArchiveOrFail`, which returns an Either value.

* Markdown, HTML readers: be more forgiving about unescaped `&` in
HTML (#2410). We are now more forgiving about parsing invalid HTML with
unescaped `&` as raw HTML. (Previously any unescaped `&`
would cause pandoc not to recognize the string as raw HTML.)

* Markdown reader:

+ Added bracket syntax for native spans (#168).
+ Fix pandoc title blocks with lines ending in 2 spaces (#2799).
+ Added `-s` to markdown-reader-more test.

* HTML reader: fixed bug in `pClose`. This caused exponential parsing
behavior in documnets with unclosed tags in `dl`, `dd`, `dt`.

* MediaWiki reader: Allow spaces before `!` in MediaWiki table header

* RST reader: Support `:class:` option for code block in RST reader
(Sidharth Kapur).

* Org reader (all Albert Krewinkel, except where noted otherwise):

+ Stop padding short table rows.
Emacs Org-mode doesn't add any padding to table rows. The first
row (header or first body row) is used to determine the column count,
no other magic is performed.
+ Refactor rows-to-table conversion. This refactors
the codes conversing a list table lines to an org table ADT.
The old code was simplified and is now slightly less ugly.
+ Fix handling of empty table cells, rows (Albert Krewinkel, #2616).
This fixes Org mode parsing of some corner cases regarding empty cells
and rows. Empty cells weren't parsed correctly, e.g. `|||` should be
two empty cells, but would be parsed as a single cell containing a pipe
character. Empty rows where parsed as alignment rows and dropped from
the output.
+ Fix spacing after LaTeX-style symbols.
The org-reader was dropping space after unescaped LaTeX-style symbol
commands: `\ForAll \Auml` resulted in `∀Ä` but should give `∀ Ä`
instead. This seems to be because the LaTeX-reader treats the
command-terminating space as part of the command. Dropping the trailing
space from the symbol-command fixes this issue.
+ Print empty table rows. Empty table rows should not
be dropped from the output, so row-height is always set to be at least 1.
+ Move parser state into separate module.
The org reader code has become large and confusing. Extracting smaller
parts into submodules should help to clean things up.
+ Add support for sub/superscript export options.
Org-mode allows to specify export settings via `#+OPTIONS` lines.
Disabling simple sub- and superscripts is one of these export options,
this options is now supported.
+ Support special strings export option Parsing of special strings
(like `...` as ellipsis or `--` as en dash) can be toggled using the `-`
+ Support emphasized text export option. Parsing of emphasized text can
be toggled using the `*` option. This influences parsing of text marked
as emphasized, strong, strikeout, and underline. Parsing of inline math,
code, and verbatim text is not affected by this option.
+ Support smart quotes export option. Reading of smart quotes can be
toggled using the `'` option.
+ Parse but ignore export options. All known export options are parsed
but ignored.
+ Refactor block attribute handling. A parser state attribute was used
to keep track of block attributes defined in meta-lines. Global state
is undesirable, so block attributes are no longer saved as part of the
parser state. Old functions and the respective part of the parser state
are removed.
+ Use custom `anyLine`. Additional state changes need to be made after
a newline is parsed, otherwise markup may not be recognized correctly.
This fixes a bug where markup after certain block-types would not be
+ Add support for `ATTR_HTML` attributes (#1906).
Arbitrary key-value pairs can be added to some block types using a
`#+ATTR_HTML` line before the block. Emacs Org-mode only includes these
when exporting to HTML, but since we cannot make this distinction here,
the attributes are always added. The functionality is now supported
for figures.
+ Add `:PROPERTIES:` drawer support (#1877).
Headers can have optional `:PROPERTIES:` drawers associated with them.
These drawers contain key/value pairs like the header's `id`. The
reader adds all listed pairs to the header's attributes; `id` and
`class` attributes are handled specially to match the way `Attr` are
defined. This also changes behavior of how drawers of unknown type
are handled. Instead of including all unknown drawers, those are not
read/exported, thereby matching current Emacs behavior.
+ Use `CUSTOM_ID` in properties. See above on Org writer changes.
+ Respect drawer export setting. The `d` export option can be used
to control which drawers are exported and which are discarded.
Basic support for this option is added here.
+ Ignore leading space in org code blocks (Emanuel Evans, #2862).
Also fix up tab handling for leading whitespace in code blocks.
+ Support new syntax for export blocks. Org-mode version 9
uses a new syntax for export blocks. Instead of `#+BEGIN_<FORMAT>`,
where `<FORMAT>` is the format of the block's content, the new
format uses `#+BEGIN_export <FORMAT>` instead. Both types are
+ Refactor `BEGIN...END` block parsing.
+ Fix handling of whitespace in blocks, allowing content to be indented
less then the block header.
+ Support org-ref style citations. The *org-ref* package is an
org-mode extension commonly used to manage citations in org
documents. Basic support for the `cite:citeKey` and
`[[cite:citeKey][prefix text::suffix text]]` syntax is added.
+ Split code into separate modules, making for cleaner code and
better decoupling.

* Added `docbook5` template.

* `--mathjax` improvements:

+ Use new CommonHTML output for MathJax (updated default MathJax URL,
+ Change default mathjax setup to use `TeX-AMS_CHTML` configuration.
This is designed for cases where the input is always TeX and maximal
conformity with TeX is desired. It seems to be smaller and load faster
than what we used before. See #2858.
+ Load the full MathJax config to maximize loading speed (KolenCheung).

* Bumped upper version bounds to allow use of latest packages
and compilation with ghc 8.

* Require texmath Closes several texmath-related bugs (#2775,
#2310, #2310, #2824). This fixes behavior of roots, e.g.
`\sqrt[3]{x}`, and issues with sub/superscript positioning
and matrix column alignment in docx.


+ Clarified documentation of `implicit_header_references` (#2904).
+ Improved documentation of `--columns` option.

* Added appveyor setup, with artifacts (Jan Schulz).

* stack.yaml versions: Use proper flags used for texmath, pandoc-citeproc.

* LaTeX template: support for custom font families (vladipus).
Needed for correct polyglossia operation with Cyrillic fonts and perhaps
can find some other usages. Example usage in YAML metadata:

- name: \cyrillicfont
font: Liberation Serif
- name: \cyrillicfonttt
options: Scale=MatchLowercase
font: Liberation

* Create unsigned msi as build artifact in appveyor build.

* On travis, test with ghc 8.0.1; drop testing for ghc 7.4.1.

pandoc (

* LaTeX writer: Fixed position of label in figures (#2813).
Previously the label wasn't in the right place, and `\ref`
wouldn't work properly.
* Added .tei test files to pandoc.cabal so they'll be included
in tarball (#2811).
* Updated copyright dates.

pandoc (

* Fixed serious regression in `htmlInBalanced`, which caused
newlines to be omitted in some raw HTML blocks in Markdown

pandoc (

* File scope is no longer used when there are no input files (i.e.,
when input comes from stdin). Previously file scope was triggered
when the `json` reader was specified and input came from `stdin`,
and this caused no output to be produced. (Fix due to Jesse Rosenthal;
thanks to Fedor Sheremetyev for calling the bug to our attention.)
* Improved documentation of templates (#2797).

pandoc (1.17)

* Added `--file-scope` option (Jesse Rosenthal).
By default pandoc operates on multiple files by first concatenating
them (around extra line breaks) and then processing the joined file. So
it only parses a multi-file document at the document scope. This has the
benefit that footnotes and links can be in different files, but for
some purposes it is useful to parse the individual files first
and then combine their outputs (e.g. when the files use footnotes
or links with the same labels). The `--file-scope` option causes
pandoc to parse the files first, and then combine the parsed output,
instead of combining before parsing. `--file-scope` is selected
automatically for binary input files (which cannot be concatenated)
and for pandoc json.

* Add TEI Writer (Chris Forster) and `tei` output format.

* Added a general `ByteStringReader` with warnings, used by the docx
reader (API change, Jesse Rosenthal).

* Add `readDocxWithWarnings` (API change, Jesse Rosenthal).

* Changed type of `Shared.uniqueIdent`'s argument from `[String]`
to `Set String.` This avoids performance problems in documents with
many identically named headers (API change, #2671).

* Removed `tex_math_single_backslash` from `markdown_github` options

* Make language extensions as well as full language names
trigger syntax highlighting. For example, `py` will now work as
well as `python` (jgm/highlighting-kate#83).

* Added `institute` variable to latex, beamer templates (Fraser
Tweedale, Josef Svenningsson).

* Docx reader (Jesse Rosenthal):

+ Handle alternate content. Some word functions (especially graphics)
give various choices for content so there can be backwards compatibility.
+ Don't turn numbered headers into lists.
+ Docx Reader: Add state to the parser, for warnings
+ Update feature checklist in source code.
+ Get rid of `Modifiable` typeclass.
+ Add tests for adjacent hyperlinks.
+ Add a "Link" modifier to `Reducible`. We want to make sure that
links have their spaces removed, and are appropriately smushed
together (#2689).

* HTML reader:

+ Fixed behavior of base tag (#2777).
If the base path does not end with slash, the last component
will be replaced. E.g. base = ``
combines with `bar.html` to give ``.
If the href begins with a slash, the whole path of the base
is replaced. E.g. base = `` combines
with `/bar.html` to give ``.
+ Rewrote `htmlInBalanced`. This version avoids an exponential
performance problem with `<script>` tags, and it should be faster
in general (#2730).
+ Properly handle an empty cell in a simple table (#2718).
+ Handle multiple `<meta>` tags with same name. Put them in a list
in the metadata so they are all preserved, rather than (as before)
throwing out all but one..

* Markdown reader:

+ Improved pipe table parsing (#2765).
+ Allow `+` separators in pipe table cells. We already allowed
them in the header, but not in the body rows, for some reason.
This gives compatibility with org-mode tables.
+ Don't cross line boundary parsing pipe table row.
Previously an Emph element could be parsed across the newline
at the end of the pipe table row.
+ Use `htmlInBalanced` for `rawVerbatimBlock`, for better
performance (#2730).
+ Fixed bug with smart quotes around tex math.

* LaTeX reader:

+ Handle interior `$` characters in math (#2743). For example,
+ `inlineCommand` now gobbles an empty `{}` after any command (#2687).
This gives better results when people write e.g. `\TeX{}` in Markdown.
+ Properly handle LaTeX "math" environment as inline math (#2171).

* Textile reader: Support `>`, `<`, `=`, `<>` text alignment attributes.
Closes #2674.

* Org reader (Albert Krewinkel):

+ Prefix even empty figure names with "fig:" (#2643). The
convention used by pandoc for figures is to mark them by prefixing
the name with `fig:`. The org reader failed to do this if a figure
had no name.
+ Refactor link-target processing (#2684).

* ConTeXt writer: Fix whitespace at line beginning in line blocks (#2744).
Thanks to @c-foster.

* HTML writer: Don't include alignment attribute for default table columns.
Previously these were given "left" alignment. Better to leave off
alignment attributes altogether (#2694).

* Markdown writer: Use hyphens for YAML metadata block bottom line, for
better compatibility with other Markdown flavors (Henrik Tramberend).

* LaTeX writer:

+ Use image identifier to create a label and hypertarget for
figures (Mauro Bieg).
+ Avoid double toprule in headerless table with caption (#2742).
+ Clean up options parser (Jesse Rosenthal).
+ Treat `memoir` template with `article` option as article, instead
of treating all `memoir` templates as books.
+ Allow more flexible table alignment (Henrik Tramberend, #2665).
New default is not to include `[c]` option (which is the default
anyway if no positioning is specified). Now LaTeX emplates can
control the overall table alignment in a document by setting the
longtable length variables `LTleft` and `LTright`. For example,
will create left-aligned tables that respect paragraph indentation.

* Docx writer: Handle image alt text (#2754, Mauro Bieg).

* Org writer - pass through RawInline with format "org".

* DokuWiki writer: use `$$` for display math.

* Custom writer: Pass attributes parameter to CaptionedImage (#2697).

* Make protocol-relative URIs work again (#2737).

* Use env variable for developer id certs.

* Raise `tagsoup` lower bound to 0.13.7 to fix entity-related
problems (#2734).

* Allow `zip-archive` 0.3.

* Allow `aeson` 0.11.

pandoc (

* Depend on deepseq rather than deepseq-generics (fpco/stackage#1096).

* Fixed regression in latex smart quote parsing (#2645).
In cases where a match was not found for a quote, everything
from the open quote to the end of the paragraph was being dropped.

pandoc (

* Fixed regression with `--latex-engine` (#2618). In 1.16 `--latex-engine`
raises an error if a full path is given.

* Org reader: Fix function dropping subtrees tagged `:noexport`
(Albert Krewinkel, #2628):

* Markdown reader: renormalize table column widths if they exceed 100%

* Textile reader: don't allow block HTML tags in inline contexts.
The reader previously did allow this, following redcloth,
which happily parses

Html blocks can be <div>inlined</div> as well.


<p>Html blocks can be <div>inlined</div> as well.</p>

This is invalid HTML. The above sample now produces;

<p>Html blocks can be</p>
<p>as well.</p>

* Improved default template lookup for custom lua scripts (#2625).
Previously, if you tried to do `pandoc -s -t /path/to/lua/script.lua`,
pandoc would look for the template in
With this change it will look in the more reasonable
`~/.pandoc/templates/default.script.lua`. This makes it possible to
store default templates for custom writers.

* RST, Markdown writers: Fixed rendering of grid tables with blank rows

* LaTeX writer: restore old treatment of Span (#2624). A Span is
now rendered with surrounding `{}`, as it was before 1.16.

* Entity handling fixes: improved handling of entities like
`&lang;` that require a trailing semicolon. Allow uppercase
`x` in numerical hexadecimal character references, working
around a tagsoup bug.

* `stack.yaml` - use lts-4.0, but with older aeson to avoid excessive
memory use on compile. With aeson 0.10 we were getting an out of
memory error on a 2GB Ubuntu 64-bit VM.

* Improved deb package creation script. Made `DPKGVER` work.
Renamed `COMMIT` to `TREE`. You should now be able to do
`TREE= DPKGVER=2 make deb`.

pandoc (1.16)

* Added `Attr` field to `Link` and `Image` (Mauro Bieg, #261, API change).

+ Added syntax for link and image attributes to pandoc's Markdown.
+ Updated readers and writers to use link and image attributes
when appropriate.
+ Support image attributes in Docx, Textile, RST readers.

* Renamed link attribute extensions. The old `link_attributes` is
now `mmd_link_attributes`, and `link_attributes` now enables the
new pandoc-style link and image attributes (API change).
Note: this change could break some existing workflows.

* Implemented `SoftBreak` and new `--wrap` option (#1701, API change).
Added threefold wrapping option.

+ Command line option: deprecated `--no-wrap`, added
+ Added `WrapOption`, exported from `Text.Pandoc.Options`
+ Changed type of `writerWrapText` in `WriterOptions` from
`Bool` to `WrapOption`.
+ Modified `Text.Pandoc.Shared` functions to allow `SoftBreak`.
+ Supported `SoftBreak` in readers and writers.

* Text.Pandoc.Options: Added `writerDpi` to `WriterOptions` (API
change, Mauro Bieg).

* Added `--dpi` command-line option (Mauro Bieg).

* Rationalized behavior of `--no-tex-ligatures` and `--smart` (#2541).
This change makes `--no-tex-ligatures` affect the LaTeX reader
as well as the LaTeX and ConTeXt writers. If it is used,
the LaTeX reader will parse characters `` ` ``, `'`, and `-`
literally, rather than parsing ligatures for quotation marks
and dashes. And the LaTeX writer will print unicode quotation
mark and dash characters literally, rather than converting
them to the standard ASCII ligatures. Note that `--smart` has
no effect on the LaTeX reader. `--smart` is still the default
for all input formats when LaTeX or ConTeXt is the output format,
*unless* `--no-tex-ligatures` is used.

Some examples to illustrate the logic:

% echo "'hi'" | pandoc -t latex
% echo "'hi'" | pandoc -t latex --no-tex-ligatures
% echo "'hi'" | pandoc -t latex --no-tex-ligatures --smart
% echo "'hi'" | pandoc -f latex --no-tex-ligatures
% echo "'hi'" | pandoc -f latex

* Removed deprecated options `--offline` and `--html5`.

* Fixed language code for Czech (`cs` not `cz`) (#2597).

* Implemented `east_asian_line_breaks` extension (#2586).
In `Text.Pandoc.Options`, added `Ext_east_asian_line_breaks` constructor
to `Extension` (API change). This extension is like
`ignore_line_breaks`, but smarter -- it only ignores line breaks
between two East Asian wide characters. This makes it better suited
for writing with a mix of East Asian and non-East Asian scripts.

* Added support for PDF creation via `wkhtmltopdf`.
To use this: `pandoc -t html5 -o result.pdf` (and add `--mathjax`
if you have math.) Margins can be set using the variables
`margin-top`, `margin-bottom`, `margin-left`, `margin-right`.
Other styling can be done through CSS.

* Fixed cite key parsing regression (jgm/pandoc-citeproc#201).
We were capturing final colons as in `[@foo: bar]`; the citation id
was being parsed as `@foo:`.

* ICML writer:

+ Fixed image syntax for local files (#2589).
+ Changed type of `writeICML` (Mauro Bieg).
API change: It is now `WriterOptions -> Pandoc -> IO String`.
Also handle new image attributes.
+ Intersperse line breaks instead of appending them to
every `ParagraphStyleRange` (Mauro Bieg, #2501).
+ Add `Cite` style to citations (Mauro Bieg).
+ Added figure handling (#2590, Mauro Bieg).
+ Better handling of math. Instead of just printing the raw tex,
we now try to fake it with unicode characters.

* HTML writer: Include `example` class for example lists (#2524).

* ODT/OpenDocument writer: improved image attributes (Mauro Bieg).

+ Support for percentage widths/heights
+ Use `Attr` instead of title to get dimensions from ODT walker
to `writeOpenDocument`.

* AsciiDoc writer:

+ Support anchors in spans and divs with id elements
+ Fixed code blocks (#1861).

* Haddock writer: omit formatting inside links, which isn't supported
by Haddock (#2515).

* MediaWiki writer: Fixed spacing issues in table cells.

+ Start cell on new line unless it's a single Para or Plain
+ For single Para or Plain, insert a space after the `|` to
avoid problems when the text begins with a character like
`-` (#2604).

* Beamer writer: mark frame as fragile when it contains verbatim (#1613).

* LaTeX writer:

+ Add support for GAP highlighting using listings (Raniere Silva).
+ Consider `header-includes` content as well as templates
when determining whether to use csquotes (Andreas Lööw).
+ Create defaults for geometry using `margin-left` etc.
If `geometry` has no value, but `margin-left`, `margin-right`,
`margin-top`, and/or `-margin-bottom` are given, a default value
for `geometry` is created from these. Note that these variables
already affect PDF production via HTML5 with `wkhtmltopdf`.

* ConTeXt writer: set default layout based on `margin-left`, etc.
This sets up `\setuplayout` based on the variables `margin-left`,
`margin-right`, `margin-bottom`, and `margin-top`, if no layout
is given.

* Docx writer: better handling of PDF images. Previously we tried
to get the image size from the image even if an explicit size was
specified. Since we still can't get image size for PDFs, this made
it impossible to use PDF images in docx. Now we don't try to get
the image size when a size is already explicitly specified.

* Markdown writer: use raw HTML for link/image attributes when
the `link_attributes` extension is unset and `raw_html` is set (#2554).

* MediaWiki reader: interpret markup inside `<tt>`, `<code>` (#2607).

* LaTeX reader:

+ Improved smart quote parsing (#2555). This fixes redering of
unmatched quotes.
+ Use curly quotes for unmatched ` (#2555).
+ Allow blank space between braced arguments of commands (#2592).

* Markdown reader:

+ Improved pipe table relative widths. Previously pipe table
columns got relative widths (based on the header underscore lines)
when the source of one of the rows was greater in width than the
column width. This gave bad results in some cases where much of
the width of the row was due to nonprinting material (e.g. link
URLs). Now pandoc only looks at printable width (the width of a
plain string version of the source), which should give better results.
Thanks to John Muccigrosso for bringing up the issue.
+ Fixed parsing bug with macros. Previously macro definitions in
indented code blocks were being parsed as macro definitions, not code.

* Textile reader: skip over attribute in image source (#2515).
We don't have a place yet for styles or sizes on images, but
we can skip the attributes rather than incorrectly taking them
to be part of the filename.

* Docx reader: Handle dummy list items (Jesse Rosenthal).
These come up when people create a list item and then delete the
bullet. It doesn't refer to any real list item, and we used to ignore

* CommonMark reader/writer rewritten to use latest `cmark`.

* Fixed Emoji character definitions (#2523). There were many bugs in the

* `Text.Pandoc.CSS`:

+ Added `pickStylesToKVs` function to extract multiple properties at
once (API change, Mauro Bieg).
+ Parse CSS that doesn't contain the optional semicolon (Mauro Bieg).

* `trypandoc`: sort drop-down lists.

* Beamer template:

+ Made `\euro` conditional on presence of character.
for xelatex and lualatex, as it is for pdflatex (Andrew Dunning).
+ Moved `header-includes` before setting of title (Thomas Hodgson),
to match the LaTeX template (jgm/pandoc-templates#168).
+ Added `section-titles` variable (defaults to true)
to enable/suppress section title pages in beamer
slide shows (Thomas Hodgson).
+ Moved beamer themes after fonts, so that themes can
change fonts. (Previously the fonts set were being
clobbered by lmodern.sty.) (Thomas Hodgson).

* Beamer/LaTeX template changes (Thomas Hodgson):
+ Added `thanks` variable
+ Use `parskip.sty` when `indent` isn't set (fall back to using
`setlength` as before if `parskip.sty` isn't available).
+ Use `biblio-style` with biblatex.
+ Added `biblatexoptions` variable.

* LaTeX template changes:

+ Added `paper` after `$papersize$` variable in latex template.
Thus you can say `papersize: a4` and the latex will contain
`a4paper`. This change may break some existing workflows; if
you currently specify `a4paper`, you'll get `a4paperpaper` which
is meaningless. However, the change seems worth it, as it will
make the `papersize` variable work uniformly across ConTeXt, LaTeX,
and html->pdf via wkhtmltopdf.
+ Only pass options to color package if `colorlinks` is set
(Andrew Dunning).
+ Make definition of `\euro` conditional in xelatex/lualatex,
as it is already for pdflatex (Andrew Dunning).
+ Removed setting of `subject` in PDF metadata.
This used to be set to the subtitle, but really the subtitle
need not give the subject. Also, `subtitle` can contain formatting,
so we'd need, at least, a plain text version for this.
+ Moved `header-includes` before setting of `\title`, `\author`,
etc. This allows these macros to be redefined.
+ Use `\subtitle` command for `subtitle`, instead of tacking it
on to the title as before. We give a no-op fallback definition if it
is not defined. This change should produce much better results
in classes that support `\subtitle`. With the default article
class, which does not define `\subtitle`, subtitles will no
longer be printed unless the user defines `\subtitle` and
redefines `\maketitle`.
+ Moved redefinitions of `\paragraph` and `\subparagraph` to
before header-includes.

* Context template:

+ Use `simplefonts` for font loading (Paolo Rodríguez). This is
needed for things to work on ConTeXt stable from TeXLive 2015.
+ Revert use of `\setuphead` in title block (Andrew Dunning,
Rik Kabel).

* Update LaTeX/ConTeXt link colour usage (Andrew Dunning).

* Fixed man template so disabling hyphenation actually works.
The command needs to come after .TH.

* Added 'navigation' variable to beamer template (#2543).
Valid values are `empty` (the default), `horizontal`, `vertical`,
and `frame`. Note that this changes the default behavior from
`horizontal` to `empty`. Closes #2543.

* Added `toc` to HTML slide format templates (Andrew Dunning),
so that `--toc` creates a contents slide.

* Added `stack.full.yaml` to build `pandoc-citeproc` as well.

* Allow pipe tables with no body rows (#2556).
Previously this raised a runtime error.

* Shared: Improved `fetchItem` so that `C:/Blah/Blah.jpg` isn't treated
as URL. The Haskell URI parsing routines will accept "C:" as a
scheme, so we rule that out manually. This helps with
`--self-contained` and absolute Windows paths.

* Define a `meta-json` variable for all writers (#2019). This contains
a JSON version of all the metadata, in the format selected for the
writer. So, for example, to get just the YAML metadata, you can run
pandoc with the following custom template: `$meta-json$`. The intent
is to make it easier for static site generators and other tools to get
at the metadata.

* Document limitations of --self-contained (#2553).

* Improved Citations section of README (#2551). Added information
about `link-citations` and a link to the pandoc-citeproc man page.

* `ImageSize`: use `safeRead` instead of `readMaybe`, which isn't
in base < 4.6.

* Allow .adoc file extension for AsciiDoc (Andrew Dunning).

* Improved implicit pandoc-citeproc inclusion.
The filter pandoc-citeproc is automatically used when
`--bibliography` is specified on the command line, unless
`--natbib` or `--biblatex` is used. However, previously this
only worked if `--bibliography` was spelled out in full, and not
if `--biblio` was used.

* reveal.js: Interpret pauses correctly for all headers (#2530).
Previously, when using headers below the slide level, pauses are left
uninterpreted into pauses. In my opinion, unexpected behavior but
intentional looking at the code.

* Remove redundant `center` variable for reveal.js (Andrew Dunning).

* Parsing: Add `extractIdClass`, modified type of `KeyTable` (Mauro
Bieg, API change).

* ImageSize: Added functions for converting between image dimensions
(Mauro Bieg).

* Use lts-3.18 in stack.yaml. This avoids Windows build
issues with the HTTP library.

* Bump version bounds for dependencies.

pandoc (

* Added two missing test files, and `stack.yaml`, to
`extra-source-files` so they're included in the source tarball.

* reveal.js template: Fixed parallaxBackground options.
`parallaxBackgroundHorizontal` and `parallaxBackgroundVertical`
need integer values, not strings. (Vaughn Iverson)

pandoc (1.15.2)

* `pandoc -t context -o my.pdf` will now create a PDF using
ConTeXt rather than LaTeX (#2463).

* Fixed omitted `url(...)` in CSS data-uri with `--self-contained` (#2489).

* Added `emoji` Markdown extension, enabled by default in `markdown_github`
(#2523). Added `Ext_emoji` to `Extension` in `Text.Pandoc.Options`
(API change).

* `Text.Pandoc.Readers.HTML.parseTags`: Fixed over-eager raw HTML inline
parsing (#2469). Tightened up the inline HTML parser so it disallows

* Derive `Generic` instances for the types in `Text.Pandoc.Options`.

* Org reader:

+ Fix paragraph/list interaction (Albert Krewinkel, #2464).
Paragraphs can be followed by lists, even if there is no blank line
between the two blocks. However, this should only be true if the
paragraph is not within a list, were the preceding block should be
parsed as a plain instead of paragraph (to allow for compact lists).
Thanks to @rgaiacs for bringing this up.
+ Allow toggling header args (Albert Krewinkel, #2269).
Org-mode allows to skip the argument of a code block header argument if
it's toggling a value. Argument-less headers are now recognized,
avoiding weird parsing errors.
+ Fix markup parsing in headers (Albert Krewinkel, #2504).
Markup as the very first item in a header wasn't recognized. This was
caused by an incorrect parser state: positions at which inline markup
can start need to be marked explicitly by changing the parser state.
This wasn't done for headers. The proper function to update the state
is now called at the beginning of the header parser, fixing this issue.
+ Fix emphasis rules for smart parsing (Albert Krewinkel, #2513).
Smart quotes, ellipses, and dashes should behave like normal quotes,
single dashes, and dots with respect to text markup parsing.
+ Require whitespace around definition list markers (#2518).
This rule was not checked before, resulting in bugs with footnotes
and some link types.

* Markdown reader:

+ Pipe tables with long lines now get relative cell widths (#2471).
If a pipe table contains a line longer than the column width (as set by
`--columns` or 80 by default), relative widths are computed based on the
widths of the separator lines relative to the column width. This should
solve persistent problems with long pipe tables in LaTeX/PDF output, and
give more flexibility for determining relative column widths in other
formats, too. For narrower pipe tables, column widths of 0 are used,
telling pandoc not to specify widths explicitly in output formats that
permit this.
+ Improved parser for `mmd_title_block`. We now allow blank metadata
fields. These were explicitly disallowed before.
+ Citation keys can now contain `://`, so URLs and DOIs can be used
as citation keys (jgm/pandoc-citeproc#166).

* Beamer template: fix incompatibility of section slides with natbib.
Natbib (and presumably biblatex) bibliography commands create
their own section. Since these are in frame environments,
we have an incompatibility with the `\AtBeginSection` macro
which creates a special frame when a new section occurs.
(We can't have a frame inside another frame.) This change disables
`\AtBeginSection` inside bibliography slides. Thinks to Yihui Xie for
bringing the problem to my attention. This supersedes #145. See
discussion there.

* Textile reader: don't do smart punctuation unless explicitly asked
(#2480). Note that although smart punctuation is part of the textile
spec, it's not always wanted when converting from textile
to, say, Markdown. So it seems better to make this an option.

* LaTeX reader: Handle `comment` environment (Arata Mizuki).
The `comment` environment is handled in a similar way to the
`verbatim` environment, except that its content is discarded.

* Docx reader: Follow relationships correctly in foot/endnotes (#2258,
Jesse Rosenthal). This fixes a problem with links in notes.

* LaTeX and ConTeXt writers: support `lang` attribute on divs and spans
(Mauro Bieg). For LaTeX, also collect `lang` and `dir` attributes on
spans and divs to set the `lang`, `otherlangs` and `dir` variables if
they aren’t set already. See #895.

* LaTeX writer:

+ Use proper command for `\textarabic` (Mauro Bieg).
+ Added `de-CH-1901`, fixed `el-polyton` in `toPloyglossia` (Nick Bart).
+ Use `\hypertarget` and `\hyperlink` for links. This works correctly
to link to Div or Span elements. We now don't bother defining `\label`
for Div or Span elements. Closes jgm/pandoc-citeproc#174.
+ Avoid footnotes in list of figures (#1506).
+ Properly handle footnotes in captions (#1506).
+ Add `\protect` to `\hyperlink` (#2490). Thanks to Hadrien Mary.
+ Set `colorlinks` if `linkcolor`, `urlcolor`, `citecolor`, or
`toccolor` is set (#2508).

* Textile writer: support start number in ordered lists (#2465).

* OpenDocument writer: Allow customization of opendocument
automatic styles. Automatic styles can now be inserted in the
template, which now provides the enclosing `<office:automatic-styles>`
tags (#2520).

* Docx writer: insert space between footnote reference and note (#2527).
This matches Word's default behavior.

* EPUB writer: don't download linked media when `data-external` attribute
set (#2473). By default pandoc downloads all linked media and includes it
in the EPUB container. This can be disabled by setting `data-external` on
the tags linking to media that should not be downloaded. Example:

<audio controls="1">
<source src=""
data-external="1" type="audio/mpeg">

* HTML writer: use width on whole table if col widths sum to < 100%.
Otherwise some browsers display the table with the columns
separated far apart.

* AsciiDoc template: Fix `author` and `date`; add `keywords`,
`abstract` (Andrew Dunning).

* HTML-based templates (Andrew Dunning):

+ Use en dash instead of hyphen between title prefix and title.
+ Add `keywords` to metadata.
+ Add `lang`, `dir`, `quotes` where missing.
+ Always make author and date display conditional.
+ Updated dzslides template from source.

* Man template: make "generated by" comment conditional.

* LaTeX, Beamer templates:

+ Add `babel-otherlangs` for language divs/spans; `babel-newcommands`,
filled by commands that make babel understand the polyglossia-style
language directives (Mauro Bieg, #137).
+ Improved formatting of conditionals; `$for$` is always provided to allow
multiple options (Andrew Dunning, #141).
+ Use `Ligatures=TeX` rather than `Mapping=tex-text` with `fontspec`
to improve support for LuaTeX (Andrew Dunning, #135).
+ Revise `hyperref` usage (Andrew Dunning, #139, #141):
- use same options for all LaTeX engines;
- add `subtitle` and `keywords` to PDF metadata;
- do not override `hyperref` link coloring without user input, effectively making
the `hidelinks` option the default (removed as a separate variable);
- link colors can be enabled (using a slightly darker version of the old
defaults) using a new `colorlinks` variable, automatically used by
the LaTeX writer when custom colors are specified;
- `pdfborder={0 0 0}` is automatically set by `hyperref` with
`colorlinks`, and is only applied if `colorlinks` is disabled.

* ConTeXt template (Andrew Dunning):

+ New variables for controlling styles: `linkstyle`, `linkcolor`,
`linkcontrastcolor`, `layout`, `pagenumbering`, `whitespace`, `indenting`,
`interlinespace`, `headertext`, `footertext`, `mainfont`, `sansfont`,
`monofont`, `mathfont`, `fontsize`.
+ Default template no longer supports MkII.
+ Improve writing of title block (suppressing numbering of first page).
+ Add `title` `subtitle`, `author`, `date`, `keywords` to PDF metadata.
+ Support `subtitle`, `abstract`.
+ Support list of figures (`lof`), list of tables (`lot`).
+ Disable link styling by default.
+ Define styles for all section types.
+ Enable microtype.
+ Improved formatting of conditionals.

* Beamer template: added code to prevent slide breaks inside paragraphs
(#2422, thanks to Nick Bart). This will matter, in practice, only when
`allowframebreaks` is used. It is especially helpful for bibliography

* OpenDocument template: Add `<office:automatic-styles>` tag around
automatic styles. The writer now longer provides this (see #2520).

* Restored Text.Pandoc.Compat.Monoid.

* Do not export (<>) from custom Prelude. The Prelude now matches
base 4.8 Prelude's API.

* Don't use custom prelude with ghc 7.10. Use the custom prelude
only for earlier versions. This change makes `stack ghci` and
`cabal repl` work (#2503), at least with ghc 7.10.

* Changed § to % in operators from Odt.Arrows.Utils (#2457).
This prevents problems building haddocks with "C" locale.

* Change default for old-locale flag to False.

* Use stack in deb, osx, and Windows package generators.

* Added Vagrantfile for building deb in vm.
This should help in automating binary package creation. 'make package'
will make the package. 'make package COMMIT=blah' will make the package
from commit blah.


+ Consistent capitalization for pandoc and Markdown.
+ Fixed `auto_identifiers` examples (Benoit Schweblin).
+ Improved documentation of template variables (Andrew Dunning).

pandoc (

* `Text.Pandoc.Data`: store paths in dataFiles using posix separators.
This way we have uniform separators, whether on Windows or Linux.
This should solve a problem where on some Windows versions
the data files weren't being found (#2459).

pandoc (1.15.1)

* `pandocVersion` is now defined in `Text.Pandoc.Shared`
and reexported from `Text.Pandoc` (Alex Vong). This allows
writers to access it. (Alex Vong) (API change)

* For `markdown_mmd`, add: `implicit_figures`, `superscripts`,
`subscripts` (#2401).

* Added `odt` as input format (Martin Linnemann). Added new module
`Text.Pandoc.Reader.ODT` (API change). Fully implemented features:
Paragraphs, Headers, Basic styling, Unordered lists, Ordered lists,
External Links, Internal Links, Footnotes, Endnotes, Blockquotes.
Partly implemented features: Citations, Tables.

* Markdown Reader:

+ Add basic tests for each header style (Ophir Lifshitz).
+ Add implicit header ref tests for headers with spaces (Ophir Lifshitz).
+ Skip spaces in headers (Ophir Lifshitz).
+ Handle 'id' and 'class' in parsing key/value attributes (#2396).
`# Header {id="myid" class="foo bar"}`
is now equivalent to `# Header {#myid .foo .bar}`.
+ Use '=' instead of '#' for atx-style headers in markdown+lhs.
(Kristof Bastiaensen)
+ Pipe tables: allow indented columns. Previously the left-hand column
could not start with 4 or more spaces indent. This was inconvenient
for right-aligned left columns. Note that the first (header column)
must still have 3 or fewer spaces indentation, or the table will be
treated as an indented code block.
+ Fix regression: allow HTML comments containing `--`.
Technically this isn't allowed in an HTML comment, but
we've always allowed it, and so do most other implementations.
It is handy if e.g. you want to put command line arguments
in HTML comments.

* LaTeX reader:

+ Don't eat excess whitespace after macros with only optional
arguments (#2446).
+ Support longtable (#2411).
+ Implement `\Cite` (#2335).
+ Support abstract environment. The abstract populates an
`abstract` metadata field.
+ Properly handle booktabs lines. Lines aren't part of the
pandoc table model, so we just ignore them (#2307).

* HTML reader:

+ Handle type attribute on ol, e.g. `<ol type="i">` (#2313).
+ Updated for new automatic header attributes.
+ Add auto identifiers if not present on headers. This makes
TOC linking work properly.
+ Detect `font-variant` with `pickStyleAttrProps` (Ophir Lifshitz).
+ Test `<ol>` type, class, and inline list-style(-type) CSS
(Ophir Lifshitz).
+ Better handling of "section" elements (#2438). Previously
`<section>` tags were just parsed as raw HTML blocks. With
this change, section elements are parsed as Div elements with
the class "section".

* MediaWiki reader: handle unquoted table attributes (#2355).

* DocBook reader:

+ Added proper support for DocBook `xref` elements (Frerich Raabe).
Added `dbContent` field to reader state, so we can lookup
cross refs.
+ Handle `informalexample` (#2319).

* Docx Reader:

+ Create special punctuation test (Ophir Lifshitz).
+ Parse soft, no-break hyphen elements (Ophir Lifshitz).
+ Updated headers test (Ophir Lifshitz). Replaced `styles.xml`
in `headers.docx` with pandoc's current `styles.xml`, which
contains styles for Heading 1 through 6. Added Heading 4
through 7 to the test document. Note that Heading 7 is not
parsed as a Heading because there is no Heading 7 style.

* RST reader: better handling of indirect roles.
Previously the parser failed on this kind of case

.. role:: indirect(code)

.. role:: py(indirect)
:language: python


Now it correctly recognizes `:py:` as a code role.

* Org reader:

+ Add auto identifiers if not present on headers
(#2354, Juliusz Gonera).
+ Allow verse blocks to contain empty lines (#2402,
Albert Krewinkel).

* EPUB reader: stop mangling external URLs (#2284).

* RST writer:

+ Don't insert `\ ` when complex expression in matched pairs.
E.g. `` [:sup:`3`] `` is okay; you don't need `` [:sup:`3`\ ] ``.
+ Ensure that `\ ` is inserted when needed before Cite and Span
elements that begin with a "complex" element (jgm/pandoc-citeproc#157).
+ Normalize headers only in "standalone" mode (#2394).

* Haddock writer: escape `*` and `^` (G. Bataille).

* Markdown writer:

+ In TOC, add links to headers (#829).
+ Use unicode super/subscripts for digits in plain output
(when the `superscripts` and `subscripts` extensions are
not enabled).

* Docx writer:

+ Moved invalid character stripping to `formattedString`.
This avoids an inefficient generic traversal (#2356).
+ Use user data directory for `reference.docx` archive.
This allows the test suite to work without installing pandoc first.
It also brings the docx writer in line with the odt writer.
+ Tests: docx writer tests now use `../data` for data directory.
This allows tests to be run without installing first.
+ Tests: Use real jpg (not empty) for docx tests to avoid warning.

* LaTeX writer:

+ Fixed detection of 'chapters' from template.
If a documentclass isn't specified in metadata, but the
template has a hardwired bookish documentclass, act as if
`--chapters` was used. This was the default in earlier
versions, but it has been broken for a little while.
+ Correctly recognize book documentclass in metadata (#2395).
+ Set language-related variables automatically, depending
on the value of the `lang` field, which is now always
assumed to be in BCP47 format (Mauro Bieg, #1614, #2437).
+ Add `\protect` to `\hyperdef` in inline context. This way we
don't get an error when this is used as a moveable argument (#2136).
+ Support all frame attributes in Beamer.
+ Percent-encode more special characters in URLs (#1640, #2377).
The special characters are '<','>','|','"','{','}','[',']','^', '`'.

* HTML writer:

+ Update KaTeX JS and CSS versions (Emily Eisenberg).
+ For dzslides, add `role="note"` for speaker notes (#1693).
+ Percent-encode more special characters in URLs (#1640, #2377).
The special characters are '<','>','|','"','{','}','[',']','^', '`'.
+ Render Div with class `section` as `<section>` in HTML5.

* EPUB writer:

+ In TOC, replace literal `<br/>` with space (#2105).
+ With `--webtex`, include image file rather than `data:` URI (#2363).

* Native writer: format Div properly, with blocks separated.

* Support bidirectional text output with XeLaTeX, ConTeXt and HTML
(#2191, Mauro Bieg).

* Reference Docx:

+ Add missing Header 6 style (steel blue) (Ophir Lifshitz).
+ Correct `outlineLvl` for Header styles (Ophir Lifshitz).

* Templates

+ Beamer: Add `innertheme`, `outertheme` variables
(Guilhem Bonnefille, #121). Add space after colon in figure caption.
Integrate recent font and language updates from LaTeX template;
allow use of `mainfont` variable for changing the slide text
in XeTeX and LuaTeX (Andrew Dunning, #131).
+ LaTeX: Add `mainfontoptions`, `sansfontoptions`,
`monofontoptions`, `mathfontoptions`, `fontfamilyoptions`
(Andrew Dunning, #122). Support handling of bidirectional
text (Mauro Bieg, #120). Improve reliability of superscripts/subscripts
under XeTeX and prevent letters and numbers from appearing on a
different baseline by removing use of the `realscripts` package
(via `xltxtra`). To restore use of OpenType characters for these
features under XeTeX or LuaTeX, add `\usepackage{realscripts}` to
`header-includes` (Andrew Dunning, #130). Remove redundant
reference to `xunicode` (Andrew Dunning, #130). Add `fontenc`,
`indent`, `subparagraph` variables (Andrew Dunning).
Allow use of `hidelinks` variable for `hyperref` package (Hugo Roy,
#113). Prevent package clash with `tufte-latex` and other classes that
include `hyperref` or `color` (Xavier Olive, #115).
+ ConTeXt: Support handling of bidirectional text (Mauro Bieg, #120).
+ LaTeX and ConTeXt: Use more specific language variables.
Instead of directly using `lang`, we now use `babel-lang` and
`polyglossia-lang` and `context-lang`. These variables are set by
the writers to the necessary values, based on the `lang` variable
(which now always takes a value in BCP47 format). (Mauro Bieg, #114,
+ HTML: Support handling of bidirectional text (Mauro Bieg, #120).
Move HTML5 shiv after CSS and fix URL (Andrew Dunning).
Add dir attribute in html5 (Andrew Dunning).
+ reveal.js: Add `controls`, `progress` variables (Grégoire Pineau, #127).
Add `width`, `height` variables (Andrew Dunning). Update template
from 3.1 source (Andrew Dunning). All configuration options are now
available as variables, but are only be included if set (reveal.js
uses defaults otherwise).
+ man: Added comment stating that the page is autogenerated by pandoc,
giving version. Added `adjusting` and `hyphenate` variables
(Alex Vong, #123).

* epub.css: added selectors for nested emphasis (Pablo Rodriguez).

* MediaBag: ensure that `/` is always used as path separator.

* `sample.lua`: define `CaptionedImage`, add newline at end (#2393).

* Added `--bash-completion` option. This generates a bash completion
script. To use: `eval "$(pandoc --bash-completion)"`.

* Text.Pandoc.Error: Define Typeable and Exception instances
for PandocError (#2386).

* Text.Pandoc.Parsing: `toKey`: strip off outer brackets.
This makes keys with extra space at the beginning and end
work: e.g.

[foo]: bar

[ foo ]

will now be a link to bar (it wasn't before).

* Text.Pandoc: disable `auto_identifiers` for epub.
The epub writer inserts its own auto identifiers;
this is more complex due to splitting into "chapter" files.

* Renamed Text.Pandoc.Compat.Locale -> Text.Pandoc.Compat.Time.
It now reexports Data.Time.

* Use custom Prelude to avoid compiler warnings.

+ The (non-exported) prelude is in prelude/Prelude.hs.
+ It exports Monoid and Applicative, like base 4.8 prelude,
but works with older base versions.
+ It exports (<>) for mappend.
+ It hides 'catch' on older base versions.

* Added a `stack.ymal` and stack install instructions to INSTALL.

* Clarified what is "out of scope" in README and

* Added note to about ghc versions and travis.

* Clarify docs on block quotes. The space after `>` is optional (#2346).

* Removed obsolete reference to default.csl (#2372).

* List all styles in manual for `--reference-docx` (Chris Black)

* Don't capitalize header links in man page.

* Added section on repl to

* README: Added space after backslash in image example (#2329).

* Document details of citation locator terms (Nick Bart).

* Fixed some internal links in README (#2309).

* Improve CSL documentation, variables documentations,
links, and cross-references in README. (Andrew Dunning)

* Fix build failure with `--flags=-https` (Sergei Trofimovich).

* Use `newManager` instead of `withManager` in recent `http-client`.
This avoids a deprecation warning.

* Allow building with latest versions of http-types,
HUnit, criterion, syb, aeson.

* Updated benchmark program for new criterion API.

* Setup.hs: rewrite so as not to use process, directory, filepath.
Using anything outside base is dangerous, since older
versions of ghc may link against two different versions.

* Added appveyor (Windows continuous integration) builds.

* New `.travis.yml`. Autgenerated using `make_travis_yml.hs`.
This script has been modified in a few ways, e.g. to add `GHCOPTS`.
`make .travis.yml` regenerates it based on the tested-with
field of the cabal file.

pandoc (

* `--self-contained`: Fixed overaggressive CSS minimization (#2301, 2286).
Previously `--self-contained` wiped out all spaces in CSS,
including semantically significant spaces. This was a regression
from 1.14.x.

* Markdown reader: don't allow bare URI links or autolinks in link
label (#2300). Added test cases.

* `Text.Pandoc.Parsing`, `uri`: Improved bare autolink detection (#2299).
Previously we disallowed `-` at the end of an autolink,
and disallowed the combination `=-`. This commit liberalizes the
rules for allowing punctuation in a bare URI, and adds test cases.
One potential drawback is that you can no longer put a bare
URI in em dashes like this:
`this uri--- an example.`
But in this respect we now match github's treatment of bare URIs.

* HTML writer: support speaker notes in dzslides.
With this change `<div class="notes">` and also `<div class="notes"
role="note">` will be output if `-t dzslides` is used. So we can
have speaker notes in dzslides too. Thanks to maybegeek.

* Updated dzslides template.

* Improved documentation of options to print system default files (#2298).
`--print-default-data-file` and `--print-default-template`.

* DokuWiki writer: use `$..$` for Math instead of `<math>..</math>`
(Tiziano Müller). MathJax seems currently to be the only maintained
math rendering extension for DokuWiki.

* `Text.Pandoc.Shared`: Changed `hierarchicalize` so it treats references
div as top-level header (#2294). This fixes a bug with `--section-divs`,
where the final references section added by pandoc-citeproc, enclosed in
its own div, got nested in the div for the section previous to it.

* Allow vector 0.11.

* Require cmark > 0.4.

pandoc (

* HTML writer: Fixed email javascript obfuscation with `mailto:`
URLs (#2280). This fixes a potential security issue. Because
single quotes weren't being escaped in the link portion, a
specially crafted email address could allow javascript code injection.

* Markdown/HTML readers: Avoid parsing partial URLs like
`<>` as HTML tags (#2277).

* RST reader: allow inline formatting in definition list field
names (Lars-Dominik Braun).

* PDF: Make sure `--latex-engine-opt` goes before the filename
on the command line. LaTeX needs the argument to come after
the options (#1779).

* CommonMark writer: fixed tags used for super/subscript.

* ConTeXt template: activate hanging indent for definition lists
(Mauro Bieg).

* Make cabal require `hsb2hs` >= 0.3.1 if `embed_data_files` specified.
This is done by adding `hookedPrograms` in `Setup.hs`, which allows us
to include `hsb2hs` in Build-Tools in cabal.

* Improved Windows installer (thanks to nkalvi).

+ When per-machine installation is chosen, the system path
is updated instead of the user's.
+ An appropriate default is used for per-machine installation
+ Admin privileges are no longer required for a per-user install

* Travis: unpack sdist for build to catch packaging bugs.

* Improved documentation on where user templates go (#2272).

pandoc (

* Added pandoc.1 man page to the repository. It is no longer
built as part of the cabal build process. (This proved too
fragile.) pandoc.1 can be regenerated (`make man/pandoc.1`)
when `README` is changed.

* Copying of the man page now respects `--destdir` (#2262).

* Improved error messages for filters. User is now informed if
the filter requires an interpreter that isn't found in the path,
or if the filter returns an error status.

pandoc (

* Ensure target directory is created when installing man page.

pandoc (

* Added files needed for building man page to Extra-Source-Files.

pandoc (

* Man page is now built and installed as part of the cabal build
process. Removed Makefile target for man page.

pandoc (1.15)

* Man page changes:

+ Removed `--man1`, `--man5` options (breaking change).
+ Removed `Text.Pandoc.ManPages` module (breaking API change).
+ Makefile target for `man/man1/pandoc.1`. This uses pandoc to
create the man page from README using a custom template and filters.
+ Added `man/` directory with template and filters needed to build
man page.
+ We no longer have two man pages: `pandoc.1` and `pandoc_markdown.5`.
Now there is just pandoc.1, which has all the content from README.
This change was needed because of the extensive cross-references
between parts of the README.
+ Removed old `data/pandoc.1.template` and

* OpenDocument writer: Do not add a carriage return after a hard
line break (Michael Chladek).

* ConTeXt writer:

+ use `\goto` for internal links.
+ Added a `%` at end for `\reference` to avoid spurious space.

* Ignore sandbox on 'make quick'

pandoc (1.14.1)

* Added `--man1` and `--man5` options to pandoc, allowing pandoc
to generate its own man pages. Man pages are no longer automatically
generated in the build process (the process for this was too complex
and prone to failure, #2190). The `make-pandoc-man-pages` executable
has been removed. The `man/` directory has been removed, and man page
templates have been moved to `data/`. NOTE TO PACKAGERS: You will no
longer find pandoc's man pages in `man/`, but you can generate them using
`pandoc --man1 > pandoc.1` and `pandoc --man5 > pandoc_markdown.5`.

* Added new unexported module: `Text.Pandoc.ManPages`.

* `README` now acts like a data file (even though it isn't in
`data/`). So, for example, `pandoc --print-default-data-file README`
will produce the README.) This change was required for the `--man1`
and `--man5` options, since the man pages are produced from the
README, but it may be useful for other purposes as well.

* Allow `reference.docx` and `reference.odt` to be used with
`--print-default-data-file` and to shadow defaults if placed in
the user data directory. Note that as of 1.14, we no longer
include these files as data files; instead, we include their
components. This change causes pandoc to behave as if it has
these data files; they are constructed on demand when needed
using `getDefaultReferenceDocx` and `getDefaultReferenceODT`.

* Fixed regression in CSS parsing with `--self-contained` (#2224).
Pandoc 1.14.0.x used css-text to parse the CSS, but its parser
silently drops big sections of CSS. This commit replaces the
use of css-text with a small but principled CSS preprocessor,
which removes whitespace and comments and replaces `url()` with
base 64 data when possible.

* Use `https://` instead of `//` for MathJax and KaTeX CDN URLs (#1920).
This will allow math to work when pages are being viewed locally.

* `Text.Pandoc.Options`: Export `plainExtensions`.
These are the extensions used in `plain` output.

* LaTeX reader: Don't parse `_` and `^` as sub/superscript outside of
math mode; treat them as regular inline text. Normally these will
cause an error in LaTeX, but there are contexts (e.g. `alltt`
environments) where they are allowed.

* HTML reader: allow `<body>` to close `<head>`.

* DocBook reader: support `mediaobject`s and `figures` (#2184, Mauro Bieg).

* RST reader: Fix reference names with special characters
(Lars-Dominik Braun).

* Textile writer: escape `+` and `-` as entities (#2225).

* DokuWiki writer: Use proper `<code>` tags for code blocks (#2213).

* Plain writer: don't use symbols for super/subscript (#2237).
Simplified code by using `plainExtensions`.

* InDesign writer: Properly escape URLs containing more than one
colon character (gohai).

* Docx writer: Make sure we use dist version of `reference.docx`
(and not the user's version) for certain settings. Taking some
settings values from a user-supplied reference.docx can lead to
corruption. This fixes a regression from the last release (#2249).

* `Text.Pandoc.Shared`: exports `getDefaultReferenceDocx` and
`getDefaultReferenceODT` (API change). These functions have been
removed from the Docx and ODT writers.

* LaTeX template (Xavier Olive):
+ Added `CJKmainfont` and `CJKoptions` variables.
+ Allow dvipsnames (e.g. `MidnightBlue`) for colors (Xavier Olive).

* Epub templates: use `author.role`, not `author.type`.

* Bump cmark version to >= 0.3.4.

* Improved Windows installer (#2205, thanks to nkalvi).
Users can now select a per-user or systemwide install, and can set
the installation path. At the end of installation, the install location
is given. The install location is also now given in the list of
installed programs in Control Panel. Cleaner WiX syntax is used for
setting the path.

* Added `download_stats` target to Makefile.

pandoc (

* Added missing commonmark template.

* Improved try pandoc (moved button, show raw command).

pandoc (

* Allow compilation with syb 0.5.*.

* Custom writer: fixed some compiler warnings for ghc < 7.10.

pandoc (

* Allow building with hslua 0.4.

pandoc (

* Fixed problem with building of `reference.docx` and `reference.odt`
when the `embed_data_files` flag is used. Instead of having a phase
of the build where `reference.docx` and `reference.odt` are created
from their constituent data files, we now construct these archives
from their constituents when a `docx` or `odt` is built. The
constituent files have been moved from `extra-source-files` to
`data-files`, and `reference.docx` and `reference.odt` have been
removed. Users can create their own `reference.docx` or
`reference.odt` by using pandoc to create a simple `docx` or `odt`.
`make-reference-files.hs` has been removed, simplifying the build
process (#2187)

* Don't include generated man pages in extra-source-files (#2189).

* Bumped upper bound for aeson.

* ConTeXt writer: create internal link anchors for Div elements with
identifiers. (This is needed for linked citations to work.)

pandoc (1.14)

[new features]

* Added `commonmark` as input and output format.

* Added `--verbose` flag for debugging output in PDF production (#1840,

* Allow wildcards in `--epub-embed-font` arguments (#1939).

* Added `--latex-engine-opt` option (#969, #1779, Sumit Sahrawat).

* Added `shortcut_reference_links` extension (Konstantin Zudov, #1977).
This is enabled by default for those markdown flavors that
support reading shortcut reference links, namely: `markdown`,
`markdown_strict`, `markdown_github`, `markdown_php`.
If the extension is enabled, the reader parses shortcut reference
links like `[foo]`, and the writer creates such links unless doing
so would cause problems. Users of markdown flavors that support
shortcut reference links should not notice a difference in reading
markdown, but the markdown pandoc produces may differ.
If shortcut links are not desired, the extension can be disabled
in the normal way.

[behavior changes]

* `--toc` is now supported for `docx` output (#458, Nikolay Yakimov).
A "dirty" TOC is created at the beginning of document.
It can be regenerated after the document has been opened.

* An implicit `--filter pandoc-citeproc` is now triggered only when the
`--bibliography` option is used, and not when the `bibliography`
field in metadata is specified (#1849).

* Markdown reader:

+ Reference links with `implicit_header_references` are no longer
case-sensitive (#1606).
+ Definition lists no longer require indentation for first line (#2087).
Previously the body of the definition (after the `:` or `~` marker)
needed to be in column 4. This commit relaxes that requirement,
to better match the behavior of PHP Markdown Extra. So, now
this is a valid definition list:

: bar
+ Resolve a potentially ambiguity with table captions:


: bar


Is "bar" a definition, or the caption for the table? We'll count
it as a caption for the table.
+ Disallow headerless pipe tables (#1996), to conform to GFM and PHP
Markdown Extra. Note: If you have been using headerless pipe tables,
this change may cause existing tables to break.
+ Allow pipe tables with header but no body (#2017).
+ Allow a digit as first character of a citation key (Matthias Troffaes).

* LaTeX reader:

+ Don't limit includes to `.tex` extension (#1882).
If the extension is not `.tex`, it must be given explicitly in
the `\input` or `\include`.

* Docx reader:

+ Allow numbering in the style file. This allows inherited styles
with numbering (lists) (Jesse Rosenthal).

* Org reader:

+ Support smart punctuation (Craig Bosma).
+ Drop trees with a :noexport: tag (Albert Krewinkel). Trees having a
`:noexport:` tag set are not exported. This mirrors org-mode.
+ Put header tags into empty spans (Albert Krewinkel, #2160).
Org mode allows headers to be tagged: `* Headline :TAG1:TAG2`.
Instead of being interpreted as part of the headline, the tags are now
put into the attributes of empty spans. Spans without textual content
won't be visible by default, but they are detectable by filters. They
can also be styled using CSS when written as HTML.
+ Generalize code block result parsing (Albert Krewinkel).
Previously, only code blocks were recognized as result blocks;
now, any kind of block can be the result.

* Append newline to the LineBreak in Dokuwiki, HTML, EPUB,
LaTeX, MediaWiki, OpenDocument, Texinfo writers (#1924, Tim Lin).

* HTML writer:

+ Add "inline" or "display" class to math spans (#1914).
This allows inline and display math to be styled differently.
+ Include raw latex blocks if `--mathjax` specified (#1938).
+ Require highlighting-kate >= 0.5.14 (#1903).
This ensures that all code blocks will be wrapped in a `div`
with class `sourceCode`. Also, the default highlighting CSS
now adds `div.sourceCode { x-overflow: auto; }`, which means
that code blocks (even with line numbers) will acquire a scroll
bar on screens too small to display them (e.g. mobile phones).
See also jgm/highlighting-kate#65.

* LaTeX writer:

+ Use a declaration for tight lists (Jose Luis Duran, Joseph
Harriott). Previously, pandoc hard-coded some commands to make
tight lists in LaTeX. Now we use a custom command instead,
allowing the styling to be changed in a macro in the header.
(Note: existing templates may need to be modified to include
the definition of this macro. See the current template.)
+ Beamer output: if the header introducing a slide has the
class `fragile`, add the `[fragile]` option to the slide (#2119).

* MediaWiki writer:

+ Use `File:` instead of the deprecated `Image:` for images and
other media files (Greg Rundlett).

* DocBook writer:

+ Render a `Div (id,_,_) [Para _]` element as a `para` element
with an `id` attribute. This makes links to citations work in
DocBook with pandoc-citeproc.

* RST writer:

+ Normalize headings to sequential levels (Nikolay Yakimov).
This is pretty much required by docutils.
+ Treat headings in block quotes, etc as rubrics (Nikolay Yakimov).
+ Better handling of raw latex inline (#1961). We use
`` :raw-latex:`...` `` and add a definition for this role to
the template.

* EPUB writer:

+ Remove `linear=no` from cover `itemref` (#1609).
+ Don't use `sup` element for epub footnotes (#1995).
Instead, just use an a element with class `footnoteRef`.
This allows more styling options, and provides better results
in some readers (e.g. iBooks, where anything inside the a
tag breaks popup footnotes).
+ Take TOC title from `toc-title` metadata field.

* Docx writer:

+ Implemented `FirstParagraph` style (Jesse Rosenthal).
Following the ODT writer, we add the `FirstParagraph` style to the
first text paragraph following an image, blockquote, table, heading,
or beginning of document. This allows it to be styled differently.
The default is for it to be the same as `Normal`.
+ Added `BodyText` style (Jesse Rosenthal).
We apply a `BodyText` style to all unstyled paragraphs. This is,
essentially, the same as `Normal`, except that since not everything
inherits from `BodyText` (the metadata won't, for example, or
the headers or footnote numbers), we can change the text in the body
without having to make exceptions for everything. If we do want to
change *everything*, we can still do it through `Normal`.
+ Altered `Blockquote` style slightly (Jesse Rosenthal).
Since `BlockQuote` derives from `BodyText`, we just want to specify
by default that it won't indent, regardless of what `BodyText` does.
Note that this will not produce any visible difference in the default
+ Take TOC title from `toc-title` metadata field (Nikolay Yakimov).
+ Added a style to figure images (Nikolay Yakimov).
Figures with empty captions use style `Figure`.
Figures with nonempty captions use style `Figure with Caption`, which
is based on `Figure`, and additionally has `keepNext` set.

* ODT writer:

+ Added figure captions (Nikolay Yakimov). The following styles are
used for figures:
`Figure` -- for figure with empty caption),
`FigureWithCaption` (based on `Figure`) -- for figure with caption,
`FigureCaption` (based on `Caption`) -- for figure captions.
Also, `TableCaption` (based on `Caption`) is used for table captions.

[API changes]

* New `Text.Pandoc.Error` module with `PandocError` type
(Matthew Pickering).

* All readers now return `Either PandocError Pandoc` instead of `Pandoc`
(Matthew Pickering). This allows better handling of errors.

* Added `Text.Pandoc.Writers.CommonMark`, exporting `writeCommonMark`.

* Added `Text.Pandoc.Readers.CommonMark`, exporting `readCommonMark`.

* Derive `Data` and `Typeable` instances for `MediaBag`, `Extension`,
`ReaderOptions`, `EPUBVersion`, `CiteMethod`, `ObfuscationMethod`,
`HTMLSlideVariant`, `TrackChanges`, `WriterOptions` (Shabbaz

* New `Ext_shortcut_reference_links` constructor for `Extension`
(Konstantin Zudov).

[bug fixes]

* Markdown reader:

+ Allow smart `'` after inline math (#1909, Nikolay Yakimov).
+ Check for tex macros after indented code (#1973).
+ Rewrote `charsInBalancedBrackets` for efficiency.
+ Make sure a closing `</div>` doesn't get included in a
definition list item (#2127).
+ Don't parse bracketed text as citation if it might be a link,
image, or footnote (Nikolay Yakimov).
+ Require space after key in mmd title block (#2026, Nikolay
Yakimov). Require space after key-value delimiter colon in mmd title
+ Require nonempty value in mmd title block (Nikolay Yakimov).
+ Disable all metadata block extensions when parsing
metadata field values (#2026, Nikolay Yakimov). Otherwise we
could get a mmd title block inside YAML metadata, for example.

* HTML reader:

+ Improve self-closing tag detection in `htmlInBalanced` (#2146).
+ Handle tables with `<th>` in body rows (#1859, Mauro Bieg).
+ Fixed `htmlTag` (#1820). If the tag parses as a comment, we check
to see if the input starts with `<!--`. If not, it's bogus comment
mode and we fail `htmlTag`.
+ Handle `base` tag; if it has an `href` value, this is added to
all relative URLs in links and images.

* DocBook reader:

+ Look inside "info" elements for section titles (#1931).

* Docx reader:

+ Parse images in deprecated vml format (Jesse Rosenthal).
+ Allow sub/superscript verbatims (Jesse Rosenthal).
Verbatim usually shuts off all other run styles, but we don't want it
to shut off sub/superscript.

* LaTeX reader:

+ Handle `tabular*` environment (#1850).
Note that the table width is not actually parsed or taken into
account, but pandoc no longer chokes on it.
+ Ignore options in `\lstinline` rather than raising error (#1997).
+ Add some test cases for simple tables (Mathias Schenner).
+ Handle valign argument in tables (Mathias Schenner) (currently
we just ignore this).
+ Allow non-empty colsep in tables (Mathias Schenner).
The `tabular` environment allows non-empty column separators
with the "@{...}" syntax. Previously, pandoc would fail to
parse tables if a non-empty colsep was present. With this
commit, these separators are still ignored, but the table gets
parsed. A test case is included.
+ Recognize `\newpage` as a block command.
+ Allow block content in \title{} (#2001).
+ Check for block-level newcommand aliases in blockCommand (Nikolay
+ Guard against paragraph starting with inline macro (Nikolay Yakimov).
+ Properly gobble spaces after `\\` (#2007).

* Textile reader:

+ Handle newlines in table cells, and empty cells (#1919).

* Org reader:

+ Allow image links with non-image targets (Hans-Peter Deifel).
This matches behavior of Org-Mode for links like

* Docbook writer:

+ Don't print empty id attributes (thanks to Steve Horne).

* HTML writer:

+ Fixed list-style-type for numbered example lists.
Should be "decimal," not "example" (#1902).
+ Do not omit missing `alt` attribute on `img` tag (#1131,
Konstantin Zudov).
+ Allow multiple colgroups in table (#2122).
+ In revealjs, ensure that lists in speaker notes don't add "fragment"
classes, which can cause additional keypresses to be needed to
advance a slide (#1394).

* LaTeX writer:

+ Don't escape `$` in URL (#1913).
+ Don't use listings in headers (Matthew Pickering, #1963).
+ Recognize book documentclass if set in metadata (#1971).
This sets `--chapters` implicitly if the documentclass in metadata
is a book documentclass. Previously this was done only if a book
documentclass was set in a variable.
+ Add a `\label` in `\hyperdef` for Div, Span (or links don't work).
+ Make `mainlang` work when `lang` is in metadata (#2174).

* Texinfo writer:

+ Fix wrapping by using breakable spaces (Tim Lin).

* RST writer:

+ Fixed toc depth in RST writer. Previously the depth was being
rendered as a floating point number with a decimal point.

* Markdown writer:

+ Improved escaping (#2086). `<` should not be escaped as `\<`, for
compatibility with original Markdown. We now escape `<` and `>`
with entities. Also, we now backslash-escape square brackets.
+ Avoid introducing spurious list items through wrapping (#1946).
+ Don't emit span tags if plain or raw HTML disabled.

* MediaWiki writer:

+ Convert spaces to underscores in wikilink URL (#1982), like MediaWiki.

* AsciiDoc writer:

+ Insert some needed blank lines (#1860).
+ Avoid wrapping after list marker (#1858).

* EPUB writer:

+ Properly handle internal links to IDs in spans, divs (#1884).
+ Use plain writer for metadata dc: fields (#2121).
This gives better results when we have, e.g. multiple paragraphs.
Note that tags aren't allowed in these fields.
+ Properly handle image links without an extension (#1855).
+ Improved chapter splitting and internal link rewriting (#1887,
#2162, #2163). This will ensure that internal links work and
that the references section produced by pandoc-citeproc is
in its own chapter.
+ Fixed handling of svg images (#2183).

* ICML writer:

+ Better handling of raw blocks and inlines (#1951).
Previously these were always escaped and printed verbatim.
Now they are ignored unless the format is `icml`, in which
case they are passed through unescaped.
+ Fixed image URIs in ICML output (gohai).

* Custom writer:

+ Raise error if loadstring returns an error status.
+ Raise `PandocLuaException` instead of using 'error'.
Eventually we'll change the return type so that no exception
is involved, but at least this can be trapped.
+ Use UTF-8 aware bytestring conversion.
+ Set foreign encoding to UTF-8 (Nikolay Yakimov, #2101, #1634).
Also factored out ByteString, since it's only used as an intermediate

* Docx writer:

+ Copy hyphenation settings from reference.docx (Nikolay Yakimov).
+ Filter out illegal XML characters (#1992, Matthew Pickering).
+ Added `noProof` to docx syntax highlighting `SourceCode` style.
+ Added footnotes id -1 and 0 (Jesse Rosenthal).
Word uses, by default, footnotes with id -1 and 0 for separators. If a
user modifies `reference.docx`, they will end up with a `settings.xml`
file that references these footnotes, but no such footnotes in the
document. This will produce a corruption error. Here we add these to the
document and `settings.xml` file, so future modifications won't break
the file.
+ Handle lists correctly inside table cells (Jesse Rosenthal).
Previously we didn't transform lists inside table cells.
+ Set firstRow information in tables (Nikolay Yakimov).
+ Don't replace `SourceCode` style in `reference.docx` if it is defined
there (Nikolay Yakimov, #1872). If `--no-highlight` specified, remove
any `SourceCode` and `*Tok` styles in `reference.docx`.
+ Attempt to match international style names (#1607, Nikolay Yakimov).
+ Set these styles as custom (Nikolay Yakimov): `Author`, `Abstract`,
`Compact`, `Image Caption`, `Table Caption`, `Definition Term`,
`Definition`, `First Paragraph`.
+ Rename these styles to correspond with Word `Normal.dotm` (Nikolay
Yakimov): `Block Quote -> Block Text`, `Link -> Hyperlink`,
`Footnote Ref -> Footnote Reference`.
+ Added `Caption` style (Nikolay Yakimov).
+ Changed these styles' inheritance (Nikolay Yakimov):
`Image Caption <- Caption`, `Table Caption <- Caption`.
+ Remove `SourceCode` style from `reference.docx` (#1872).
This is added automatically by the docx writer.
+ Added toc heading style to `reference.docx` (Nikolay Yakimov).

* `Text.Pandoc.PDF`

+ Don't suggest "Try xelatex" if xelatex already in use (Mauro Bieg,
+ More comprehensible errors on image conversion (#2067).
EPS can't be supported without shelling out to something like
ImageMagick, but at least we can avoid mysterious error messages.

* `Text.Pandoc.Shared`:

+ Make safeRead safe (#1801, Matthew Pickering).
+ Added `mapLeft`, `hush` (Matthew Pickering).

* `Text.Pandoc.Pretty`:

+ Remove partial function (Matthew Pickering).

* `Text.Pandoc.SelfContained`:

+ Add `;charset=utf-8` to script mime type if missing (#1842).
+ Improved building of data URIs (#1940). Now base64 is used except
for `text/*` mime types.
+ `cssURLs` no longer tries to fetch fragment URLs (#2121).
+ Properly handle data URIs in css urls (#2129).
Use a proper CSS parser (adds dependency on `text-css`).

* `Text.Pandoc.UTF8`:

+ Better handling of bare CRs in input files (#2132).
Previously we just stripped them out; now we convert
other line ending styles to LF line endings.

* `Text.Pandoc.ImageSize`:

+ Fixed some exif header parsing bugs (#1834).
+ Make imageSize return an Either, not a Maybe (#1834).
Use `runGetOrFail` (with `binary >= 0.7`) to return `Left` on
parse failure (rather than `error`).
+ Improved warnings when image size can't be determined.
+ Removed error landmines (Matthew Pickering).

* Added woff2 to MIME types (Alfred Wechselberger).

* pandoc: When a binary input format is used, warn that file
arguments past the first one are being ignored (Matthew Pickering).

[template changes]

* LaTeX template:

+ Degrade gracefully if `\paragraph` not defined.
+ Include `grffile` together with `graphicx` (#2074).
This properly handles filenames containing spaces and dots.
+ Redefine `\paragraph`, `\subparagraph`... to behave more
like section headers (#1658).
+ Import hyperref before polyglossia to avoid an error with xelatex,
"please load package hyperref before bidi package" (Nick Bart).
+ Added `toccolor` variable to control link color in toc (Kaixhin).

* LaTeX, Beamer templates:

+ Provide `\tightlist`, which is now used by the LaTeX writer.
+ Use polyglossia in beamer (#85).
+ Use `bibliography` instead of `biblio-files`
(#1661). Also use `\addbibresource` instead of `\bibliography` for
+ Added `setotherlanguages` in polyglossia. This uses an `otherlang`
variable that is derived from a comma-separated list in `lang`;
the last language is `mainlang` and the others are `otherlang`.

* EPUB templates:

+ Use `div`, not `p`, for "rights" on title page.
+ Added header-includes, include-before, include-after (#1987).

* OpenDocument template:

+ Use `text:p` instead of `text:h` for title.
Using `text:h` causes problems with numbering. Closes #2059.
Thanks to @nkalvi for diagnosing this.

* reveal.js template:

+ Link to non-minified css, js. The minified versions no longer
ship with the library.
+ Correctly include style CSS (#1949).
+ New configurable options options: `center`, `maxScale`, `slideNuber`
(Dmitry Smirnov, pandoc-templates#89).
+ Moved custom CSS after theme. This allows custom CSS to modify
themes, instead of being replaced by them.
+ Allow `center` to be set to false.

[under the hood improvements]

* Removed pre-built `reference.docx` and `reference.odt` (Nikolay
Yakimov). Instead the repository now includes the component text files,
and the zipped binaries are built from these using a helper
program, `make-reference-files`. This should make maintenance of
these components easier going forward.

* `Text.Pandoc.Parsing`:

+ Added new `<+?>` combinator (Nikolay Yakimov).
+ Added `stateHeaderKeys` to `ParserState`.

* `` fixes:

+ Detect architecture.
+ Add Installed-Size to debian package control file (#1900).
+ Use `fakeroot` to get permissions right.
+ Use `mkdir` and `cp` instead of `install`.
+ Set permissions of directories to 755.
+ Install in `/usr` rather than `/usr/local`.
+ Compress man pages.
+ Combine copyright files for `pandoc`, `pandoc-citeproc`.

* Added `Text.Pandoc.Compat.Locale` and `old-locale` flag
to assist with transition to `time` 1.5.

* Updated with information about issue tags (Matthew

* Updated travis installs to the new sudo-less syntax (Tim Lin).

* Updated dependency version bounds.

* EPUB tests: don't use `joinPath`, which varies across platforms.
Instead, use a forward-slash to join paths, regardless of the
platform. This matches the way `MediaBag` now works.

* Clarify JSON input and output in usage message (Caleb McDaniel).

* Improved INSTALL instructions.

* Always build man pages. Removed make-pandoc-man-pages flag.

* Makefile: removed man target, now that we generate man pages by default.


+ Fixed typos (J. Lewis Muir).
+ Added documentation on backtick_code_blocks (#2135, Nikolay Yakimov).
+ Added note on in-field markup in biblio databases (Nick Bart).
+ Fixed misleading example of raw HTML block.
+ Various minor formatting and consistency fixes for the program
options (Andreas Lööw).
+ Made definition lists for options all "loose" for consistency.
+ Added YAML biblio format to table, and note on `pandoc-citeproc`'s
`--bib2json` and `--bib2yaml` options (Nick Bart).
+ Removed obsolete reference to `mods2yaml` (Nick Bart).
+ Added section on syntax highlighting.
+ Documented `toccolor` variable.

pandoc (

* Updated to build with ghc 7.10.1.

* Bumped package upper bounds for filepath, blaze-html,

pandoc (1.13.2)

* TWiki Reader: add new twiki reader (API change, Alexander Sulfrian).

* Markdown reader:

+ Better handling of paragraph in div (#1591).
Previously text that ended a div would be parsed as Plain
unless there was a blank line before the closing div tag.
+ Don't treat a citation as a reference link label (#1763).
+ Fix autolinks with following punctuation (#1811).
The price of this is that autolinked bare URIs can no longer
contain `>` characters, but this is not a big issue.
+ Fix `Ext_lists_without_preceding_blankline` bug (#1636, Artyom).
+ Allow `startnum` to work without `fancy_lists`. Formerly
`pandoc -f markdown-fancy_lists+startnum` did not work properly.

* RST reader (all Daniel Bergey):

+ Parse quoted literal blocks (#65). RST quoted literal blocks are
the same as indented literal blocks (which pandoc already supports)
except that the quote character is preserved in each line.
+ Parse RST class directives. The class directive accepts one or more
class names, and creates a Div value with those classes. If the
directive has an indented body, the body is parsed as the children of
the Div. If not, the first block following the directive is made a
child of the Div. This differs from the behavior of rst2xml, which
does not create a Div element. Instead, the specified classes are
applied to each child of the directive. However, most Pandoc Block
constructors to not take an Attr argument, so we can't duplicate this
+ Warn about skipped directives.
+ Literal role now produces Code. Code role should have "code" class.
+ Improved support for custom roles

- Added `sourceCode` to classes for `:code:` role, and anything
inheriting from it.
- Add the name of the custom role to classes if the Inline
constructor supports Attr.
- If the custom role directive does not specify a parent role,
inherit from the `:span:` role.

This differs somewhat from the `` behavior. If a custom
role inherits from another custom role, Pandoc will attach both
roles' names as classes. `` will only use the class of
the directly invoked role (though in the case of inheriting from a
`:code:` role with a `:language:` defined, it will also provide the
inherited language as a class).
+ Warn about ignored fields in role directives.

* LaTeX reader:

+ Parse label after caption into a span instead of
inserting an additional paragraph of bracketed text (#1747).
+ Parse math environments as inline when possible (#1821).
+ Better handling of `\noindent` and `\greektext` (#1783).
+ Handle `\texorpdfstring` more gracefully.
+ Handle `\cref` and `\sep` (Wikiwide).
+ Support `\smartcite` and `\Smartcite` from biblatex.

* HTML reader:

+ Retain display type of MathML output (#1719, Matthew Pickering).
+ Recognise `<br>` tags inside `<pre>` blocks (#1620, Matthew Pickering).
+ Make `embed` tag either block or inline (#1756).

* DocBook reader:

+ Handle `keycombo`, `keycap` (#1815).
+ Get string content in inner tags for literal elements (#1816).
+ Handle `menuchoice` elements better, with a `>` between (#1817).
+ Include `id` on section headers (#1818).
+ Document/test "type" as implemented (Brian O'Sullivan).
+ Add support for calloutlist and callout (Brian O'Sullivan).
We treat a calloutlist as a bulleted list. This works well in practice.
+ Add support for `classname` (Bryan O'Sullivan).

* Docx reader:

+ Fix window path for image lookup (Jesse Rosenthal).
Don't use os-sensitive "combine", since we always want the paths in our
zip-archive to use forward-slashes.
+ Single-item headers in ordered lists are headers (Jesse Rosenthal).
When users number their headers, Word understands that as a single item
enumerated list. We make the assumption that such a list is, in fact,
a header.
+ Rewrite rewriteLink to work with new headers (Jesse Rosenthal).
There could be new top-level headers after making lists, so we have to
rewrite links after that.
+ Use polyglot header list (Jesse Rosenthal).
We're just keeping a list of header formats that different languages
use as their default styles. At the moment, we have English, German,
Danish, and French. We can continue to add to this.
This is simpler than parsing the styles file, and perhaps less
error-prone, since there seems to be some variations, even within a
language, of how a style file will define headers.
+ Remove header class properly in other langs (Jesse Rosenthal).
When we encounter one of the polyglot header styles, we want to remove
that from the par styles after we convert to a header. To do that, we
have to keep track of the style name, and remove it appropriately.
+ Account for external link URLs with anchors. Previously, if a URL
had an anchor, the reader would incorrectly identify it as an
internal link and return only the anchor as URL. (Caleb McDaniel)
+ Fix for Issue #1692 (i18n styles) (Nikolay Yakimov).

* Org reader:

+ Added state changing blanklines (Jesse Rosenthal).
This allows us to emphasize at the beginning of a new paragraph (or, in
general, after blank lines).
+ Fixed bug with bulleted lists:

- a
- b
* c

was being parsed as a list, even though an unindented `*`
should make a heading. See
+ Org reader: absolute, relative paths in link (#1741, Albert
Krewinkel). The org reader was too restrictive when parsing links;
some relative links and links to files given as absolute paths
were not recognized correctly.
+ Org reader: allow empty links (jgm/gitit#471, Albert Krewinkel).
This is important for use in gitit, which uses empty links
for wikilinks.
+ Respect indent when parsing Org bullet lists (#1650, Timothy
Humphries). Fixes issue with top-level bullet list parsing.
+ Fix indent issue for definition lists (Timothy Humphries,
see #1650, #1698, #1680).
+ Parse multi-inline terms correctly in definition list (#1649,
Matthew Pickering).
+ Fix rules for emphasis recognition (Albert Krewinkel).
Things like `/hello,/` or `/hi'/` were falsy recognized as emphasised
strings. This is wrong, as `,` and `'` are forbidden border chars and
may not occur on the inner border of emphasized text.
+ Drop COMMENT document trees (Albert Krewinkel).
Document trees under a header starting with the word `COMMENT` are
comment trees and should not be exported. Those trees are dropped
silently (#1678).
+ Properly handle links to `file:target` (Albert Krewinkel).
Org links like `[[file:target][title]]` were not handled correctly,
parsing the link target verbatim. The org reader is changed such that
the leading `file:` is dropped from the link target (see #756, #1812).
+ Parse LaTeX-style MathML entities (#1657, Albert Krewinkel).
Org supports special symbols which can be included using LaTeX syntax,
but are actually MathML entities. Examples for this are
`\nbsp` (non-breaking space), `\Aacute` (the letter A with accent acute)
or `\copy` (the copyright sign ©)

* EPUB reader:

+ URI handling improvements. Now we outsource most of the work to
`fetchItem'`. Also, do not include queries in file extensions (#1671).

* LaTeX writer:

+ Use `\texorpdfstring` for section captions when needed (Vaclav Zeman).
+ Handle consecutive linebreaks (#1733).
+ Protect graphics in headers (Jesse Rosenthal).
Graphics in `\section`/`\subsection` etc titles need to be `\protect`ed.
+ Put `~` before header in list item text (Jesse Rosenthal).
Because of the built-in line skip, LaTeX can't handle a section header
as the first element in a list item.
+ Avoid using reserved characters as `\lstinline` delimiters (#1595).
+ Better handling of display math in simple tables (#1754).
We convert display math to inline math in simple tables,
since LaTeX can't deal with display math in simple tables.
+ Escape spaces in code (#1694, Bjorn Buckwalter).

* MediaWiki writer:

+ Fixed links with URL = text. Previously these were rendered as bare
words, even if the URL was not an absolute URL (#1825).

* ICML writer:

+ Don't force all citations into footnotes.

* RTF writer:

+ Add blankline at end of output (#1732, Matthew Pickering).

* RST writer:

+ Ensure blank line after figure.
+ Avoid exces whitespace after last list item (#1777).
+ Wrap line blocks with spaces before continuations (#1656).
+ Fixed double-rendering of footnotes in RST tables (#1769).

* DokuWiki writer:

+ Better handling of block quotes. This change ensures that
multiple paragraph blockquotes are rendered using native `>`
rather than as HTML (#1738).
+ Fix external images (#1739). Preface relative links with ":",
absolute URIs without. (Timothy Humphries)

* HTML writer:

+ Use protocol-relative URL for mathjax.
+ Put newline btw img and caption paragraph.
+ MathML now outputted with tex annotation (#1635, Matthew Pickering).
+ Add support for KaTeX HTML math (#1626, Matthew Pickering).
This adds `KaTeX` to `HTMLMathMethod` (API change).
+ Don't double render when `email-obfuscation=none` (#1625, Matthew
+ Make header attributes work outside top level (#1711).
Previously they only appeared on top level header elements.
Now they work e.g. in blockquotes.

* ODT writer:

+ Correctly handle images without extensions (#1729).
+ Strip querystring in ODT write (#1682, Todd Sifleet).

* FB2 writer:

+ Add newline to output.

* EPUB writer:

+ Don't add `sourceURL` to absolute URIs (#1669).
+ Don't use unsupported `opf:title-type` for epub2.
+ Include "landmarks" section in nav document for epub3 (#1757).
+ Removed playOrder from navpoint elements in ncx file (#1760).
These aren't required, and they make manual modification of epubs
+ Extract title even from structured title.
+ Don't include nav node in spine unless `--toc` was requested.
Previously we included it in the spine with `linear="no"`, leading
to odd results in some readers (#1593).
+ Fixed absolute URI detection (#1672).
+ Correctly resolve relative URIs (#1671).
+ Use regular page template for `nav.xhtml`, including doctype (#1759).

* Docx writer:

+ Put docx table captions above tables (#1641, Nikolay Yakimov).
+ Get the page width from the reference docx file, and use
it to scale images that are too large to fit (Grégory Bataille).
+ Partial fix for #1607 (Nikolay Yakimov). International heading styles
are inferred based on `<w:name val="heading #">` fallback, if there
are no en-US "Heading#" styles
+ Look in user data dir for archive `reference.docx`.
+ Renumber header and footer relationships to avoid collisions (Jesse
Rosenthal). We previously took the old relationship names of the
headers and footer in secptr. That led to collisions. We now make
a map of availabl names in the relationships file, and then rename
in secptr.

* ConTeXt writer:

+ Add function toLabel (Mark Szepieniec).
This function can be used to sanitize reference labels so that
they do not contain any of the illegal characters \#[]",{}%()|= .
Currently only Links have their labels sanitized, because they
are the only Elements that use passed labels.

* `Text.Pandoc.Shared`:

+ Moved import of `toChunks` outside of CPP conditional (#1590).
+ Fix `inDirectory` to reset to the original directory in case
an exception occurs (Freiric Barral).

* Templates:

+ LaTeX template: load polyglossia before bibtex (jgm/pandoc-templates#70).
Thanks to bluebirch.
+ LaTeX template: Added `\VerbatimFootnotes` if there is verbatim in notes
+ LaTeX template: Add shorthands=off to babel options (#1648).
+ EPUB, EPUB3 templates: Added `id="cover"` to body of cover page.
This aids styling, making it possible for example to set 0 margins
on the title page (#1758).
+ EPUB, EPUB3 templates: Handle structured metadata on titlepage.
Previously we just expected 'title', 'subtitle', 'author', 'date'.
Now we still support those, but also support the format recommended
for epub metadata in the pandoc README:

- type: main
text: My Book
- type: subtitle
text: An investigation of metadata
- role: author
text: John Smith
- role: editor
text: Sarah Jones
- scheme: DOI
text: doi:10.234234.234/33
publisher: My Press
rights: (c) 2007 John Smith, CC BY-NC

* `Text.Pandoc.Templates.getDefaultTemplate`:
don't fail when called with "fb2" (#1660).

* `Text.Pandoc.Parsing`:

+ Fixed `inlineMath` so it handles `\text{..}` containing `$`.
For example: `$x = \text{the $n$th root of $y$}` (#1677).
+ Change `parseFromString` to fail if not all input is consumed.
(Matthew Pickering)
+ Moved `addWarning` from Markdown reader to `Parsing`, so it can be
used by more readers (API change, Daniel Bergey).

* `Text.Pandoc.Pretty`:

+ Improve performance of `realLength` (Matthew Pickering).
+ Make CR + BLANKLINE = BLANKLINE. This fixes an extra blank line we
were getting at the end of markdown fragments (as well as rst, org,
etc.) (#1705).

* `Text.Pandoc.MIME`:

+ Add mime type for WebVTT (Jason Ronallo).
+ Changed mime type for `otf` to `application/` (#1761).
This is needed for epub3 validation.

* `Text.Pandoc.MediaBag`:

+ Fix Windows specific path problems (#1597).

* `Text.Pandoc.Shared`:

+ Make `collapseFilePath` OS-agnostic (Matthew Pickering).

* Link the test suite using `-threaded`.
This allows the test suite to be run using `+RTS -N`.

* Added `network` dependency under `network-uri` flag in test section.

* Give better error messages when someone tries to convert from
pdf, doc, odt (#1683).

* Added `track` to list of tags treated by `--self-contained` (#1664).

pandoc (1.13.1)

* Fixed `--self-contained` with Windows paths (#1558).
Previously `C:\foo.js` was being wrongly interpreted as a URI.

* HTML reader: improved handling of tags that can be block or inline.
Previously a section like this would be enclosed in a paragraph,
with RawInline for the video tags (since video is a tag that can
be either block or inline):

<video controls="controls">
<source src="../videos/test.mp4" type="video/mp4" />
<source src="../videos/test.webm" type="video/webm" />
The videos can not be played back on your system.<br/>
Try viewing on Youtube (requires Internet connection):
<a href="">Relative Velocity on

This change will cause the video and source tags to be parsed
as RawBlock instead, giving better output.
The general change is this: when we're parsing a "plain" sequence
of inlines, we don't parse anything that COULD be a block-level tag.

* Docx reader:

+ Be sensitive to user styles. Note that "Hyperlink" is
"blacklisted," as we don't want the default underline styling to be
inherited by all links by default (Jesse Rosenthal).
+ Read single paragraph in table cell as `Plain` (Jesse Rosenthal).
This makes to docx reader's native output fit with the way the markdown
reader understands its markdown output.

* Textile writer: Extended the range of cases where native textile
tables will be used (as opposed to raw HTML): we now handle any
alignment type, but only for simple tables with no captions.

* Txt2Tags reader:

+ Header is now parsed only if standalone flag is set (Matthew Pickering).
+ The header is now parsed as meta information. The first line is the
`title`, the second is the `author` and third line is the `date`
(Matthew Pickering).
+ Corrected formatting of `%%mtime` macro (Matthew Pickering).
+ Fixed crash when reading from stdin.

* EPUB writer: Don't use page-progression-direction in EPUB2, which
doesn't support it. Also, if page-progression-direction not specified
in metadata, don't include the attribute even in EPUB3; not including it
is the same as including it with the value "default", as we did before.

* Org writer: Accept example lines with indentation at the beginning
(Calvin Beck).

* DokuWiki writer:

+ Refactor to use Reader monad (Matthew Pickering).
+ Avoid using raw HTML in table cells; instead, use `\\`
instead of newlines (Jesse Rosenthal).
+ Properly handle HTML table cell alignments, and use spacing
to make the tables look prettier (#1566).

* Docx writer:

+ Bibliography entries get `Bibliography` style (#1559).
+ Implement change tracking (Jesse Rosenthal).

* LaTeX writer:

+ Fixed a bug that caused a table caption to repeat across all pages
(Jose Luis Duran).
+ Improved vertical spacing in tables and made it customizable using
standard lengths set by booktab. See
(Jose Luis Duran).
+ Added `\strut` to fix spacing in multiline tables (Jose Luis Duran).
+ Use `\tabularnewline` instead of `\\` in table cells (Jose Luis Duran).
+ Made horizontal rules more flexible (Jose Luis Duran).

* Text.Pandoc.MIME:

+ Added `MimeType` (type synonym for `String`) and `getMimeTypeDef`.
Code cleanups (Artyom Kazak).

* Templates:

+ LaTeX template: disable microtype protrusion for typewriter font (#1549,
thanks lemzwerg).

* Improved OSX build procedure.

* Added `network-uri` flag, to deal with split of `network-uri` from

* Fix build dependencies for the `trypandoc` flag, so that they are
ignored if `trypandoc` flag is set to False (Gabor Pali).

* Updated README to remove outdated claim that `--self-contained`
looks in the user data directory for missing files.

pandoc (

* Docx writer:

+ Fixed regression which bungled list numbering (#1544), causing
all lists to appear as basic ordered lists.
+ Include row width in table rows (Christoffer Ackelman, Viktor Kronvall).
Added a property to all table rows where the sum of column widths
is specified in pct (fraction of 5000). This helps persuade Word
to lay out the table with the widths we specify.

* Fixed a bug in Windows 8 which caused pandoc not to find the
`pandoc-citeproc` filter (#1542).

* Docx reader: miscellaneous under-the-hood improvements (Jesse Rosenthal).
Most significantly, the reader now uses Builder, leading to some
performance improvements.

* HTML reader: Parse appropriately styled span as SmallCaps.

* Markdown writer: don't escape `$`, `^`, `~` when `tex_math_dollars`,
`superscript`, and `subscript` extensions, respectively, are
deactivated (#1127).

* Added `trypandoc` flag to build CGI executable used in the online

* Makefile: Added 'quick', 'osxpkg' targets.

* Updated README in templates to indicate templates license.
The templates are dual-licensed, BSD3 and GPL2+.

pandoc (1.13)

[new features]

* Added `docx` as an input format (Jesse Rosenthal). The docx
reader includes conversion of native Word equations to pandoc
LaTeX `Math` elements. Metadata is taken from paragraphs at the
beginning of the document with styles `Author`, `Title`, `Subtitle`,
`Date`, and `Abstract`.

* Added `epub` as an input format (Matthew Pickering). The epub
reader includes conversion of MathML to pandoc LaTeX `Math`

* Added `t2t` (Txt2Tags) as an input format (Matthew Pickering).
Txt2tags is a lightweight markup format described at

* Added `dokuwiki` as an output format (Clare Macrae).

* Added `haddock` as an output format.

* Added `--extract-media` option to extract media contained in a zip
container (docx or epub) while adjusting image paths to point to the
extracted images.

* Added a new markdown extension, `compact_definition_lists`, that
restores the syntax for definition lists of pandoc 1.12.x, allowing
tight definition lists with no blank space between items, and
disallowing lazy wrapping. (See below under behavior changes.)

* Added an extension `epub_html_exts` for parsing HTML in EPUBs.

* Added extensions `native_spans` and `native_divs` to activate
parsing of material in HTML span or div tags as Pandoc Span
inlines or Div blocks.

* `--trace` now works with the Markdown, HTML, Haddock, EPUB,
Textile, and MediaWiki readers. This is an option intended
for debugging parsing problems; ordinary users should not need
to use it.

[behavior changes]

* Changed behavior of the `markdown_attribute` extension, to bring
it in line with PHP markdown extra and multimarkdown. Setting
`markdown="1"` on an outer tag affects all contained tags,
recursively, until it is reversed with `markdown="0"` (#1378).

* Revised markdown definition list syntax (#1429). Both the reader
and writer are affected. This change brings pandoc's definition list
syntax into alignment with that used in PHP markdown extra and
multimarkdown (with the exception that pandoc is more flexible about
the definition markers, allowing tildes as well as colons). Lazily
wrapped definitions are now allowed. Blank space is required
between list items. The space before a definition is used to determine
whether it is a paragraph or a "plain" element. **WARNING: This change
may break existing documents!** Either check your documents for
definition lists without blank space between items, or use
`markdown+compact_definition_lists` for the old behavior.

* `.numberLines` now works in fenced code blocks even if no language
is given (#1287, jgm/highlighting-kate#40).

* Improvements to `--filter`:

+ Don't search PATH for a filter with an explicit path.
This fixed a bug wherein `--filter ./` would run `` from
the system path, even if there was a `` in the working directory.
+ Respect shebang if filter is executable (#1389).
+ Don't print misleading error message.
Previously pandoc would say that a filter was not found,
even in a case where the filter had a syntax error.

* HTML reader:

+ Parse `div` and `span` elements even without `--parse-raw`,
provided `native_divs` and `native_spans` extensions are set.
Motivation: these now generate native pandoc Div and Span
elements, not raw HTML.
+ Parse EPUB-specific elements if the `epub_html_exts`
extension is enabled. These include `switch`, `footnote`,
`rearnote`, `noteref`.

* Org reader:

+ Support for inline LaTeX. Inline LaTeX is now accepted and parsed by the
org-mode reader. Both math symbols (like `\tau`) and LaTeX commands (like
`\cite{Coffee}`), can be used without any further escaping (Albert

* Textile reader and writer:

+ The `raw_tex` extension is no longer set by default. You can
enable it with `textile+raw_tex`.

* DocBook reader:

+ Support `equation`, `informalequation`, `inlineequation` elements with
`mml:math` content. This is converted into LaTeX and put into a Pandoc
Math inline.

* Revised `plain` output, largely following the style of Project

+ Emphasis is rendered with `_underscores_`, strong emphasis
with ALL CAPS.
+ Headings are rendered differently, with space to set them off,
not with setext style underlines. Level 1 headers are ALL CAPS.
+ Math is rendered using unicode when possible, but without the
distracting emphasis markers around variables.
+ Footnotes use a regular `[n]` style.

* Markdown writer:

+ Horizontal rules are now a line across the whole page.
+ Prettier pipe tables. Columns are now aligned (#1323).
+ Respect the `raw_html` extension. `pandoc -t markdown-raw_html`
no longer emits any raw HTML, including span and div tags
generated by Span and Div elements.
+ Use span with style for `SmallCaps` (#1360).

* HTML writer:

+ Autolinks now have class `uri`, and email autolinks have class
`email`, so they can be styled.

* Docx writer:

+ Document formatting is carried over from `reference.docx`.
This includes margins, page size, page orientation, header,
and footer, including images in headers and footers.
+ Include abstract (if present) with `Abstract` style (#1451).
+ Include subtitle (if present) with `Subtitle` style, rather
than tacking it on to the title (#1451).

* Org writer:

+ Write empty span elements with an id attribute as org anchors.
For example `Span ("uid",[],[]) []` becomes `<<uid>>`.

* LaTeX writer:

+ Put table captions above tables, to match the conventional
standard. (Previously they appeared below tables.)
+ Use `\(..\)` instead of `$..$` for inline math (#1464).
+ Use `\nolinkurl` in email autolinks. This allows them to be styled
using `\urlstyle{tt}`. Thanks to Ulrike Fischer for the solution.
+ Use `\textquotesingle` for `'` in inline code. Otherwise we get
curly quotes in the PDF output (#1364).
+ Use `\footnote<.>{..}` for notes in beamer, so that footnotes
do not appear before the overlays in which their markers appear
+ Don't produce a `\label{..}` for a Div or Span element. Do produce
a `\hyperdef{..}` (#1519).

* EPUB writer:

+ If the metadata includes `page-progression-direction` (which can be
`ltr` or `rtl`, the `page-progression-direction` attribute will
be set in the EPUB spine (#1455).

* Custom lua writers:

+ Custom writers now work with `--template`.
+ Removed HTML header scaffolding from `sample.lua`.
+ Made citation information available in lua writers.

* `--normalize` and `Text.Pandoc.Shared.normalize` now consolidate
adjacent `RawBlock`s when possible.

[API changes]

* Added `Text.Pandoc.Readers.Docx`, exporting `readDocx` (Jesse Rosenthal).

* Added `Text.Pandoc.Readers.EPUB`, exporting `readEPUB` (Matthew

* Added `Text.Pandoc.Readers.Txt2Tags`, exporting `readTxt2Tags` (Matthew

* Added `Text.Pandoc.Writers.DokuWiki`, exporting `writeDokuWiki`
(Clare Macrae).

* Added `Text.Pandoc.Writers.Haddock`, exporting `writeHaddock`.

* Added `Text.Pandoc.MediaBag`, exporting `MediaBag`, `lookupMedia`,
`insertMedia`, `mediaDirectory`, `extractMediaBag`. The docx and epub
readers return a pair of a `Pandoc` document and a `MediaBag` with
the media resources they contain. This can be extracted using
`--extract-media`. Writers that incorporate media (PDF, Docx,
ODT, EPUB, RTF, or HTML formats with `--self-contained`) will look
for resources in the `MediaBag` generated by the reader, in addition to
the file system or web.

* `Text.Pandoc.Readers.TexMath`: Removed deprecated `readTeXMath`.
Renamed `readTeXMath'` to `texMathToInlines`.

* `Text.Pandoc`: Added `Reader` data type (Matthew Pickering).
`readers` now associates names of readers with `Reader`
structures. This allows inclusion of readers, like the docx
reader, that take binary rather than textual input.

* `Text.Pandoc.Shared`:

+ Added `capitalize` (Artyom Kazak), and replaced uses of
`map toUpper` (which give bad results for many languages).
+ Added `collapseFilePath`, which removes intermediate `.` and
`..` from a path (Matthew Pickering).
+ Added `fetchItem'`, which works like `fetchItem` but searches
a `MediaBag` before looking on the net or file system.
+ Added `withTempDir`.
+ Added `removeFormatting`.
+ Added `extractSpaces` (from HTML reader) and generalized its type
so that it can be used by the docx reader (Matthew Pickering).
+ Added `ordNub`.
+ Added `normalizeInlines`, `normalizeBlocks`.
+ `normalize` is now `Pandoc -> Pandoc` instead of
`Data a :: a -> a`. Some users may need to change their uses of
`normalize` to the newly exported `normalizeInlines` or

* `Text.Pandoc.Options`:

+ Added `writerMediaBag` to `WriterOptions`.
+ Removed deprecated and no longer used `readerStrict` in
`ReaderOptions`. This is handled by `readerExtensions` now.
+ Added `Ext_compact_definition_lists`.
+ Added `Ext_epub_html_exts`.
+ Added `Ext_native_divs` and `Ext_native_spans`.
This allows users to turn off the default pandoc behavior of
parsing contents of div and span tags in markdown and HTML
as native pandoc Div blocks and Span inlines.

* `Text.Pandoc.Parsing`:

+ Generalized `readWith` to `readWithM` (Matthew Pickering).
+ Export `runParserT` and `Stream` (Matthew Pickering).
+ Added `HasQuoteContext` type class (Matthew Pickering).
+ Generalized types of `mathInline`, `smartPunctuation`, `quoted`,
`singleQuoted`, `doubleQuoted`, `failIfInQuoteContext`,
`applyMacros` (Matthew Pickering).
+ Added custom `token` (Matthew Pickering).
+ Added `stateInHtmlBlock` to `ParserState`. This is used to keep
track of the ending tag we're waiting for when we're parsing inside
HTML block tags.
+ Added `stateMarkdownAttribute` to `ParserState`. This is used
to keep track of whether the markdown attribute has been set in
an enclosing tag.
+ Generalized type of `registerHeader`, using new type classes
`HasReaderOptions`, `HasIdentifierList`, `HasHeaderMap` (Matthew
Pickering). These allow certain common functions to be reused
even in parsers that use custom state (instead of `ParserState`),
such as the MediaWiki reader.
+ Moved `inlineMath`, `displayMath` from Markdown reader to Parsing,
and generalized their types (Matthew Pickering).

* `Text.Pandoc.Pretty`:

+ Added `nestle`.
+ Added `blanklines`, which guarantees a certain number of blank lines
(and no more).

[bug fixes]

* Markdown reader:

+ Fixed parsing of indented code in list items. Indented code
at the beginning of a list item must be indented eight spaces
from the margin (or edge of the container), or four spaces
from the list marker, whichever is greater.
+ Fixed small bug in HTML parsing with `markdown_attribute`, which
caused incorrect tag nesting for input like
`<aside markdown="1">*hi*</aside>`.
+ Fixed regression with intraword underscores (#1121).
+ Improved parsing of inline links containing quote characters (#1534).
+ Slight rewrite of `enclosure`/`emphOrStrong` code.
+ Revamped raw HTML block parsing in markdown (#1330).
We no longer include trailing spaces and newlines in the
raw blocks. We look for closing tags for elements (but without
backtracking). Each block-level tag is its own `RawBlock`;
we no longer try to consolidate them (though `--normalize` will do so).
+ Combine consecutive latex environments. This helps when you have
two minipages which can't have blank lines between them (#690, #1196).
+ Support smallcaps through span.
`<span style="font-variant:small-caps;">foo</span>` will be
parsed as a `SmallCaps` inline, and will work in all output
formats that support small caps (#1360).
+ Prevent spurious line breaks after list items (#1137). When the
`hard_line_breaks` option was specified, pandoc would formerly
produce a spurious line break after a tight list item.
+ Fixed table parsing bug (#1333).
+ Handle `c++` and `objective-c` as language identifiers in
github-style fenced blocks (#1318).
+ Inline math must have nonspace before final `$` (#1313).

* LaTeX reader:

+ Handle comments at the end of tables. This resolves the issue
illustrated in <>.
+ Correctly handle table rows with too few cells. LaTeX seems to
treat them as if they have empty cells at the end (#241).
+ Handle leading/trailing spaces in `\emph` better.
`\emph{ hi }` gets parsed as `[Space, Emph [Str "hi"], Space]`
so that we don't get things like `* hi *` in markdown output.
Also applies to `\textbf` and some other constructions (#1146).
+ Don't assume preamble doesn't contain environments (#1338).
+ Allow (and discard) optional argument for `\caption` (James Aspnes).

* HTML reader:

+ Fixed major parsing problem with HTML tables. Table cells were
being combined into one cell (#1341).
+ Fixed performance issue with malformed HTML tables.
We let a `</table>` tag close an open `<tr>` or `<td>` (#1167).
+ Allow space between `<col>` and `</col>`.
+ Added `audio` and `source` in `eitherBlockOrInline`.
+ Moved `video`, `svg`, `progress`, `script`, `noscript`, `svg` from
`blockTags` to `eitherBlockOrInline`.
+ `map` and `object` were mistakenly in both lists; they have been removed
from `blockTags`.
+ Ignore `DOCTYPE` and `xml` declarations.

* MediaWiki reader:

+ Don't parse backslash escapes inside `<source>` (#1445).
+ Tightened up template parsing.
The opening `{{` must be followed by an alphanumeric or `:`.
This prevents the exponential slowdown in #1033.
+ Support "Bild" for images.

* DocBook reader:

+ Better handle elements inside code environments. Pandoc's document
model does not allow structure inside code blocks, but at least this way
we preserve the text (#1449).
+ Support `<?asciidoc-br?>` (#1236).

* Textile reader:

+ Fixed list parsing. Lists can now start without an intervening
blank line (#1513).
+ HTML block-level tags that do not start a line are parsed as
inline HTML and do not interrupt paragraphs (as in RedCloth).

* Org reader:

+ Make tildes create inline code (#1345). Also relabeled `code` and
`verbatim` parsers to accord with the org-mode manual.
+ Respect `:exports` header argument in code blocks (Craig Bosma).
+ Fixed tight lists with sublists (#1437).

* EPUB writer:

+ Avoid excess whitespace in `nav.xhtml`. This should improve
TOC view in iBooks (#1392).
+ Fixed regression on cover image.
In 1.12.4 and, the cover image would not appear properly,
because the metadata id was not correct. Now we derive the id from the
actual cover image filename, which we preserve rather than using
+ Keep newlines between block elements. This allows
easier diff-ability (#1424).
+ Use `stringify` instead of custom `plainify`.
+ Use `renderTags'` for all tag rendering. This properly handles tags
that should be self-closing. Previously `<hr/>` would appear in EPUB
output as `<hr></hr>` (#1420).
+ Better handle HTML media tags.
+ Handle multiple dates with OPF `event` attributes. Note: in EPUB3 we
can have only one dc:date, so only the first one is used.

* LaTeX writer:

+ Correctly handle figures in notes. Notes can't contain figures in
LaTeX, so we fake it to avoid an error (#1053).
+ Fixed strikeout + highlighted code (#1294).
Previously strikeout highlighted code caused an error.

* ConTeXt writer:

+ Improved detection of autolinks with URLs containing escapes.

* RTF writer:

+ Improved image embedding: `fetchItem'` is now used to get the
images, and calculated image sizes are indicated in the RTF.
+ Avoid extra paragraph tags in metadata (#1421).

* HTML writer:

+ Deactivate "incremental" inside slide speaker notes (#1394).
+ Don't include empty items in the table of contents for
slide shows. (These would result from creating a slide
using a horizontal rule.)

* MediaWiki writer:

+ Minor renaming of `st` prefixed names.

* AsciiDoc writer:

+ Double up emphasis and strong emphasis markers in intraword
contexts, as required by asciidoc (#1441).

* Markdown writer:

+ Avoid wrapping that might start a list, blockquote, or header (#1013).
+ Use Span instead of (hackish) `SmallCaps` in `plainify`.
+ Don't use braced attributes for fenced code (#1416).
If `Ext_fenced_code_attributes` is not set, the first class
attribute will be printed after the opening fence as a bare word.
+ Separate adjacent lists of the same kind with an HTML comment (#1458).

* PDF writer:

+ Fixed treatment of data uris for images (#1062).

* Docx writer:

+ Use Compact style for empty table cells (#1353).
Otherwise we get overly tall lines when there are empty
table cells and the other cells are compact.
+ Create overrides per-image for `media/` in reference docx.
This should be somewhat more robust and cover more types of images.
+ Improved `entryFromArchive` to avoid an unneeded parse.
+ Section numbering carries over from reference.docx (#1305).
+ Simplified `abstractNumId` numbering. Instead of sequential numbering,
we assign numbers based on the list marker styles.

* `Text.Pandoc.Options`:

+ Removed `Ext_fenced_code_attributes` from `markdown_github`

* `Text.Pandoc.ImageSize`:

+ Use default instead of failing if image size not found
in exif header (#1358).
+ ignore unknown exif header tag rather than crashing.
Some images seem to have tag type of 256, which was causing
a runtime error.

* `Text.Pandoc.Shared`:

+ `fetchItem`: unescape URI encoding before reading local file (#1427).
+ `fetchItem`: strip a fragment like `?#iefix` from the extension before
doing mime lookup, to improve mime type guessing.
+ Improved logic of `fetchItem`: absolute URIs are fetched from the net;
other things are treated as relative URIs if `sourceURL` is `Just _`,
otherwise as file paths on the local file system.
+ `fetchItem` now properly handles links without a protocol (#1477).
+ `fetchItem` now escapes characters not allowed in URIs before trying
to parse the URIs.
+ Fixed runtime error with `compactify'DL` on certain lists (#1452).

* `pandoc.hs`: Don't strip path off of `writerSourceURL`: the path is
needed to resolve relative URLs when we fetch resources (#750).

* `Text.Pandoc.Parsing`

+ Simplified `dash` and `ellipsis` (#1419).
+ Removed `(>>~)` in favor of the equivalent `(<*)` (Matthew Pickering).
+ Generalized functions to use `ParsecT` (Matthew Pickering).
+ Added `isbn` and `pmid` to list of recognized schemes (Matthew

[template changes]

* Added haddock template.
* EPUB3: Added `type` attribute to `link` tags. They are supposed to
be "advisory" in HTML5, but kindlegen seems to require them.
* EPUB3: Put title page in section with `epub:type="titlepage"`.
* LaTeX: Made `\subtitle` work properly (#1327).
* LaTeX/Beamer: remove conditional around date (#1321).
* LaTeX: Added `lot` and `lof` variables, which can be set to
get `\listoftables` and `\listoffigures` (#1407). Note that
these variables can be set at the command line with `-Vlot -Vlof`
or in YAML metadata.

[under the hood improvements]

* Rewrote normalize for efficiency (#1385).

* Rewrote Haddock reader to use `haddock-library` (#1346).

+ This brings pandoc's rendering of haddock markup in line
with the new haddock.
+ Fixed line breaks in `@` code blocks.
+ alex and happy are no longer build-depends.

* Added `Text.Pandoc.Compat.Directory` to allow building against
different versions of the `directory` library.

+ Added `Text.Pandoc.Compat.Except` to allow building against
different versions of `mtl`.

* Code cleanup in some writers, using Reader monad to avoid
passing options parameter around (Matej Kollar).

* Improved readability in `pandoc.hs`.

* Miscellaneous code cleanups (Artyom Kazak).

* Avoid `import Prelude hiding (catch)` (#1309, thanks to Michael

* Changed `http-conduit` flag to `https`. Depend on `http-client`
and `http-client-tls` instead of `http-conduit`. (Note: pandoc still
depends on `conduit` via `yaml`.)

* Require `highlighting-kate >=` (#1271, #1317, Debian #753299).
This change to highlighting-kate means that PHP fragments no longer need
to start with `<?php`. It also fixes a serious bug causing failures with
ocaml and fsharp.

* Require latest `texmath`. This fixes `\tilde{E}` and allows
`\left` to be used with `]`, `)` etc. (#1319), among many other

* Require latest `zip-archive`. This has fixes for unicode path names.

* Added tests for plain writer.

* `Text.Pandoc.Templates`:

+ Fail informatively on template syntax errors.
With the move from parsec to attoparsec, we lost good error
reporting. In fact, since we weren't testing for end of input,
malformed templates would fail silently. Here we revert back to
Parsec for better error messages.
+ Use `ordNub` (#1022).

* Benchmarks:

+ Made benchmarks compile again (Artyom Kazak).
+ Fixed so that the failure of one benchmark does not prevent others
from running (Artyom Kazak).
+ Use `nfIO` instead of the `getLength` trick to force full evaluation.
+ Changed benchmark to use only the test suite, so that benchmarks
run more quickly.

* Windows build script:

+ Add `-windows` to file name.
+ Use one install command for pandoc, pandoc-citeproc.
+ Force install of pandoc-citeproc.

* `make_osx_package`: Call zip file ``.
The zip should not be named ``, or OSX finder
will extract it into a folder named `SOMETHING.pkg`, which it
will interpret as a defective package (#1308).


+ Made headers for all extensions so they have IDs and can be
linked to (Beni Cherniavsky-Paskin).
+ Fixed typos (Phillip Alday).
+ Fixed documentation of attributes (#1315).
+ Clarified documentation on small caps (#1360).
+ Better documentation for `fenced_code_attributes` extension
(Caleb McDaniel).
+ Documented fact that you can put YAML metadata in a separate file

pandoc (

* Require highlighting-kate >= 0.5.8. Fixes a performance regression.

* Shared: `addMetaValue` now behaves slightly differently:
if both the new and old values are lists, it concatenates their
contents to form a new list.

* LaTeX reader:

+ Set `bibliography` in metadata from `\bibliography` or
`\addbibresource` command.
+ Don't error on `%foo` with no trailing newline.

* Org reader:

+ Support code block headers (`#+BEGIN_SRC ...`) (Albert Krewinkel).
+ Fix parsing of blank lines within blocks (Albert Krewinkel).
+ Support pandoc citation extension (Albert Krewinkel). This can
be turned off by specifying `org-citations` as the input format.

* Markdown reader:

+ `citeKey` moved to `Text.Pandoc.Parsing` so it can be used by
other readers (Albert Krewinkel).

* `Text.Pandoc.Parsing`:

+ Added `citeKey` (see above).
+ Added `HasLastStrPosition` type class and `updateLastStrPos`
and `notAfterString` functions.

* Updated copyright notices (Albert Krewinkel).

* Added default.icml to data files so it installs with the package.

* OSX package:

+ The binary is now built with options to ensure that it can be
used with OSX 10.6+.
+ Moved OSX package materials to osx directory.
+ Added OSX package uninstall script, included in the zip container
(thanks to Daniel T. Staal).

pandoc (1.12.4)

* Made it possible to run filters that aren't executable (#1096).
Pandoc first tries to find the executable (searching the path
if path isn't given). If it fails, but the file exists and has
a `.py`, `.pl`, `.rb`, `.hs`, or `.php` extension, pandoc runs the filter
using the appropriate interpreter. This should make it easier to
use filters on Windows, and make it more convenient for everyone.

* Added Emacs org-mode reader (Albert Krewinkel).

* Added InDesign ICML Writer (Mauro Bieg).

* MediaWiki reader:

+ Accept image links in more languages (Jaime Marquínez Ferrándiz).
+ Fixed bug in certain nested lists (#1213). If a level 2 list was
followed by a level 1 list, the first item of the level 1 list
would be lost.
+ Handle table rows containing just an HTML comment (#1230).

* LaTeX reader:

+ Give better location information on errors, pointing to line
numbers within included files (#1274).
+ LaTeX reader: Better handling of `table` environment (#1204).
Positioning options no longer rendered verbatim.
+ Better handling of figure and table with caption (#1204).
+ Handle `@{}` and `p{length}` in tabular. The length is not actually
recorded, but at least we get a table (#1180).
+ Properly handle `\nocite`. It now adds a `nocite` metadata
field. Citations there will appear in the bibliography but not
in the text (unless you explicitly put a `$nocite$` variable
in your template).

* Markdown reader:

+ Ensure that whole numbers in YAML metadata are rendered without
decimal points. (This became necessary with changes to aeson
and yaml libraries. aeson >= 0.7 and yaml >= are now required.)
+ Fixed regression on line breaks in strict mode (#1203).
+ Small efficiency improvements.
+ Improved parsing of nested `div`s. Formerly a closing `div` tag
would be missed if it came right after other block-level tags.
+ Avoid backtracking when closing `</div>` not found.
+ Fixed bug in reference link parsing in `markdown_mmd`.
+ Fixed a bug in list parsing (#1154). When reading a raw list
item, we now strip off up to 4 spaces.
+ Fixed parsing of empty reference link definitions (#1186).
+ Made one-column pipe tables work (#1218).

* Textile reader:

+ Better support for attributes. Instead of being ignored, attributes
are now parsed and included in Span inlines. The output will be a bit
different from stock textile: e.g. for `*(foo)hi*`, we'll get
`<em><span class="foo">hi</span></em>` instead of
`<em class="foo">hi</em>`. But at least the data is not lost.
+ Improved treatment of HTML spans (%) (#1115).
+ Improved link parsing. In particular we now pick up on attributes.
Since pandoc links can't have attributes, we enclose the whole link in
a span if there are attributes (#1008).
+ Implemented correct parsing rules for inline markup (#1175, Matthew
+ Use Builder (Matthew Pickering).
+ Fixed list parsing bug (#1500).
+ Don't allow inline formatting to extend over newlines.
This matches the behavior of RedCarpet, avoids some ugly bugs,
and improves performance.

* DocBook reader:

+ Better treatment of `formalpara`. We now emit the title (if present)
as a separate paragraph with boldface text (#1215).
+ Set metadata `author` not `authors`.
+ Added recognition of `authorgroup` and `releaseinfo` elements (#1214,
Matthew Pickering).
+ Converted current meta information parsing in DocBook to a more
extensible version which is aware of the more recent meta
representation (Matthew Pickering).

* HTML reader:

+ Require tagsoup 0.13.1, to fix a bug with parsing of script tags
+ Treat processing instructions & declarations as block. Previously
these were treated as inline, and included in paragraph tags in HTML
or DocBook output, which is generally not what is wanted (#1233).
+ Updated `closes` with rules from HTML5 spec.
+ Use Builder (Matthew Pickering, #1162).

* RST reader:

+ Remove duplicate `http` in PEP links (Albert Krewinkel).
+ Make rst figures true figures (#1168, CasperVector)
+ Enhanced Pandoc's support for rST roles (Merijn Verstaaten).
rST parser now supports: all built-in rST roles, new role definition,
role inheritance, though with some limitations.
+ Use `author` rather than `authors` in metadata.
+ Better handling of directives. We now correctly handle field
lists that are indented more than three spaces. We treat an
`aafig` directive as a code block with attributes, so it can be
processed in a filter (#1212).

* LaTeX writer:

+ Mark span contents with label if span has an ID (Albert Krewinkel).
+ Made `--toc-depth` work well with books in latex/pdf output (#1210).
+ Handle line breaks in simple table cells (#1217).
+ Workaround for level 4-5 headers in quotes. These previously produced
invalid LaTeX: `\paragraph` or `\subparagraph` in a `quote` environment.
This adds an `mbox{}` in these contexts to work around the problem.
See <> (#1221).
+ Use `\/` to avoid en-dash ligature instead of `-{}-` (Vaclav Zeman).
This is to fix LuaLaTeX output. The `-{}-` sequence does not avoid the
ligature with LuaLaTeX but `\/` does.
+ Fixed string escaping in `hyperref` and `hyperdef` (#1130).

* ConTeXt writer: Improved autolinks (#1270).

* DocBook writer:

+ Improve handling of hard line breaks in Docbook writer
(Neil Mayhew). Use a `<literallayout>` for the entire paragraph, not
just for the newline character.
+ Don't let line breaks inside footnotes influence the enclosing
paragraph (Neil Mayhew).
+ Distinguish tight and loose lists in DocBook output, using
`spacing="compact"` (Neil Mayhew, #1250).

* Docx writer: When needed files are not present in the user's
`reference.docx`, fall back on the versions in the `reference.docx`
in pandoc's data files. This fixes a bug that occurs when a
`reference.docx` saved by LibreOffice is used. (#1185)

* EPUB writer:

+ Include extension in epub ids. This fixes a problem with duplicate
extensions for fonts and images with the same base name but different
extensions (#1254).
+ Handle files linked in raw `img` tags (#1170).
+ Handle media in `audio` source tags (#1170).
Note that we now use a `media` directory rather than `images`.
+ Incorporate files linked in `video` tags (#1170). `src` and `poster`
will both be incorporated into `content.opf` and the epub container.

* HTML writer:

+ Add colgroup around col tags (#877). Also affects EPUB writer.
+ Fixed bug with unnumbered section headings. Unnumbered section
headings (with class `unnumbered`) were getting numbers.
+ Improved detection of image links. Previously image links with
queries were not recognized, causing `<embed>` to be used instead
of `<img>`.

* Man writer: Ensure that terms in definition lists aren't line wrapped

* Markdown writer:

+ Use proper escapes to avoid unwanted lists (#980). Previously we used
0-width spaces, an ugly hack.
+ Use longer backtick fences if needed (#1206). If the content contains a
backtick fence and there are attributes, make sure longer fences are
used to delimit the code. Note: This works well in pandoc, but github
markdown is more limited, and will interpret the first string of three
or more backticks as ending the code block.

* RST writer: Avoid stack overflow with certain tables (#1197).

* RTF writer: Fixed table cells containing paragraphs.

* Custom writer:

+ Correctly handle UTF-8 in custom lua scripts (#1189).
+ Fix bugs with lua scripts with mixed-case filenames and
paths containing `+` or `-` (#1267). Note that `getWriter`
in `Text.Pandoc` no longer returns a custom writer on input

* AsciiDoc writer: Handle multiblock and empty table cells
(#1245, #1246). Added tests.

* `Text.Pandoc.Options`: Added `readerTrace` to `ReaderOptions`

* `Text.Pandoc.Shared`:

+ Added `compactify'DL` (formerly in markdown reader) (Albert Krewinkel).
+ Fixed bug in `toRomanNumeral`: numbers ending with '9' would
be rendered as Roman numerals ending with 'IXIV' (#1249). Thanks to
Jesse Rosenthal.
+ `openURL`: set proxy with value of http_proxy env variable (#1211).
Note: proxies with non-root paths are not supported, due to
limitations in `http-conduit`.

* `Text.Pandoc.PDF`:

+ Ensure that temp directories deleted on Windows (#1192). The PDF is
now read as a strict bytestring, ensuring that process ownership will
be terminated, so the temp directory can be deleted.
+ Use `/` as path separators in a few places, even on Windows.
This seems to be necessary for texlive (#1151, thanks to Tim Lin).
+ Use `;` for `TEXINPUTS` separator on Windows (#1151).
+ Changes to error reporting, to handle non-UTF8 error output.

* `Text.Pandoc.Templates`:

+ Removed unneeded datatype context (Merijn Verstraaten).

+ YAML objects resolve to "true" in conditionals (#1133).
Note: If `address` is a YAML object and you just have `$address$`
in your template, the word `true` will appear, which may be
unexpected. (Previously nothing would appear.)

* `Text.Pandoc.SelfContained`:

+ `mkSelfContained` now takes just two arguments, `WriterOptions` and
the string.
* It no longer looks in data files. This only made sense when we
had copies of slidy and S5 code there.
* `fetchItem'` is used instead of the nearly duplicate `getItem`.
+ Handle `poster` attribute in `video` tags (#1188).

* `Text.Pandoc.Parsing`:

+ Made `F` an instance of Applicative (#1138).
+ Added `stateCaption`.
+ Added `HasMacros`, simplified other typeclasses.
Removed `updateHeaderMap`, `setHeaderMap`, `getHeaderMap`,
`updateIdentifierList`, `setIdentifierList`, `getIdentifierList`.
+ Changed the smart punctuation parser to return `Inlines`
rather than `Inline` (Matthew Pickering).
+ Changed `HasReaderOptions`, `HasHeaderMap`, `HasIdentifierList`
from typeclasses of monads to typeclasses of states. This simplifies
the instance definitions and provides more flexibility. Generalized
type of `getOption` and added a default definition. Removed
`askReaderOption`. Added `extractReaderOption`. Added
`extractHeaderMap` and `updateHeaderMap` in `HasHeaderMap`.
Gave default definitions for `getHeaderMap`, `putHeaderMap`,
`modifyHeaderMap`. Added `extractIdentifierList` and
`updateIdentifierList` in `HasIdentifierList`. Gave defaults
for `getIdentifierList`, `putIdentifierList`, and
`modifyIdentifierList`. The ultimate goal here is to allow different
parsers to use their own, tailored parser states (instead of
`ParserState`) while still using shared functions.

* Template changes:

+ LaTeX template: Use `fontenc` package only with `pdflatex` (#1164).
+ LaTeX template: Add `linestretch` and `fontfamily` variables.
+ LaTeX template: Conditionalize author and date commands.
+ Beamer template: Consistent styles for figure and table captions
+ LaTeX and beamer template: Adjust widths correctly for oversized
images. Use `\setkeys{Gin}{}` to set appropriate defaults for
`\includegraphics` (Yihui Xie, Garrick Aden-Buie). Load
`upquote` only after `fontenc` (Yihui Xie).
+ Beamer template: Added caption package (#1200).
+ Beamer template: changes for better unicode handling (KarolS).
+ DocBook template: use `authorgroup` if there are authors.
+ revealjs template: Move `include-after` to end (certainlyakey).
+ revealjs template: Fixed PDF print function (#1220, kevinkenan).

* Bumped version bounds of dependencies.

* Added a `--trace` command line option, for debugging backtracking
bugs. So far this only works with the markdown reader.

* MathMLinHTML: Fixed deprecation warning (#362, gwern, Albert Krewinkel).

* Updated travis script to test with multiple GHC versions.

* Force failure of a Travis build if GHC produces warnings (Albert

* Add `.editorconfig` (Albert Krewinkel).
See <> for details.

* Give more useful error message if '-t pdf' is specified (#1155).

* Added `Cite`, `SmallCaps` to `Arbitrary` instance (#1269).

* Allow `html4` as a synonym of `html` as a reader (it already works
as a writer).


+ Added an explanation of how to use YAML metadata to
force items to appear in the bibliography without citations in
the text (like LaTeX `\nocite`).
+ Added note to `--bibtex/--natbib`: not for use in making PDF
(#1194, thanks to nahoj).
+ Added explanatory notes about `--natbib` and `--biblatex`.
+ Added specification of legal syntax for citation keys.
+ Fixed variable defaults documentation (Albert Krewinkel).

* Removed copyright statements for files that have been removed
(Albert Krewinkel).

* Moved some doc files from `data-files` to `extra-source-files` (#1123).
They aren't needed at runtime. We keep README and COPYRIGHT in data
to ensure that they'll be available on all systems on which pandoc
is installed.

* Use cabal sandboxes in Windows build script.

pandoc (

* To changes to source; recompiled tarball with latest alex and
happy, so they will work with GHC 7.8.

pandoc (

* Bumped version bounds for blaze-html, blaze-markup.

* ImageSize: Avoid use of lookAhead, which is not in binary >= 0.6

* Fixed mediawiki ordered list parsing (#1122).

* HTML reader: Fixed bug reading inline math with `$$` (#225).

* Added support for LaTeX style literate Haskell code blocks in rST
(Merijn Verstraaten).

pandoc (

* Relaxed version constraint on binary, allowing the use of binary 0.5.

pandoc (1.12.3)

* The `--bibliography` option now sets the `biblio-files` variable.
So, if you're using `--natbib` or `--biblatex`, you can just use
`--bibliography=foo.bib` instead of `-V bibliofiles=foo`.

* Don't run pandoc-citeproc filter if `--bibliography` is
used together with `--natbib` or `--biblatex` (Florian Eitel).

* Template changes:

+ Updated beamer template to include booktabs.
+ Added `abstract` variable to LaTeX template.
+ Put `header-includes` after `title` in LaTeX template (#908).
+ Allow use of `\includegraphics[size]` in beamer.
This just required porting a macro definition from the default
LaTeX template to the default beamer template.

* `reference.docx`: Include `FootnoteText` style.
Otherwise Word ignores the style, even when specified in the `pPr`.

* `reference.odt`: Tidied `styles.xml`.

* Relaxed version bounds for dependencies.

* Added `withSocketsDo` around http conduit code in `openURL`,
so it works on Windows (#1080).

* Added `Cite` function to `sample.lua`.

* Markdown reader:

+ Fixed regression in title blocks (#1089).
If author field was empty, date was being ignored.
+ Allow backslash-newline hard line breaks in grid and
multiline table cells.
+ Citation keys may now start with underscores, and may contain
underscores adjacent to internal punctuation.

* LaTeX reader:

+ Add support for `Verb` macro (jrnold) (#1090).
+ Support babel-style quoting: `` "`..."' ``.

* Properly handle script blocks in strict mode. (That is,
`markdown-markdown_in_html_blocks`.) Previously a spurious
`<p>` tag was being added (#1093).

* Docbook reader: Avoid failure if `tbody` contains no `tr` or `row`

* LaTeX writer:

+ Factored out function for table cell creation.
+ Better treatment of footnotes in tables.
Notes now appear in the regular sequence, rather than in the
table cell. (This was a regression in 1.10.)

* HTML reader: Parse name/content pairs from meta tags as metadata.
Closes #1106.

* Moved `fixDisplayMath` from Docx writer to `Writer.Shared`.

* OpenDocument writer: Fixed `RawInline`, `RawBlock` so they don't escape.

* ODT writer: Use mathml for proper rendering of formulas.
Note: LibreOffice's support for this seems a bit buggy. But
it should be better than what we had before.

* RST writer: Ensure no blank line after def in definition list (#992).

* Markdown writer: Don't use tilde code blocks with braced attributes in
`markdown_github` output. A consequence of this change is that the
backtick form will be preferred in general if both are enabled. That
is good, as it is much more widespread than the tilde form. (#1084)

* Docx writer: Fixed problem with some modified reference docx files.
Include `word/_rels/settings.xml.rels` if it exists, as well as other
`rels` files besides the ones pandoc generates explicitly.

* HTML writer:

+ With `--toc`, headers no longer link to themselves (#1081).
+ Omit footnotes from TOC entries. Otherwise we get doubled
footnotes when headers have notes!

* EPUB writer:

+ Avoid duplicate notes when headings contain notes.
This arose because the headings are copied into the metadata
"title" field, and the note gets rendered twice. We strip the
note now before putting the heading in "title".
+ Strip out footnotes from toc entries.
+ Fixed bug with `--epub-stylesheet`. Now the contents of
`writerEpubStylesheet` (set by `--epub-stylesheet`)
should again work, and take precedence over a stylesheet specified
in the metadata.

* `Text.Pandoc.MIME`: Added `wmf`, `emf`.

* `Text.Pandoc.Shared`: `fetchItem` now handles image URLs beginning
with `//`.

* `Text.Pandoc.ImageSize`: Parse EXIF format JPEGs. Previously
we could only get size information for JFIF format, which led
to squished images in Word documents. Closes #976.

* Removed old `MarkdownTest_1.0.3` directory (#1104).

pandoc (

* Markdown reader: Fixed regression in list parser, involving
continuation lines containing raw HTML (or even verbatim raw HTML).

pandoc (1.12.2)

* Metadata may now be included in YAML blocks in a markdown document.
For example,

- type: main
text: My Book
- type: subtitle
text: An investigation of metadata
- role: author
text: John Smith
- role: editor
text: Sarah Jones
- scheme: DOI
text: doi:10.234234.234/33
publisher: My Press
rights: (c) 2007 John Smith, CC BY-NC
cover-image: img/mypic.jpg

Metadata may still be provided using `--epub-metadata`; it will
be merged with the metadata in YAML blocks.

* EPUB writer:

+ `meta` tags are now used instead of `opf` attributes for EPUB3.
+ Insert "svg" property as needed in opf (EPUB 3).
+ Simplify `imageTypeOf` using `getMimeType`.
+ Add properties attribute to `cover-image` item for EPUB 3.
+ Don't include node for `cover.xhtml` if no cover!
+ Ensure that same identifier is used throughout (#1044).
If an identifier is given in metadata, we use that; otherwise
we generate a random uuid.
+ Add cover reference to guide element (EPUB 2) (Shaun Attfield).
Fixes an issue with Calibre putting the cover at the end of the book
if the spine has `linear="no"`. Apparently this is best practice
for other converters as well:
+ Allow `stylesheet` in metadata. The value is a path to the stylesheet.
+ Allow partial dates: `YYYY`, `YYYY-MM`.

* Markdown writer: Fix rendering of tight sublists (#1050).
Previously a spurious blank line was included after a tight sublist.

* ODT writer: Add `draw:name` attribute to `draw:frame` elements (#1069).
This is reported to be necessary to avoid an error from recent
versions of Libre Office when files contain more than one image
Thanks to wmanley for reporting and diagnosing the problem.

* ConTeXt writer: Don't hardcode figure/table placement and numbering.
Instead, let this be set in the template, using `\setupfloat`.
Thanks to on4aa and Aditya Mahajan for the suggestion (#1067).

* Implemented CSL flipflopping spans in DOCX, LaTeX, and HTML writers.

* Fixed bug with markdown intraword emphasis. Closes #1066.

* Docbook writer: Hierarchicalize block content in metadata.
Previously headers just disappeared from block-level metadata
when it was used in templates. Now we apply the 'hierarchicalize'
transformation. Note that a block headed by a level-2 header will
turn into a `<sect1>` element.

* OpenDocument writer: Skip raw HTML (#1035).
Previously it was erroneously included as verbatim text.

* HTML/EPUB writer, footnotes: Put `<sup>` tag inside `<a>` tags.
This allows better control of formatting, since the `<a>`
tags have a distinguishing class (#1049).

* Docx writer:

+ Use mime type info returned by fetchItem.
+ Fixed core metadata (#1046).
Don't create empty date nodes if no date given.
Don't create multiple `dc:creator` nodes; instead separate by
+ Fix URL for core-properties in `_rels/.rels` (#1046).

* Plain writer: don't print `<span>` tags.

* LaTeX writer:

+ Fix definition lists with internal links in terms (#1032).
This fix puts braces around a term that contains an internal
link, to avoid problems with square brackets.
+ Properly escape pdftitle, pdfauthor (#1059).
+ Use booktabs package for tables (thanks to Jose Luis Duran).

* Updated beamer template. Now references should work properly
(in a slide) when `--biblatex` or `--natbib` is used.

* LaTeX reader:

+ Parse contents of curly quotes or matched `"` as quotes.
+ Support `\textnormal` as span with class `nodecor`.
This is needed for pandoc-citeproc.
+ Improved citation parsing. This fixes a run-time error that occurred
with `\citet{}` (empty list of keys). It also ensures that empty keys
don't get produced.

* MediaWiki reader: Add automatic header identifiers.

* HTML reader:

+ Use pandoc `Div` and `Span` for raw `<div>`, `<span>` when
+ Recognize `svg` tags as block level content (thanks to MinRK).
+ Parse LaTeX math if appropriate options are set.

* Markdown reader:

+ Yaml block must start immediately after `---`. If there's a blank
line after `---`, we interpreted it as a horizontal rule.
+ Correctly handle empty bullet list items.
+ Stop parsing "list lines" when we hit a block tag.
This fixes exponential slowdown in certain input, e.g.
a series of lists followed by `</div>`.

* Slides: Preserve `<div class="references">` in references slide.

* `Text.Pandoc.Writer.Shared`:

+ Fixed bug in `tagWithAttrs`. A space was omitted before key-value
attributes, leading to invalid HTML.
+ `normalizeDate`: Allow dates with year only (thanks to Shaun Attfield).
+ Fixed bug in `openURL` with `data:` URIs. Previously the base-64
encoded bytestring was returned. We now decode it so it's a proper

* DocBook reader: Handle numerical attributes starting with decimal.
Also use `safeRead` instead of `read`.

* `Text.Pandoc.Readers.TexMath`: Export `readTeXMath'`, which attends
to display/inline. Deprecate `readTeXMath`, and use `readTeXMath'`
in all the writers. Require `texmath >=`.

* `Text.Pandoc.MIME`:

+ Add entry for `jfif`.
+ In looking up extensions, drop the encoding info.
E.g. for 'image/jpg;base64' we should lookup 'image/jpg'.

* Templates: Changed how array variables are resolved. Previously if
`foo` is an array (which might be because multiple values were set on
the command line), `$foo$` would resolve to the concatenation of the
elements of foo. This is rarely useful behavior. It has been changed
so that the first value is rendered. Of course, you can still iterate
over the values using `$for(foo)$`. This has the result that you can
override earlier settings using `-V` by putting new values later on the
command line, which is useful for many purposes.

* `Text.Pandoc`: Don't default to `pandocExtensions` for all writers.

* Allow "epub2" as synonym for "epub", "html4" for "html".

* Don't look for slidy files in data files with `--self-contained`.

* Allow `https:` command line arguments to be downloaded.

* Fixed `` so data files embedded in `pandoc-citeproc`.

pandoc (1.12.1)

* `Text.Pandoc.Definition`: Changed default JSON serialization format.
Instead of `{"Str": "foo"}`, for example, we now have `{"t": "Str",
"c": "foo"}`. This new format is easier to work with outside of Haskell.
Incidentally, "t" stands for "tag", "c" for "contents".

* MediaWiki reader: Trim contents of `<math>` tags, to avoid problems
when converting to markdown (#1027).

* LaTeX reader:

+ Ensure that preamble doesn't contribute to the text of
the document.
+ Fixed character escaping in \url{}. Previously `\~` wasn't handled
properly, among others.
+ Parse `{groups}` as `Span`. This is needed for accurate conversion of
bibtex titles, since we need to know what was protected from
titlecase conversions.

* LaTeX writer:

+ Specially escape non-ascii characters in labels.
Otherwise we can get compile errors and other bugs when
compiled with pdflatex (#1007). Thanks to begemotv2718 for the fix.
+ Add link anchors for code blocks with identifiers (#1025).

* Throughout the code, use `isURI` instead of `isAbsoluteURI`.
It allows fragments identifiers.

* Slide formats:

+ A Div element with class "notes" is treated as speaker
notes. Currently beamer goes to `\note{}`, revealjs to
`<aside class="notes">`, and the notes are simply suppressed in
other formats (#925).
+ Fixed `. . .` (pause) on HTML slide formats. Closes #1029.
The old version caused a pause to be inserted before the first
material on a slide. This has been fixed.
+ Removed data files for s5, slideous, slidy.
Users of s5 and slideous will have to download the needed
files, as has been documented for some time in the README.
By default, slidy code will be sought on the web, as before.

* HTML writer: Insert command to typeset mathjax only in slideous output
(#966, #1012).

* RST writer: Skip spaces after display math. Otherwise we get indentation
problems, and part of the next paragraph may be rendered as part of the

* OpenDocument writer: Fix formatting of strikeout code (#995),
thanks to wilx. don't use `font-face-decls` variable.

* Fixed test suite so it works with cabal sandboxes.

pandoc (

* Removed `stringable` dependency.

pandoc (

* Allow `--metadata` to be repeated for the same key to form a list.
This also has the effect that `--bibliography` can be repeated,
as before.

* Handle boolean values in `--metadata`. Note that anything not parseable
as a YAML boolean or string is treated as a literal string.
You can get a string value with "yes", or any of the strings interpretable
as booleans, by quoting it:

-M boolvalue=yes -M stringvalue='"yes"'

* LaTeX writer: Don't print references if `--natbib` or `--biblatex`
option used.

* DOCX writer: Add `settings.xml` to the zip container. Fixes a bug
in which docx files could not be read by some versions of Word
and LibreOffice (#990).

* Fixed a regression involving slide shows with bibliographies.
The Div container around references messed up the procedure for carving
a document into slides. So we now remove the surrounding Div in

* More informative error message when a filter is not found in path.

* Depend on pandoc-types 1.12.1. This provide `ToJSONFilter`
instances for `Data a => a -> [a]` and `Data a => a -> IO [a]`.

* Don't use unicode_collation in building OSX package:
it adds something like 50MB of dependencies to the package.

* Declare alex and happy as build-tools (#986).

pandoc (1.12)

[new features]

* Much more flexible metadata, including arbitrary fields and structured
values. Metadata can be specified flexibly in pandoc markdown using
YAML metadata blocks, which may occur anywhere in the document:

title: Here is my title.
abstract: |
This is the abstract.

1. It can contain
2. block content
and *inline markup*

tags: [cat, dog, animal]

Metadata fields automatically populate template variables.

* Added `opml` (OPML) as input and output format. The `_note` attribute,
used in OmniOutliner and supported by multimarkdown, is supported.
We treat the contents as markdown blocks under a section header.

* Added `haddock` (Haddock markup) as input format (David Lazar).

* Added `revealjs` output format, for reveal.js HTML 5 slide shows.
(Thanks to Jamie F. Olson for the initial patch.)
Nested vertical stacks are used for hierarchical structure.
Results for more than one level of nesting may be odd.

* Custom writers can now be written in lua.

pandoc -t data/sample.lua

will load the script sample.lua and use it as a custom writer.
(For a sample, do `pandoc --print-default-data-file sample.lua`.)
Note that pandoc embeds a lua interpreter, so lua need not be
installed separately.

* New `--filter/-F` option to make it easier to run "filters"
(Pandoc AST transformations that operate on JSON serializations).
Filters are always passed the name of the output format, so their
behavior can be tailored to it. The repository
<> contains
a python module for writing pandoc filters in python, with
a number of examples.

* Added `--metadata/-M` option.
This is like `--variable/-V`, but actually adds to metadata, not
just variables.

* Added `--print-default-data-file` option, which allows printing
of any of pandoc's data files. (For example,
`pandoc --print-default-data-file reference.odt` will print

* Added syntax for "pauses" in slide shows:

This gives

. . .

me pause.

* New markdown extensions:

+ `ignore_line_breaks`: causes intra-paragraph line breaks to be ignored,
rather than being treated as hard line breaks or spaces. This is useful
for some East Asian languages, where spaces aren't used between words,
but text is separated into lines for readability.
+ `yaml_metadata_block`: Parse YAML metadata blocks. (Default.)
+ `ascii_identifiers`: This will force `auto_identifiers` to use ASCII
only. (Default for `markdown_github`.) (#807)
+ `lists_without_preceding_blankline`: Allow lists to start without
preceding blank space. (Default for `markdown_github`.) (#972)

[behavior changes]

* `--toc-level` no longer implies `--toc`.
Reason: EPUB users who don't want a visible TOC may still want
to set the TOC level for in the book navigation.

* `--help` now prints in and out formats in alphabetical order, and
says something about PDF output (#720).

* `--self-contained` now returns less verbose output (telling you
which URLs it is fetching, but not giving the full header). In
addition, there are better error messages when fetching a URL fails.

* Citation support is no longer baked in to core pandoc. Users who
need citations will need to install and use a separate filter
(`--filter pandoc-citeproc`). This filter will take `bibliography`,
`csl`, and `citation-abbreviations` from the metadata, though it
may still be specified on the command line as before.

* A `Cite` element is now created in parsing markdown whether or not
there is a matching reference.

* The `pandoc-citeproc` script will put the bibliography at the
end of the document, as before. However, it will be put inside a `Div`
element with class "references", allowing users some control
over the styling of references. A final header, if any, will
be included in the `Div`.

* The markdown writer will not print a bibliography if the
`citations` extension is enabled. (If the citations are formatted
as markdown citations, it is redundant to have a bibliography,
since one will be generated automatically.)

* Previously we used to store the directory of the first input file,
even if it was local, and used this as a base directory for finding
images in ODT, EPUB, Docx, and PDF. This has been confusing to many
users. So we now look for images relative to the current
working directory, even if the first file argument is in another
directory. Note that this change may break some existing workflows.
If you have been assuming that relative links will be interpreted
relative to the directory of the first file argument, you'll need
to make that the current directory before running pandoc. (#942)

* Better error reporting in some readers, due to changes in `readWith`:
the line in which the error occurred is printed, with a caret pointing
to the column.

* All slide formats now support incremental slide view for definition lists.

* Parse `\(..\)` and `\[..\]` as math in MediaWiki reader.
Parse `:<math>...</math>` as display math. These notations are used with
the MathJax MediaWiki extension.

* All writers: template variables are set automatically from metadata
fields. However, variables specified on the command line with
`--variable` will completely shadow metadata fields.

* If `--variable` is used to set many variables with the same name,
a list is created.

* Man writer: The `title`, `section`, `header`, and `footer` can now
all be set individually in metadata. The `description` variable has been
removed. Quotes have been added so that spaces are allowed in the
title. If you have a title that begins

COMMAND(1) footer here | header here

pandoc will still parse it into a title, section, header, and
footer. But you can also specify these elements explicitly (#885).

* Markdown reader

+ Added support for YAML metadata blocks, which can come anywhere
in the document (not just at the beginning). A document can contain
multiple YAML metadata blocks.
+ HTML span and div tags are parsed as pandoc Span and Div elements.

* Markdown writer

+ Allow simple tables to be printed as grid tables,
if other table options are disabled. This means you can do
`pandoc -t markdown-pipe_tables-simple_tables-multiline_tables`
and all tables will render as grid tables.
+ Support YAML title block (render fields in alphabetical order
to make output predictable).

[API changes]

* `Meta` in `Text.Pandoc.Definition` has been changed to allow
structured metadata. (Note: existing code that pattern-matches
on `Meta` will have to be revised.) Metadata can now contain
indefinitely many fields, with content that can be a string,
a Boolean, a list of `Inline` elements, a list of `Block`
elements, or a map or list of these.

* A new generic block container (`Div`) has been added to `Block`,
and a generic inline container (`Span`) has been added to `Inline`.
These can take attributes. They will render in HTML, Textile,
MediaWiki, Org, RST and and Markdown (with `markdown_in_html`
extension) as HTML `<div>` and `<span>` elements; in other formats
they will simply pass through their contents. But they can be
targeted by scripts.

* `Format` is now a newtype, not an alias for String.
Equality comparisons are case-insensitive.

* Added `Text.Pandoc.Walk`, which exports hand-written tree-walking
functions that are much faster than the SYB functions from
`Text.Pandoc.Generic`. These functions are now used where possible
in pandoc's code. (`Tests.Walk` verifies that `walk` and `query`
match the generic traversals `bottomUp` and `queryWith`.)

* Added `Text.Pandoc.JSON`, which provides `ToJSON` and `FromJSON`
instances for the basic pandoc types. They use GHC generics and
should be faster than the old JSON serialization using

* Added `Text.Pandoc.Process`, exporting `pipeProcess`.
This is a souped-up version of `readProcessWithErrorcode` that
uses lazy bytestrings instead of strings and allows setting
environment variables. (Used in `Text.Pandoc.PDF`.)

* New module `Text.Pandoc.Readers.OPML`.

* New module `Text.Pandoc.Writers.OPML`.

* New module `Text.Pandoc.Readers.Haddock` (David Lazar).
This is based on Haddock's own lexer/parser.

* New module `Text.Pandoc.Writers.Custom`.

* In `Text.Pandoc.Shared`, `openURL` and `fetchItem` now return an
Either, for better error handling.

* Made `stringify` polymorphic in `Text.Pandoc.Shared`.

* Removed `stripTags` from `Text.Pandoc.XML`.

* `Text.Pandoc.Templates`:

+ Simplified `Template` type to a newtype.
+ Removed `Empty`.
+ Changed type of `renderTemplate`: it now takes a JSON context
and a compiled template.
+ Export `compileTemplate`.
+ Export `renderTemplate'` that takes a string instead of a compiled
+ Export `varListToJSON`.

* `Text.Pandoc.PDF` exports `makePDF` instead of `tex2pdf`.
The signature of `makePDF` has changed and now contains
an additional argument for pdf engine options.

* `Text.Pandoc`:

+ Made `toJsonFilter` an alias for `toJSONFilter` from `Text.Pandoc.JSON`.
+ Removed `ToJsonFilter` typeclass. `ToJSONFilter` from
`Text.Pandoc.JSON` should be used instead. (Compiling against
pandoc-types instead of pandoc will also produce smaller executables.)
* Removed the deprecated `jsonFilter` function.
+ Added `readJSON`, `writeJSON` to the API (#817).

* `Text.Pandoc.Options`:

+ Added `Ext_lists_without_preceding_blankline`,
`Ext_ascii_identifiers`, `Ext_ignore_line_breaks`,
`Ext_yaml_metadataBlock` to `Extension`.
+ Changed `writerSourceDirectory` to `writerSourceURL` and changed the
type to a `Maybe`. `writerSourceURL` is set to 'Just url' when the
first command-line argument is an absolute URL. (So, relative links
will be resolved in relation to the first page.) Otherwise, 'Nothing'.
+ All bibliography-related fields have been removed from
`ReaderOptions` and `WriterOptions`: `writerBiblioFiles`,
`readerReferences`, `readerCitationStyle`.
+ Removed `writerPdfArgs` from `WriterOptions`. These are now
passed in as a parameter to `makePDF`.

* The `Text.Pandoc.Biblio` module has been removed. Users of the
pandoc library who want citation support will need to use
`Text.CSL.Pandoc` from `pandoc-citeproc`.

[bug fixes]

* In markdown, don't autolink a bare URI that is followed by `</a>`

* `Text.Pandoc.Shared`

+ `openURL` now follows redirects (#701), properly handles `data:`
URIs, and prints diagnostic output to stderr rather than stdout.
+ `readDefaultDataFile`: normalize the paths. This fixes bugs in
`--self-contained` on pandoc compiled with `embed_data_files` (#833).
+ Fixed `readDefaultDataFile` so it works on Windows.
+ Better error messages for `readDefaultDataFile`. Instead of
listing the last path tried, which can confuse people who are
using `--self-contained`, so now we just list the data file name.
+ URL-escape pipe characters. Even though these are legal, `Network.URI`
doesn't regard them as legal in URLs. So we escape them first (#535).

* Mathjax in HTML slide shows: include explicit "Typeset" call.
This seems to be needed for some formats (e.g. slideous) and won't
hurt in others (#966).

* `Text.Pandoc.PDF`

+ On Windows, create temdir in working directory, since the system
temp directory path may contain tildes, which can cause
problems in LaTeX (#777).
+ Put temporary output directory in `TEXINPUTS` (see #917).
+ `makePDF` tries to download images that are not found locally,
if the first argument is a URL (#917).
+ If compiling with `pdflatex` yields an encoding error, offer
the suggestion to use `--latex-engine=xelatex`.

* Produce automatic header identifiers in parsing textile, RST,
and LaTeX, unless `auto_identifiers` extension is disabled (#967).

* `Text.Pandoc.SelfContained`: Strip off fragment, query of relative URL
before treating as a filename. This fixes `--self-contained` when used
with CSS files that include web fonts using the method described here:
(#739). Handle `src` in `embed`, `audio`, `source`, `input` tags.

* `Text.Pandoc.Parsing`: `uri` parser no longer treats punctuation before
percent-encoding, or a `+` character, as final punctuation.

* `Text.Pandoc.ImageSize`: Handle EPS (#903). This change will make
EPS images properly sized on conversion to Word.

* Slidy: Use slidy.js rather than slidy.js.gz.
Reason: some browsers have trouble with the gzipped js file,
at least on the local file system (#795).

* Markdown reader

+ Properly handle blank line at beginning of input (#882).
+ Fixed bug in unmatched reference links. The input
`[*infile*] [*outfile*]` was getting improperly parsed:
"infile" was emphasized, but "*outfile*" was literal (#883).
+ Allow internal `+` in citation identifiers (#856).
+ Allow `.` or `)` after `#` in ATX headers if no `fancy_lists`.
+ Do not generate blank title, author, or date metadata elements.
Leave these out entirely if they aren't present.
+ Allow backtick code blocks not to be preceded by blank line (#975).

* Textile reader:

+ Correctly handle entities.
+ Improved handling of `<pre>` blocks (#927). Remove internal HTML tags
in code blocks, rather than printing them verbatim. Parse attributes
on `<pre>` tag for code blocks.

* HTML reader: Handle non-simple tables (#893). Column widths are read from
`col` tags if present, otherwise divided equally.

* LaTeX reader

+ Support alltt environment (#892).
+ Support `\textasciitilde`, `\textasciicircum` (#810).
+ Treat `\textsl` as emphasized text reader (#850).
+ Skip positional options after `\begin{figure}`.
+ Support `\v{}` for hacek (#926).
+ Don't add spurious ", " to citation suffixes.
This is added when needed in pandoc-citeproc.
+ Allow spaces in alignment spec in tables, e.g. `{ l r c }`.
+ Improved support for accented characters (thanks to Scott Morrison).
+ Parse label after section command and set id (#951).

* RST reader:

+ Don't insert paragraphs where docutils doesn't.
`rst2html` doesn't add `<p>` tags to list items (even when they are
separated by blank lines) unless there are multiple paragraphs in the
list. This commit changes the RST reader to conform more closely to
what docutils does (#880).
+ Improved metadata. Treat initial field list as metadata when
standalone specified. Previously ALL fields "title", "author",
"date" in field lists were treated as metadata, even if not at
the beginning. Use `subtitle` metadata field for subtitle.
+ Fixed 'authors' metadata parsing in reST. Semicolons separate
different authors.

* MediaWiki reader

+ Allow space before table rows.
+ Fixed regression for `<ref>URL</ref>`.
`<` is no longer allowed in URLs, according to the uri parser
in `Text.Pandoc.Parsing`. Added a test case.
+ Correctly handle indented preformatted text without preceding
or following blank line.
+ Fixed `|` links inside table cells. Improved attribute parsing.
+ Skip attributes on table rows. Previously we just crashed if
rows had attributes, now we ignore them.
+ Ignore attributes on headers.
+ Allow `Image:` for images (#971).
+ Parse an image with caption in a paragraph by itself as a figure.

* LaTeX writer

+ Don't use ligatures in escaping inline code.
+ Fixed footnote numbers in LaTeX/PDF tables. This fixes a bug
wherein notes were numbered incorrectly in tables (#827).
+ Always create labels for sections. Previously the labels were only
created when there were links to the section in the document (#871).
+ Stop escaping `|` in LaTeX math.
This caused problems with array environments (#891).
+ Change `\` to `/` in paths. `/` works even on Windows in LaTeX.
`\` will cause major problems if unescaped.
+ Write id for code block to label attribute in LaTeX when listings
is used (thanks to Florian Eitel).
+ Scale LaTeX tables so they don't exceed columnwidth.
+ Avoid problem with footnotes in unnumbered headers (#940).

* Beamer writer: when creating beamer slides, add `allowframebreaks` option
to the slide if it is one of the header classes. It is recommended
that your bibliography slide have this attribute:

# References {.allowframebreaks}

This causes multiple slides to be created if necessary, depending
on the length of the bibliography.

* ConTeXt writer: Properly handle tables without captions. The old output
only worked in MkII. This should work in MkIV as well (#837).

* MediaWiki writer: Use native mediawiki tables instead of HTML (#720).

* HTML writer:

+ Fixed `--no-highlight` (Alexander Kondratskiy).
+ Don't convert to lowercase in email obfuscation (#839).
+ Ensure proper escaping in `<title>` and `<meta>` fields.

* AsciiDoc writer:

+ Support `--atx-headers` (Max Rydahl Andersen).
+ Don't print empty identifier blocks `([[]])` on headers (Max
Rydahl Andersen).

* ODT writer:

+ Fixing wrong numbered-list indentation in open document format
(Alexander Kondratskiy) (#369).
+ `reference.odt`: Added pandoc as "generator" in `meta.xml`.
+ Minor changes for ODF 1.2 conformance (#939). We leave the
nonconforming `contextual-spacing` attribute, which is provided by
LibreOffice itself and seems well supported.

* Docx writer:

+ Fixed rendering of display math in lists.
In 1.11 and 1.11.1, display math in lists rendered as a new list
item. Now it always appears centered, just as outside of lists,
and in proper display math style, no matter how far indented the
containing list item is (#784).
+ Use `w:br` with `w:type` `textWrapping` for linebreaks.
Previously we used `w:cr` (#873).
+ Use Compact style for Plain block elements, to
differentiate between tight and loose lists (#775).
+ Ignore most components of `reference.docx`.
We take the `word/styles.xml`, `docProps/app.xml`,
`word/theme/theme1.xml`, and `word/fontTable.xml` from
`reference.docx`, ignoring everything else. This should help
with the corruption problems caused when different versions of
Word resave the reference.docx and reorganize things.
+ Made `--no-highlight` work properly.

* EPUB writer

+ Don't add `dc:creator` tags if present in EPUB metadata.
+ Add `id="toc-title"` to `h1` in `nav.xhtml` (#799).
+ Don't put blank title page in reading sequence.
Set `linear="no"` if no title block. Addresses #797.
+ Download webtex images and include as data URLs.
This allows you to use `--webtex` in creating EPUBs.
Math with `--webtex` is automatically made self-contained.
+ In `data/epub.css`, removed highlighting styles (which
are no longer needed, since styles are added by the HTML
writer according to `--highlighting-style`). Simplified
margin fields.
+ If resource not found, skip it, as in Docx writer (#916).

* RTF writer:

+ Properly handle characters above the 0000-FFFF range.
Uses surrogate pairs. Thanks to Hiromi Ishii for the patch.
+ Fixed regression with RTF table of contents.
+ Only autolink absolute URIs. This fixes a regression, #830.

* Markdown writer:

+ Only autolink absolute URIs. This fixes a regression, #830.
+ Don't wrap attributes in fenced code blocks.
+ Write full metadata in MMD style title blocks.
+ Put multiple authors on separate lines in pandoc titleblock.
Also, don't wrap long author entries, as new lines get treated
as new authors.

* `Text.Pandoc.Templates`:

+ Fixed bug retrieving default template for markdown variants.
+ Templates can now contain "record lookups" in variables;
for example, `author.institution` will retrieve the `institution`
field of the `author` variable.
+ More consistent behavior of `$for$`. When `foo` is not a list,
`$for(foo)$...$endfor$` should behave like $if(foo)$...$endif$.
So if `foo` resolves to "", no output should be produced.
See pandoc-templates#39.

* Citation processing improvements (now part of pandoc-citeproc):

+ Fixed `endWithPunct` The new version correctly sees a sentence
ending in '.)' as ending with punctuation. This fixes a bug which
led such sentences to receive an extra period at the end: '.).'.
Thanks to Steve Petersen for reporting.
+ Don't interfere with Notes that aren't citation notes.
This fixes a bug in which notes not generated from citations were
being altered (e.g. first letter capitalized) (#898).
+ Only capitalize footnote citations when they have a prefix.
+ Changes in suffix parsing. A suffix beginning with a digit gets 'p'
inserted before it before passing to citeproc-hs, so that bare numbers
are treated as page numbers by default. A suffix not beginning with
punctuation has a space added at the beginning (rather than a comma and
space, as was done before for not-author-in-text citations).
The result is that `\citep[23]{item1}` in LaTeX will be interpreted
properly, with '23' treated as a locator of type 'page'.
+ Many improvements to citation rendering, due to fixes in citeproc-hs
(thanks to Andrea Rossato).
+ Warnings are issued for undefined citations, which are rendered
as `???`.
+ Fixed hanging behavior when locale files cannot be found.

[template changes]

* DocBook: Use DocBook 4.5 doctype.

* Org: '#+TITLE:' is inserted before the title.
Previously the writer did this.

* LaTeX: Changes to make mathfont work with xelatex.
We need the mathspec library, not just fontspec, for this.
We also need to set options for setmathfont (#734).

* LaTeX: Use `tex-ansi` mapping for `monofont`.
This ensures that straight quotes appear as straight, rather than
being treated as curly. See #889.

* Made `\includegraphics` more flexible in LaTeX template.
Now it can be used with options, if needed. Thanks to Bernhard Weichel.

* LaTeX/Beamer: Added `classoption` variable.
This is intended for class options like `oneside`; it may
be repeated with different options. (Thanks to Oliver Matthews.)

* Beamer: Added `fonttheme` variable. (Thanks to Luis Osa.)

* LaTeX: Added `biblio-style` variable (#920).

* DZSlides: title attribute on title section.

* HTML5: add meta tag to allow scaling by user (Erik Evenson)

[under-the-hood improvements]

* Markdown reader:Improved strong/emph parsing, using the strategy of
<>. The new parsing algorithm requires
no backtracking, and no keeping track of nesting levels. It will give
different results in some edge cases, but these should not affect normal

* Added `Text.Pandoc.Compat.Monoid`.
This allows pandoc to compile with `base` < 4.5, where `Data.Monoid`
doesn't export `<>`. Thanks to Dirk Ullirch for the patch.

* Added `Text.Pandoc.Compat.TagSoupEntity`.
This allows pandoc to compile with `tagsoup` 0.13.x.
Thanks to Dirk Ullrich for the patch.

* Most of `Text.Pandoc.Readers.TeXMath` has been moved to the
`texmath` module (0.6.4). (This allows `pandoc-citeproc` to
handle simple math in bibliography fields.)

* Added `Text.Pandoc.Writers.Shared` for shared functions used
only in writers. `metaToJSON` is used in writers to create a
JSON object for use in the templates from the pandoc metadata
and variables. `getField`, `setField`, and `defField` are
for working with JSON template contexts.

* Added `Text.Pandoc.Asciify` utility module.
This exports functions to create ASCII-only versions of identifiers.

* `Text.Pandoc.Parsing`

+ Generalized state type on `readWith` (API change).
+ Specialize readWith to `String` input. (API change).
+ In `ParserState`, replace `stateTitle`, `stateAuthors`, `stateDate`
with `stateMeta` and `stateMeta'`.

* `Text.Pandoc.UTF8`: use strict bytestrings in reading. The use of lazy
bytestrings seemed to cause problems using pandoc on 64-bit Windows
7/8 (#874).

* Factored out `registerHeader` from markdown reader, added to

* Removed `blaze_html_0_5` flag, require `blaze-html` >= 0.5.
Reason: < 0.5 does not provide a monoid instance for Attribute,
which is now needed by the HTML writer (#803).

* Added `http-conduit` flag, which allows fetching https resources.
It also brings in a large number of dependencies (`http-conduit`
and its dependencies), which is why for now it is an optional flag

* Added

* Improved INSTALL instructions.

* `make-windows-installer.bat`: Removed explicit paths for executables.

* `aeson` is now used instead of `json` for JSON.

* Set default stack size to 16M. This is needed for some large
conversions, esp. if pandoc is compiled with 64-bit ghc.

* Various small documentation improvements.
Thanks to achalddave and drothlis for patches.

* Removed comment that chokes recent versions of CPP (#933).

* Removed support for GHC version < 7.2, since pandoc-types now
requires at least GHC 7.2 for GHC generics.

pandoc (1.11.1)

* Markdown reader:

+ Fixed regression in which parentheses were lost in link URLs.
Added tests. Closes #786.
+ Better handling of unmatched double quotes in `--smart` mode.
These occur frequently in fiction, since it is customary not to
close quotes in dialogue if the speaker does not change between
paragraphs. The unmatched quotes now get turned into literal
left double quotes. (No `Quoted` inline is generated, however.)
Closes #99 (again).

* HTML writer: Fixed numbering mismatch between TOC and sections.
`--number-offset` now affects TOC numbering as well
as section numbering, as it should have all along. Closes #789.

* Markdown writer: Reverted 1.11 change that caused citations to be rendered
as markdown citations, even if `--bibliography` was specified, unless
`citation` extension is disabled. Now, formatted citations are always
printed if `--bibliography` was specified. If you want to reformat
markdown keeping pandoc markdown citations intact, don't use
`--bibliography`. Note that citations parsed from LaTeX documents will
be rendered as pandoc markdown citations when `--bibliography` is not

* ODT writer: Fixed regression leading to corrupt ODTs.
This was due to a change in the `Show` instance for
`Text.Pandoc.Pretty.Doc`. Closes #780.

* Fixed spacing bugs involving code block attributes in
RST reader and Markdown writer. Closes #763.

* Windows package: Various improvements due to Fyodor Sheremetyev.

+ Automatically set installation path (Program Files or Local App Data).
+ Set system PATH environment variable when installing for all users.
+ Pandoc can installed for all users using the following command.
`msiexec /i pandoc-1.11.msi ALLUSERS=1`.

* Bumped QuickCheck version bound.

pandoc (1.11)

* Added `--number-offset` option. (See README for description.)

* Added `--default-image-extension` option. (See README for description.)

* `--number-sections` behavior change: headers with class `unnumbered`
will not be numbered.

* `--version` now reports the default data directory.

* `Text.Pandoc.Parsing` is no longer exposed. (API change.)

* `Text.Pandoc.Highlighting` is no longer exposed. (API change.)

* `Text.Pandoc.Shared`: Changed type of `Element`. `Sec` now includes
a field for `Attr` rather than just `String`. (API change.)

* Added `markdown_github` as input format. This was an accidental
omission in 1.10.

* Added `readerDefaultImageExtension` field to `ReaderOptions`. (API

* Added `writerNumberOffset` field in `WriterOptions`. (API change.)

* Beamer template:

+ Fixed captions with longtable. Thanks to Joost Kremers.
+ Provide `\Oldincludegraphics` as in LaTeX template (Benjamin Bannier).

* LaTeX template:

+ Load microtype after fonts. Microtype needs to know
what fonts are being used. Thanks to dfc for the patch.
+ Set `secnumdepth` to 5 if `--number-sections` specified.
This yields behavior equivalent to the other writers, numbering
level 4 and 5 headers too. Closes #753.

* HTML reader:

+ Handle `<colgroup>` tag.
+ Preserve all header attributes.

* LaTeX reader:

+ Parse `\hrule` as `HorizontalRule`. Closes #746.
+ Parse starred variants of `\section` etc. as headers with
attribute `unnumbered`.
+ Read optional attributes in `lstlisting` and `Verbatim` environments.
We convert these to pandoc standard names, e.g. `numberLines`
for `numbers=left`, `startFrom=100` from `firstnumber=100`.
+ Handle language attribute for lstlistings.
+ Better support for Verbatim and minted environments. Closes #763.

* Markdown reader:

+ `-` in an attribute context = `.unnumbered`. The point of this
is to provide a way to specify unnumbered headers in non-English
+ Fixed bug parsing key/value attributes. Parsing failed if you
had an unquoted attribute immediately before the final '}'.
+ Make backslash escape work in attributes.
+ Fix title block parsing. Now if `mmd_title_blocks` is specified,
pandoc will parse a MMD title block if it sees one, even if
`pandoc_title_blocks` is enabled.
+ Refactoring: `litChar` now includes entities, so we don't need
to use `fromEntities` e.g. on titles.
+ Allow spaces around borders in pipe tables. Closes #772.
+ Allow all punctuation in angle-bracket autolinks. Previously
things like `----` were disallowed, because the uri parser
treated them as trailing punctuation. Closes #768.
+ Make `implicit_header_references` work properly when
headers are given explicit identifiers.
+ Check for tables before line blocks. Otherwise some pipe
tables get treated as line blocks.
+ Allow `&` in emails (for entities).
+ Properly handle entities in titles and links. A markdown link
`<http://g&ouml;>` should be a link to `http://gö`.
Closes #723.

* Textile reader:

+ Handle attributes on headers.

* LaTeX reader:

+ Add `fig:` as title for images with captions.
This is needed for them to be rendered as figures. Closes #766.
+ Never emit an empty paragraph. See #761.
+ Handle `\caption` for images in figures. Closes #766.
+ Parse `\section*`, etc. as unnumbered sections.

* HTML writer:

+ Support header attributes. The attributes go on
the enclosing `section` or `div` tag if `--section-divs` is specified.
+ Fixed a regression (only now noticed) in html+lhs output.
Previously the bird tracks were being omitted.

* LaTeX writer:

+ Omit lists with no items to avoid LaTeX errors.
+ Support line numbering with `--listings`.
If `numberLines` class is present, we add `numbers=left`;
if `startFrom` is present, we add `firstnumber=`. (#763)

* ConTeXt writer:

+ Removed `\placecontent`. This produced a duplicate toc,
in conjunction with `\placelist`.
+ Use `\title`, `\subject` etc. for headers with
`unnumbered` class.

* Textile writer:

+ Support header attributes.

* Markdown writer:

+ Use grid tables when needed, and if enabled. Closes #740.
+ Render citations as pandoc-markdown citations.
Previously citations were rendered as citeproc-formatted citations
by default. Now we render them as pandoc citations, e.g. `[@item1]`,
unless the `citations` extension is disabled.
If you still want formatted citations in your markdown output,
use `pandoc -t markdown-citations`.

* RST writer:

+ Support `:number-lines:` in code blocks.

* Docx writer:

+ Better treatment of display math. Display math inside a
paragraph is now put in a separate paragraph, so it will render
properly (centered and without extra blank lines around it).
Partially addresses #742.
+ Content types and document rels xml files are now created from
scratch, rather than being taken over from `reference.docx`.
This fixes problems that arise when you edit the `reference.docx`
with Word.
+ We also now encode mime types for each individual image rather
than using defaults. This should allow us to handle a wider
range of image types (including PDF). Closes #414.
+ Changed style names in `reference docx`.
`FootnoteReference` -> `FootnoteRef`, `Hyperlink` -> `Link`.
The old names got changed by Word when the `reference.docx` was
edited. Closes #414.

* EPUB writer:

+ Fix section numbering. Previously the numbering restarted from 1
in each chapter (with `--number-sections`), though the numbers in
the table of contents were correct.
+ Headers with "unnumbered" attribute are not numbered. (Nor do they
cause an increment in running numbering.) Section numbers now work
properly, even when there is material before the first numbered section.
+ Include HTML TOC, even in epub2. The TOC is included in `<spine>`,
but `linear` is set to `no` unless the `--toc` option is specified.
Include `<guide>` element in OPF. This should allow the TOC to
be usable in Kindles when converted with kindlegen. Closes #773.

* `Text.Pandoc.Parsing`: Optimized `oneOfStringsCI`.
This dramatically reduces the speed penalty that comes from enabling the
`autolink_bare_uris` extension. The penalty is still substantial (in one
test, from 0.33s to 0.44s), but nowhere near what it used to be.
The RST reader is also much faster now, as it autodetects URIs.

* `Text.Pandoc.Shared`: `hierarchicalize` will not number section
with class "unnumbered". Unnumbered sections get `[]` for their
section number.

* `Text.Pandoc.Pretty`:

+ Fixed `chomp` so it works inside `Prefixed` elements.
+ Changed `Show` instance so it is better for debugging.

* `Text.Pandoc.ImageSize`: Added `Pdf` to `ImageType`.

* `Text.Pandoc.UTF8`: Strip off BOM if present. Closes #743.

* Windows installer improvements:

+ The installer is now signed with a certificate (thanks to
Fyodor Sheremetyev).
+ WiX is used instead of InnoSetup. The installer is now a
standard msi file.
+ The version number is now auto-detected, and need not be
updated separately.

* OSX installer improvements:

+ The package and pandoc executable are now signed with a
certificate (thanks to Fyodor Sheremetyev).
+ RTF version of license is used.
+ Use full path for sysctl in `InstallationCheck` script (jonahbull).
Closes #580.

* Converted COPYING to markdown.

* pandoc.cabal: Require latest versions of highlighting-kate,
texmath, citeproc-hs, zip-archive.

pandoc (1.10.1)

* Markdown reader: various optimizations, leading to a
significant performance boost.

* RST reader: Allow anonymous form of inline links:
`` `hello <url>`__ `` Closes #724.

* Mediawiki reader: Don't require newlines after tables.
Thanks to jrunningen for the patch. Closes #733.

* Fixed LaTeX macro parsing. Now LaTeX macro definitions are preserved
when output is LaTeX, and applied when it is another format.
Partially addresses #730.

* Markdown and RST readers: Added parser to `block` that skips blank
lines. This fixes a subtle regression involving grid tables with
empty cells. Also added test for grid table with empty cells.
Closes #732.

* RST writer: Use `.. code:: language` for code blocks with language.
Closes #721.

* DocBook writer: Fixed output for hard line breaks, adding a newline
between `<literallayout>` tags.

* Markdown writer: Use an autolink when link text matches url.
Previously we also checked for a null title, but this
test fails for links produced by citeproc-hs in bibliographies.
So, if the link has a title, it will be lost on conversion
to an autolink, but that seems okay.

* Markdown writer: Set title, author, date variables as before.
These are no longer used in the default template, since we use
titleblock, but we set them anyway for those who use custom templates.

* LaTeX writer: Avoid extra space at start/end of table cell.
Thanks to Nick Bart for the suggestion of using @{}.

* `Text.Pandoc.Parsing`:

+ More efficient version of `anyLine`.
+ Type of `macro` has changed; the parser now returns `Blocks`
instead of `Block`.

* Relaxed old-time version bound, allowing 1.0.*.

* Removed obsolete `hsmarkdown` script. Those who need `hsmarkdown`
should create a symlink as described in the README.

pandoc (

* Markdown reader: Try `lhsCodeBlock` before `rawTeXBlock`. Otherwise
`\begin{code}...\end{code}` isn't handled properly in markdown+lhs.
Thanks to Daniel Miot for noticing the bug and suggesting the fix.

* Markdown reader: Fixed bug with headerless grid tables.
The 1.10 code assumed that each table header cell contains exactly one
block. That failed for headerless tables (0) and also for tables with
multiple blocks in a header cell. The code is fixed and tests provided.
Thanks to Andrew Lee for pointing out the bug.

* Markdown reader: Fixed regressions in fenced code blocks. Closes #722.

+ Tilde code fences can again take a bare language string
(`~~~ haskell`), not just curly-bracketed attributes
(`~~~ {.haskell}`).
+ Backtick code blocks can take the curly-bracketed attributes.
+ Backtick code blocks don't *require* a language.
+ Consolidated code for the two kinds of fenced code blocks.

* LaTeX template: Use `\urlstyle{same}` to avoid monospace URLs.

* Markdown writer: Use proportional font for email autolinks with
obfuscation. Closes #714.

* Corrected name of `blank_before_blockquote` in README. Closes #718.

* `Text.Pandoc.Shared`: Fixed bug in `uri` parser.
The bug prevented an autolink at the end of a string (e.g.
at the end of a line block line) from counting as a link. Closes #711.

* Use the `hsb2hs` preprocessor instead of TH for embed_data_files.
This should work on Windows, unlike the TH solution with

* Eliminated use of TH in test suite.

* Added `Text.Pandoc.Data` (non-exported) to hold the association
list of embedded data files, if the `embed_data_files` flag is selected.
This isolates the code that needs special treatment with file-embed or

* Changes to `make-windows-installer.bat`.

+ Exit batch file if any of the cabal-dev installs fail.
+ There's no longer any need to reinstall `highlighting-kate`.
+ Don't start with a `cabal update`; leave that to the user.
+ Force reinstall of pandoc.

* Fixed EPUB writer so it builds with blaze-html 0.4.x. Thanks to
Jens Petersen.

pandoc (

* Fixed bug with escaped % in LaTeX reader. Closes #710.

pandoc (

* Added further missing fb2 tests to cabal file.

pandoc (

* Added fb2 tests to cabal file's extra-source-files.

pandoc (

* Bump version bounds on test-framework packages.

pandoc (1.10)

[new features]

* New input formats: `mediawiki` (MediaWiki markup).

* New output formats: `epub3` (EPUB v3 with MathML),
`fb2` (FictionBook2 ebooks).

* New `--toc-depth` option, specifying how many levels of
headers to include in a table of contents.

* New `--epub-chapter-level` option, specifying the header
level at which to divide EPUBs into separate files.
Note that this normally affects only performance, not the
visual presentation of the EPUB in a reader.

* Removed the `--strict` option. Instead of using `--strict`,
one can now use the format name `markdown_strict` for either input
or output. This gives more fine-grained control that `--strict`
did, allowing one to convert from pandoc's markdown to strict
markdown or vice versa.

* It is now possible to enable or disable specific syntax
extensions by appending them (with `+` or `-`) to the writer
or reader name. For example,

pandoc -f markdown-footnotes+hard_line_breaks

disables footnotes and enables treating newlines as hard
line breaks. The literate Haskell extensions are now implemented
this way as well, using either `+lhs` or `+literate_haskell`.
For a list of extension names, see the README under
"Pandoc's Markdown."

* The following aliases have been introduced for specific
combinations of markdown extensions: `markdown_phpextra`,
`markdown_github`, `markdown_mmd`, `markdown_strict`. These aliases
work just like regular reader and writer names, and can be modified
with extension modifiers as described above. (Note that conversion
from one markdown dialect to another does not work perfectly,
because there are differences in markdown parsers besides
just the extensions, and because pandoc's internal document model is
not rich enough to capture all of the extensions.)

* New `--html-q-tags` option. The previous default was to use `<q>`
tags for smart quotes in HTML5. But `<q>` tags are also valid HTML4.
Moreover, they are not a robust way of typesetting quotes, since
some user agents don't support them, and some CSS resets (e.g.
bootstrap) prevent pandoc's quotes CSS from working properly.
We now just insert literal quote characters by default in both
`html` and `html5` output, but this option is provided for
those who still want `<q>` tags.

* The markdown reader now prints warnings (to stderr) about
duplicate link and note references. Closes #375.

* Markdown syntax extensions:

+ Added pipe tables. Thanks to François Gannaz for the initial patch.
These conform to PHP Markdown Extra's pipe table syntax. A subset
of org-mode table syntax is also supported, which means that you can
use org-mode's nice table editor to create tables.

+ Added support for RST-style line blocks. These are
useful for verse and addresses.

+ Attributes can now be specified for headers, using the same
syntax as in code blocks. (However, currently only the
identifier has any effect in most writers.) For example,

# My header {#foo}

See [the header above](#foo).

+ Pandoc will now act as if link references have been defined
for all headers without explicit identifiers.
So, you can do this:

# My header

Link to [My header].
Another link to [it][My header].

Closes #691.

* LaTeX reader:

+ Command macros now work everywhere, including non-math.
Environment macros still not supported.
+ `\input` now works, as well as `\include`. TEXINPUTS is used.
Pandoc looks recursively into included files for more included files.

[behavior changes]

* The Markdown reader no longer puts the text of autolinks in a
`Code` inline. This means that autolinks will no longer appear
in a monospace font.

* The character `/` can now appear in markdown citation keys.

* HTML blocks in strict_markdown are no longer required to begin
at the left margin. Technically this is required, according to
the markdown syntax document, but `` and other markdown
processors are more liberal.

* The `-V` option has been changed so that if there are duplicate
variables, those specified later on the command line take precedence.

* Tight lists now work in LaTeX and ConTeXt output.

* The LaTeX writer no longer relien on the `enumerate` package.
Instead, it uses standard LaTeX commands to change the list numbering

* The LaTeX writer now uses `longtable` instead of `ctable`. This allows
tables to be split over page boundaries.

* The RST writer now uses a line block to render paragraphs containing
linebreaks (which previously weren't supported at all).

* The markdown writer now applies the `--id-prefix` to footnote IDs.
Closes #614.

* The plain writer no longer uses backslash-escaped line breaks
(which are not very "plain").

* `Text.Pandoc.UTF8`: Better error message for invalid UTF8.
Read bytestring and use `Text`'s decodeUtf8 instead of using
`System.IO.hGetContents`. This way you get a message saying
"invalid UTF-8 stream" instead of "invalid byte sequence."
You are also told which byte caused the problem.

* Docx, ODT, and EPUB writers now download images specified by a URL
instead of skipping them or raising an error.

* EPUB writer:

+ The default CSS now left-aligns headers by default, instead of
centering. This is more consistent with the rest of the writers.
+ A proper multi-level table of contents is now used in `toc.ncx`.
There is no longer a subsidiary table of contents at the beginning
of each chapter.
+ Code highlighting now works by default.
+ Section divs are used by default for better semantic markup.
+ The title is used instead of "Title Page" in the table of contents.
Otherwise we have a hard-coded English string, which looks
strange in ebooks written in other languages. Closes #572.

* HTML writer:

+ Put mathjax in span with class "math". Closes #562.
+ Put citations in a span with class "citation." In HTML5, also include
a `data-cite` attribute with a space-separated list of citation

* `Text.Pandoc.UTF8`: use universalNewlineMode in reading.
This treats both `\r\n` and `\n` as `\n` on input, no matter
what platform we're running on.

* Citation processing is now done in the Markdown and LaTeX
readers, not in `pandoc.hs`. This makes it easier for library users
to use citations.

[template changes]

* HTML: Added css to template to preserve spaces in `<code>` tags.
Thanks to Dirk Laurie.

* Beamer: Remove English-centric strings in section pages.
Section pages used to have "Section" and a number as well as the
section title. Now they just have the title. Similarly for part
and subsection. Closes #566.

* LaTeX, ConTeXt: Added papersize variable.

* LaTeX, Beamer templates: Use longtable instead of ctable.

* LaTeX, Beamer templates: Don't require 'float' package for tables.
We don't actually seem to use the '[H]' option.

* LaTeX: Use `upquote` package if it is available. This fixes
straight quotes in verbatim environments.

* Markdown, plain: Fixed titleblock so it is just a single string.
Previously separate title, author, and date variables were used,
but this didn't allow different kinds of title blocks.


+ Rationalized templates. Previously there were three different
templates involved in epub production. There is now just one
template, `default.epub` or `default.epub3`. It can now be
overridden using `--template`, just like other templates.
The titlepage is now folded into the default template.
A `titlepage` variable selects it.
+ UTF-8, lang tag, meta tags, title element.

* Added scale-to-width feature to beamer template

[API changes]

* `Text.Pandoc.Definition`: Added `Attr` field to `Header`.
Previously header identifiers were autogenerated by the writers.
Now they are added in the readers (either automatically or explicitly).

* `Text.Pandoc.Builder`:

+ `Inlines` and `Blocks` are now synonyms for `Many Inline` and
`Many Block`. `Many` is a newtype wrapper around `Seq`, with
custom Monoid instances for `Many Inline` and `Many Block. This
allows `Many` to be made an instance of `Foldable` and `Traversable`.
+ The old `Listable` class has been removed.
+ The module now exports `isNull`, `toList`, `fromList`.
+ The old `Read` and `Show` instances have been removed; derived
instances are now used.
+ Added `headerWith`.

* The readers now take a `ReaderOptions` rather than a `ParserState`
as a parameter. Indeed, not all parsers use the `ParserState` type;
some have a custom state. The motivation for this change was to separate
user-specifiable options from the accounting functions of parser state.

* New module `Text.Pandoc.Options`. This includes the `WriterOptions`
formerly in `Text.Pandoc.Shared`, and its associated
data types. It also includes a new type `ReaderOptions`, which
contains many options formerly in `ParserState`, and its associated
data types:

+ `ParserState.stateParseRaw` -> `ReaderOptions.readerParseRaw`.
+ `ParserState.stateColumns` -> `ReaderOptions.readerColumns`.
+ `ParserState.stateTabStop` -> `ReaderOptions.readerTabStop`.
+ `ParserState.stateOldDashes` -> `ReaderOptions.readerOldDashes`.
+ `ParserState.stateLiterateHaskell` -> `ReaderOptions.readerLiterateHaskell`.
+ `ParserState.stateCitations` -> `ReaderOptions.readerReferences`.
+ `ParserState.stateApplyMacros` -> `ReaderOptions.readerApplyMacros`.
+ `ParserState.stateIndentedCodeClasses` ->
+ Added `ReaderOptions.readerCitationStyle`.

* `WriterOptions` now includes `writerEpubVersion`, `writerEpubChapterLevel`,
`writerEpubStylesheet`, `writerEpubFonts`, `writerReferenceODT`,
`writerReferenceDocx`, and `writerTOCDepth`. `writerEPUBMetadata` has
been renamed `writerEpubMetadata` for consistency.

* Changed signatures of `writeODT`, `writeDocx`, `writeEPUB`, since they no
longer stylesheet, fonts, reference files as separate parameters.

* Removed `writerLiterateHaskell` from `WriterOptions`, and
`readerLiterateHaskell` from `ReaderOptions`. LHS is now handled
by an extension (`Ext_literate_haskell`).

* Removed deprecated `writerXeTeX`.

* Removed `writerStrict` from `WriterOptions`. Added `writerExtensions`.
Strict is now handled through extensions.

* `Text.Pandoc.Options` exports `pandocExtensions`, `strictExtensions`,
`phpMarkdownExtraExtensions`, `githubMarkdownExtensions`,
and `multimarkdownExtensions`, as well as the `Extensions` type.

* New `Text.Pandoc.Readers.MediaWiki` module, exporting

* New `Text.Pandoc.Writers.FB2` module, exporting `writeFB2`
(thanks to Sergey Astanin).

* `Text.Pandoc`:

+ Added `getReader`, `getWriter` to `Text.Pandoc`.
+ `writers` is now an association list `(String, Writer)`.
A `Writer` can be a `PureStringWriter`, an `IOStringWriter`, or
an `IOByteStringWriter`. ALL writers are now in the 'writers'
list, including the binary writers and FB2 writer. This allows
code in `pandoc.hs` to be simplified.
+ Changed type of `readers`, so all readers are in IO.
Users who want pure readers can still get them form the reader
modules; this just affects the function `getReader` that looks up
a reader based on the format name. The point of this change is to
make it possible to print warnings from the parser.

* `Text.Pandoc.Parsing`:

+ `Text.Parsec` now exports all Parsec functions used in pandoc code.
No other module directly imports Parsec. This will make it easier
to change the parsing backend in the future, if we want to.
+ `Text.Parsec` is used instead of `Text.ParserCombinators.Parsec`.
+ Export the type synonym `Parser`.
+ Export `widthsFromIndices`, `NoteTable'`, `KeyTable'`, `Key'`, `toKey'`,
`withQuoteContext`, `singleQuoteStart`, `singleQuoteEnd`,
`doubleQuoteStart`, `doubleQuoteEnd`, `ellipses`, `apostrophe`,
`dash`, `nested`, `F(..)`, `askF`, `asksF`, `runF`, `lineBlockLines`.
+ `ParserState` is no longer an instance of `Show`.
+ Added `stateSubstitutions` and `stateWarnings` to `ParserState`.
+ Generalized type of `withQuoteContext`.
+ Added `guardEnabled`, `guardDisabled`, `getOption`.
+ Removed `failIfStrict`.
+ `lookupKeySrc` and `fromKey` are no longer exported.

* `Data.Default` instances are now provided for `ReaderOptions`,
`WriterOptions`, and `ParserState`. `Text.Pandoc` re-exports `def`.
Now you can use `def` (which is re-exported by `Text.Pandoc`) instead
of `defaultWriterOptions` (which is still defined). Closes #546.

* `Text.Pandoc.Shared`:

+ Added `safeRead`.
+ Renamed `removedLeadingTrailingSpace` to `trim`,
`removeLeadingSpace` to `triml`, and `removeTrailingSpace` to `trimr`.
+ Count `\r` as space in `trim` functions.
+ Moved `renderTags'` from HTML reader and `Text.Pandoc.SelfContained`
to `Shared`.
+ Removed `failUnlessLHS`.
+ Export `compactify'`, formerly in Markdown reader.
+ Export `isTightList`.
+ Do not export `findDataFile`.
+ `readDataFile` now returns a strict ByteString.
+ Export `readDataFileUTF8` which returns a String, like the
old `readDataFile`.
+ Export `fetchItem` and `openURL`.

* `Text.Pandoc.ImageSize`: Use strict, not lazy bytestrings.
Removed `readImageSize`.

* `Text.Pandoc.UTF8`: Export `encodePath`, `decodePath`,
`decodeArg`, `toString`, `fromString`, `toStringLazy`,

* `Text.Pandoc.UTF8` is now an exposed module.

* `Text.Pandoc.Biblio`:

+ csl parameter now a `String` rather than a `FilePath`.
+ Changed type of `processBiblio`. It is no longer in the IO monad.
It now takes a `Maybe Style` argument rather than parameters for CSL
and abbrev filenames. (`pandoc.hs` now calls the functions to parse
the style file and add abbreviations.)

* Markdown reader now exports `readMarkdownWithWarnings`.

* `Text.Pandoc.RTF` now exports `writeRTFWithEmbeddedImages` instead of

[bug fixes]

* Make `--ascii` work properly with `--self-contained`. Closes #568.

* Markdown reader:

+ Fixed link parser to avoid exponential slowdowns. Closes #620.
Previously the parser would hang on input like this:


We fixed this by making the link parser parser characters
between balanced brackets (skipping brackets in inline code spans),
then parsing the result as an inline list. One change is that

[hi *there]* bud](/url)

is now no longer parsed as a link. But in this respect pandoc behaved
differently from most other implementations anyway, so that seems okay.

+ Look for raw html/latex blocks before tables.
Otherwise the following gets parsed as a table:

-- My comment.

Closes #578.

* RST reader:

+ Added support for `:target:` on `.. image::` blocks
and substitutions.
+ Field list fixes:

- Fixed field lists items with body beginning after a new line
(Denis Laxalde).
- Allow any char but ':' in names of field lists in RST reader
(Denis Laxalde).
- Don't allow line breaks in field names.
- Require whitespace after field list field names.
- Don't create empty definition list for metadata field lists.
Previously a field list consisting only of metadata fields (author,
title, date) would be parsed as an empty DefinitionList, which is
not legal in LaTeX and not needed in any format.

+ Don't recognize inline-markup starts inside words.
For example, `2*2 = 4*1` should not contain an emphasized
section. Added test case for "Literal symbols". Closes #569.
+ Allow dashes as separator in simple tables. Closes #555.
+ Added support for `container`, `compound`, `epigraph`,
`rubric`, `highlights`, `pull-quote`.
+ Added support for `.. code::`.
+ Made directive labels case-insensitive.
+ Removed requirement that directives begin at left margin.
This was (correctly) not in earlier releases; docutils doesn't
make the requirement.
+ Added support for `replace::` and `unicode::` substitutions.
+ Ignore unknown interpreted roles.
+ Renamed image parser to `subst`, since it now handles all
substitution references.

* Textile reader:

+ Allow newlines before pipes in table. Closes #654.
+ Fixed bug with list items containing line breaks.
Now pandoc correctly handles hard line breaks inside list items.
Previously they broke list parsing.
+ Implemented comment blocks.
+ Fixed bug affected words ending in hyphen.
+ Properly handle links with surrounding brackets.
Square brackets need to be used when the link isn't surrounded by
spaces or punctuation, or when the URL ending may be ambiguous.
Closes #564.
+ Removed nullBlock. Better to know about parsing problems than
to skip stuff when we get stuck.
+ Allow ID attributes on headers.
+ Textile reader: Avoid parsing dashes as strikeout.
Previously the input


would be parsed with strikeouts rather than dashes. This fixes
the problem by requiring that a strikeout delimiting - not be
followed by a -. Closes #631.
+ Expanded list of `stringBreakers`.
This fixes a bug on input like "(_hello_)" which should
be a parenthesized emphasized "hello".
The new list is taken from the PHP source of textile 2.4.
+ Fixed autolinks. Previously the textile reader and writer
incorrectly implented RST-style autolinks for URLs and email
addresses. This has been fixed. Now an autolink is done this way:
+ Fixed footnotes bug in textile. This affected notes occurring
before punctuation, e.g. `foo[1].`. Closes #518.

* LaTeX reader:

+ Better handling of citation commands.
+ Better handling of `\noindent`.
+ Added a 'try' in rawLaTeXBlock, so we can handle `\begin` without `{`.
Closes #622.
+ Made `rawLaTeXInline` try to parse block commands as well. This
is usually what we want, given how `rawLaTeXInline` is used in
the markdown and textile readers. If a block-level LaTeX command
is used in the middle of a paragraph (e.g. `\subtitle` inside a title),
we can treat it as raw inline LaTeX.
+ Handle \slash command. Closes #605.
+ Basic `\enquote` support.
+ Fixed parsing of paragraphs beginning with a group. Closes #606.
+ Use curly quotes for bare straight quotes.
+ Support obeylines environment. Closes #604.
+ Guard against "begin", "end" in inlineCommand and
+ Better error messages for environments. Now it should tell you that
it was looking for \end{env}, instead of giving "unknown parse error."

* HTML reader:

+ Added HTML 5 tags to list of block-level tags.
+ HTML reader: Fixed bug in `htmlBalanced`, which
caused hangs in parsing certain markdown input using
strict mode.
+ Parse `<q>` as `Quoted DoubleQuote`.
+ Handle nested `<q>` tags properly.
+ Modified `htmlTag` for fewer false positives.
A tag must start with `<` followed by `!`,`?`, `/`, or a letter.
This makes it more useful in the wikimedia and markdown parsers.

* DocBook reader: Support title in "figure" element. Closes #650.

* MediaWiki writer:

+ Remove newline after `<br/>` in translation of `LineBreak`
There's no particular need for a newline (other than making the
generated MediaWiki source look nice to a human), and in fact
sometimes it is incorrect: in particular, inside an enumeration, list
items cannot have embedded newline characters. (Brent Yorgey)
+ Use `<code>` not `<tt>` for Code.

* Man writer: Escape `-` as `\-`.
Unescaped `-`'s become hyphens, while `\-`'s are left as ascii minus
signs. That is preferable for use with command-line options. See
<>. Thanks
to Andrea Bolognani for bringing the issue to our attention.

* RST writer:

+ Improved line block output. Use nonbreaking spaces for
initial indent (otherwise lost in HTML and LaTeX).
Allow multiple paragraphs in a single line block.
Allow soft breaks w continuations in line blocks.
+ Properly handle images with no alt text. Closes #678.
+ Fixed bug with links with duplicate text. We now (a) use anonymous
links for links with inline URLs, and (b) use an inline link instead
of a reference link if the reference link would require a label that
has already been used for a different link. Closes #511.
+ Fixed hyperlinked images. Closes #611. Use `:target:`
field when you have a simple linked image.
+ Don't add `:align: center` to figures.

* Texinfo writer: Fixed internal cross-references.
Now we insert anchors after each header, and use `@ref` instead of `@uref`
for links. Commas are now escaped as `@comma{}` only when needed;
previously all commas were escaped. (This change is needed, in part,
because `@ref` commands must be followed by a real comma or period.) Also
insert a blank line in from of `@verbatim` environments.

* DocBook writer:

+ Made --id-prefix work in DocBook as well as HTML.
Closes #607.
+ Don't include empty captions in figures. Closes #581.

* LaTeX writer:

+ Use `\hspace*` for nonbreaking space after line break,
since `~` spaces after a line break are just ignored.
Closes #687.
+ Don't escape `_` in URLs or hyperref identifiers.
+ Properly escape strings inside \url{}. Closes #576.
+ Use `[fragile]` only for slides containing code rendered
using listings. Closes #649.
+ Escape `|` as `\vert` in LaTeX math. This avoids a clash with
highlighting-kate's macros, which redefine `|` as a short verbatim
delimiter. Thanks to Björn Peemöller for raising this issue.
+ Use minipage rather than parbox for block containers in tables.
This allows verbatim code to be included in grid tables.
Closes #663.
+ Prevent paragraphs containing only linebreaks or spaces.

* HTML writer:

+ Included `highlighting-css` for code spans, too.
Previously it was only included if used in a code block. Closes #653.
+ Improved line breaks with `<dd>` tags. We now put a newline between
`</dd>` and `<dd>` when there are multiple definitions.
+ Changed mathjax cdn url so it doesn't use https. (This caused
problems when used with `--self-contained`.) See #609.

* EPUB writer:

+ `--number-sections` now works properly.
+ Don't strip meta and link elements in epub metadata.
Patch from aberrancy. Closes #589.
+ Fixed a couple validation bugs.
+ Use ch001, ch002, etc. for chapter filenames. This improves sorting
of chapters in some readers, which apparently sort ch2 after ch10.
Closes #610.

* ODT writer: properly set title property (Arlo O'Keeffe).

* Docx writer:

+ Fixed bug with nested lists. Previously a list like

1. one
- a
- b
2. two

would come out with a bullet instead of "2."
Thanks to Russell Allen for reporting the bug.
+ Use `w:cr` in `w:r` instead of `w:br` for linebreaks.
This seems to fix a problem viewing pandoc-generated
docx files in LibreOffice.
+ Use integer ids for bookmarks. Closes #626.
+ Added nsid to abstractNum elements. This helps when merging
word documents with numbered or bulleted lists. Closes #627.
+ Use separate footnotes.xml for notes.
This seems to help LibreOffice convert the file, even though
it was valid docx before. Closes #637.
+ Use rIdNN identifiers for r:embed in images.
+ Avoid reading image files again when we've already processed them.
+ Fixed typo in `referenc.docx` that prevented image captions from
working. Thanks to Huashan Chen.

* `Text.Pandoc.Parsing`:

+ Fixed bug in `withRaw`, which didn't correctly handle the case
where nothing is parsed.
+ Made `emailAddress` parser more correct. Now it is based on RFC 822,
though it still doesn't implement quoted strings in email addresses.
+ Revised URI parser. It now allows many more schemes, allows
uppercase URIs, and better handles trailing punctuation and
trailing slashes in bare URIs. Added many tests.
+ Simplified and improved singleQuoteStart. This makes `'s'`, `'l'`,
etc. parse properly. Formerly we had some English-centric heuristics,
but they are no longer needed. Closes #698.

* `Text.Pandoc.Pretty`: Added wide punctuation range to `charWidth`.
This fixes bug with Chinese commas in markdown and reST tables, and
a bug that caused combining characters to be dropped.

* `Text.Pandoc.MIME`: Added MIME types for .wof and .eot. Closes #640.

* `Text.Pandoc.Biblio`:

+ Run `mvPunc` and `deNote` on metadata too.
This fixed a bug with notes on titles using footnote styles.
+ Fixed bug in fetching CSL files from CSL data directory.

* `pandoc.hs`: Give correct value to `writerSourceDirectory` when a URL
is provided. It should be the URL up to the path.

* Fixed/simplified diff output for tests.
Biblio: Make sure mvPunc and deNote run on metadata too.
This fixed a bug with notes on titles using footnote styles.

[under the hood improvements]

* We no longer depend on `utf8-string`. Instead we use functions
defined in `Text.Pandoc.UTF8` that use `Data.Text`'s conversions.

* Use `safeRead` instead of using `reads` directly (various modules).

* "Implicit figures" (images alone in a paragraph) are now handled
differently. The markdown reader gives their titles the prefix `fig:`; the
writers look for this before treating the image as a figure. Though this
is a bit of a hack, it has two advantages: (i) implicit figures can be
limited to the markdown reader, and (ii) they can be deactivated by turning
off the `implicit_figures` extension.

* `catch` from `Control.Exception` is now used instead of the
old Preface `catch`.

* `Text.Pandoc.Shared`: Improved algorithm for `normalizeSpaces`
and `oneOfStrings` (which is now non-backtracking).

* `Text.Pandoc.Biblio`: Remove workaround for `toCapital`.
Now citeproc-hs is fixed upstream, so this is no longer needed.
Closes #531.

* Textile reader: Improved speed of `hyphenedWords`.
This speeds up the textile reader by about a factor of 4.

* Use `Text.Pandoc.Builder` in RST reader, for more flexibility,
better performance, and automatic normalization.

* Major rewrite of markdown reader:

+ Use `Text.Pandoc.Builder` instead of lists. This also
means that everything is normalized automatically.
+ Move to a one-pass parsing strategy, returning values in the reader
monad, which are then run (at the end of parsing) against the final
parser state.

* In HTML writer, we now use `toHtml` instead of pre-escaping.
We work around the problem that blaze-html unnecessarily escapes `'`
by pre-escaping just the `'` characters, instead of the whole string.
If blaze-html later stops escaping `'` characters, we can simplify
`strToHtml` to `toHtml`. Closes #629.

* Moved code for embedding images in RTFs from `pandoc.hs` to the
RTF writer (which now exports `writeRTFWithEmbeddedImages`).

* Moved citation processing from `pandoc.hs` into the readers.
This makes things more convenient for library users.

* The man pages are now built by an executable `make-pandoc-man-pages`,
which has its own stanza in the cabal file so that dependencies can be
handled by Cabal. Special treatment in `Setup.hs` ensures that this
executable never gets installed; it is only used to create the man pages.

* The cabal file has been modified so that the pandoc library is used
in building the pandoc executable. (This required moving `pandoc.hs`
from `src` to `.`.) This cuts compile time in half.

* `-O2` is no longer used in building pandoc. The performance improvement
it yields is so slight that it is not worth it. (Measured with
benchmarks on ghc 7.4.)

* The `executable` and `library` flags have been removed.

* `-threaded` has been removed from ghc-options.

* Version bounds of dependencies have been raised, and the
`blaze_html_0_5` flag now defaults to True. Pandoc now compiles on
GHC 7.6.

* We now require base >= 4.2.

* Integrated the benchmark program into cabal. One can now do:

cabal configure --enable-benchmarks && cabal build
cabal bench --benchmark-option='markdown' --benchmark-option='-s 20'

The benchmark now uses README + testsuite, so benchmark results
from older versions aren't comparable.

* Integrated test suite with cabal.
To run tests, configure with `--enable-tests`, then `cabal test`.
You can specify particular tests using `--test-options='-t markdown'`.
No output is shown unless tests fail. The Haskell test modules
have been moved from `src/` to `tests/`.

* Moved all data files and templates to the `data/` subdirectory.

* Added an `embed_data_files` cabal flag. This causes all
data files to be embedded in the binary, so that the binary
is self-sufficient and can be relocated anywhere, copied on
a USB key, etc. The Windows installer now uses this.
(Since we no longer have the option to build the executable
without the library, this is the only way to get a relocatable
binary on Windows.)

* Removed pcre3.dll from windows package.
It isn't needed unless highlighting-kate is compilled with the
`pcre-light` flag. By default, regex-prce-builtin is used.

pandoc (

* Don't encode/decode file paths if base >= 4.4.
Prior to base 4.4, filepaths and command line arguments were treated
as unencoded lists of bytes, not unicode strings, so we had to work
around that by encoding and decoding them. This commit adds CPP
checks for the base version that intelligibly enable encoding/decoding
when needed. Fixes a bug with multilingual filenames when pandoc was
compiled with ghc 7.4 (#540).

* Don't generate an empty H1 after hrule slide breaks.
We now use a slide-level header with contents `[Str "\0"]` to mark
an hrule break. This avoids creation of an empty H1 in these
contexts. Closes #484.

* Docbook reader: Added support for "bold" emphasis. Thanks to Mauro Bieg.

* In, ensure citeproc-hs is built with the
embed_data_files flag.

* MediaWiki writer: Avoid extra blank lines after sublists (Gavin Beatty).

* ConTeXt writer: Don't escape `&`, `^`, `<`, `>`, `_`,
simplified escapes for `}` and `{` to `\{` and `\}` (Aditya Mahajan).

* Fixed handling of absolute URLs in CSS imports with `--self-contained`.
Closes #535.

* Added webm to mime types. Closes #543.

* Added some missing exports and tests to the cabal file
(Alexander V Vershilov).

* Compile with `-rtsopts` and `-threaded` by default.

pandoc (

* Markdown reader: Added `cf.` and `cp.` to list of likely abbreviations.

* LaTeX template: Added `linkcolor`, `urlcolor` and `links-as-notes`
variables. Make TOC links black.

* LaTeX template improvements.

+ Don't print date unless one is given explicitly in the document.
+ Simplified templates.
+ Use fontenc [T1] by default, and lmodern.
+ Use microtype if available.

* Biblio:

+ Add comma to beginning of bare suffix, e.g. `@item1 [50]`.
Motivation: `@item1 [50]` should be as close as possible to
`[@item1, 50]`.
+ Added workaround for a bug in citeproc-hs 0.3.4 that causes footnotes
beginning with a citation to be empty. Closes #531.

* Fixed documentation on mixed lists. Closes #533.

pandoc (1.9.4)

* Simplified `Text.Pandoc.Biblio` and fixed bugs with citations inside
footnotes and captions. We now handle note citations by inserting
footnotes during initial citation processing, and doing a separate
pass later to remove notes inside notes.

* Added 'zenburn' highlight style from highlighting-kate.

* Added Slideous writer. Slideous is an HTML + javascript slide show
format, similar to Slidy, but works with IE 7. (Jonas Smedegaard)

* LaTeX writer:

+ Ensure we don't have extra blank lines at ends of cells.
This can cause LaTeX errors, as they are interpreted as new paragraphs.
+ More consistent interblock spacing.
+ Require highlighting-kate >= 0.5.1, for proper highlighted inline
code in LaTeX. Closes #527.
+ Ensure that a Verbatim at the end of a footnote is followed by
a newline. (Fixes a regression in the previous version.)
+ In default template, use black for internal links and TOC.
Added commented-out code to use footnotes for links, as would
be suitable in print output.

* Beamer writer: When `--incremental` is used, lists inside
a block quote should appear all at once. (This makes Beamer
output consistent with the HTML slide show formats.)

* ConTeXt writer:

+ Escape `%` as `\letterpercent{}` not `\letterpercent `,
to avoid gobbling spaces after the `%` sign.
+ Ensure space after `\stopformula`.

* Markdown writer:

+ Use `:` form instead of `~` in definition lists, for better
compatibility with other markdown implementations.
+ Don't wrap the term, because it breaks definition lists.
+ Use a nonzero space to prevent false recognition
of list marker in ordered lists. Closes #516.

* Org writer: Add space before language name. Closes #523.

* Docx writer: Simplified bullet characters so they work properly
with Word 2007. Closes #520.

* LaTeX reader: Support `\centerline`.

* RST reader: handle figures. Closes #522.

* Textile reader: fix for `<notextile>` and `==`. Closes #517.
(Paul Rivier)

pandoc (1.9.3)

* Fixed bug in `fromEntities`. The previous version would turn
`hi & low you know;` into `hi &`.

* HTML reader:

+ Don't skip nonbreaking spaces.
Previously a paragraph containing just `&nbsp;` would be rendered
as an empty paragraph. Thanks to Paul Vorbach for pointing out the bug.
+ Support `<col>` and `<caption>` in tables. Closes #486.

* Markdown reader:

+ Don't recognize references inside delimited code blocks.
+ Allow list items to begin with lists.

* Added basic docbook reader (John MacFarlane and Mauro Bieg).

* LaTeX reader:

+ Handle `\bgroup`, `\egroup`, `\begingroup`, `\endgroup`.
+ Control sequences can't be followed by a letter.
This fixes a bug where `\begingroup` was parsed as `\begin`
followed by `group`.
+ Parse 'dimension' arguments to unknown commands. e.g. `\parindent0pt`
+ Make `\label` and `\ref` sensitive to `--parse-raw`.
If `--parse-raw` is selected, these will be parsed as raw latex
inlines, rather than bracketed text.
+ Don't crash on unknown block commands (like `\vspace{10pt}`)
inside `\author`; just skip them. Closes #505.

* Textile reader:

+ Implemented literal escapes with `==` and `<notextile>`. Closes #473.
+ Added support for LaTeX blocks and inlines (Paul Rivier).
+ Better conformance to RedCloth inline parsing (Paul Rivier).
+ Parse '+text+' as emphasized (should be underlined, but this
is better than leaving literal plus characters in the output.

* Docx writer: Fixed multi-paragraph list items. Previously they each
got a list marker. Closes #457.

* LaTeX writer:

+ Added `--no-tex-ligatures` option to avoid replacing
quotation marks and dashes with TeX ligatures.
+ Use `fixltx2e` package to provide '\textsubscript'.
+ Improve spacing around LaTeX block environments:
quote, verbatim, itemize, description, enumerate.
Closes #502.
+ Use blue instead of pink for URL links in latex/pdf output.

* ConTeXt writer: Fixed escaping of `%`.
In text, `%` needs to be escaped as `\letterpercent`, not `\%`
Inside URLs, `%` needs to be escaped as `\%`
Thanks to jmarca and adityam for the fix. Closes #492.

* Texinfo writer: Escape special characters in node titles.
This fixes a problem pointed out by Joost Kremers. Pandoc used
to escape an '@' in a chapter title, but not in the corresponding
node title, leading to invalid texinfo.

* Fixed document encoding in texinfo template.
Resolves Debian Bug #667816.

* Markdown writer:

+ Don't force delimited code blocks to be flush left.
Fixes bug with delimited code blocks inside lists etc.
+ Escape `<` and `$`.

* LaTeX writer: Use `\hyperref[ident]{text}` for internal links.
Previously we used `\href{\#ident}{text}`, which didn't work on
all systems. Thanks to Dirk Laurie.

* RST writer: Don't wrap link references. Closes #487.

* Updated to use latest versions of blaze-html, mtl.

pandoc (1.9.2)

* LaTeX reader:

+ Made `lstlisting` work as a proper verbatim environment.
+ Fixed bug parsing LaTeX tables with one column.

* LaTeX writer:

+ Use `{}` around `ctable` caption, so that formatting can be used.
+ Don't require eurosym package unless document has a €.

* LaTeX template: Added variables for `geometry`, `romanfont`,
`sansfont`, `mathfont`, `mainfont` so users can more easily
customize fonts.

* PDF writer:

+ Run latex engine at least two times, to ensure
that PDFs will have hyperlinked bookmarks.
+ Added PDF metadata (title,author) in LaTeX standalone + PDF output.

* Texinfo writer: retain directories in image paths. (Peter Wang)

* RST writer: Better handling of inline formatting, in accord
with docutils' "inline markup recognition rules" (though we don't
implement the unicode rules fully). Now `hi*there*hi` gets
rendered properly as `hi\ *there*\ hi`, and unnecessary
`\ ` are avoided around `:math:`, `:sub:`, `:sup:`.

* RST reader:

+ Parse `\ ` as null, not escaped space.
+ Allow `` :math:`...` `` even when not followed by blank
or `\`. This does not implement the complex rule docutils follows,
but it should be good enough for most purposes.
+ Add support for the rST default-role directive. (Greg Maslov)

* Text.Pandoc.Parsing: Added `stateRstDefaultRole` field to `ParserState`.
(Greg Maslov)

* Markdown reader: Properly handle citations nested in other inline

* Markdown writer: don't replace empty alt in image with "image".

* DZSlides: Updated template.html and styles in default template.
Removed bizarre CSS for `q` in dzslides template.

* Avoid repeated `id` attribute in section and header in HTML slides.

* README improvements: new instructions on internal links,
removed misleading note on reST math.

* Build system:

+ Fixed Windows installer so that dzslides works.
+ Removed
+ Added .travis.yml for Travis continuous integration support..
+ Fixed upper bound for zlib (Sergei Trofimovich).
+ Fixed upper bound for test-framework.
+ Updated haddocks for haddock-2.10 (Sergei Trofimovich).

pandoc (

* Added `beamer+lhs` as output format.

* Don't escape `<` in `<style>` tags with `--self-contained`.
This fixes a bug which prevented highlighting from working
when using `--self-contained`.

* PDF: run latex engine three times if `--toc` specified.
This fixes page numbers in the table of contents.

* Docx writer: Added TableNormal style to tables.

* LaTeX math environment fixes. `aligned` is now used instead of
the nonexistent `aligned*`. `multline` instead of the nonexistent

* LaTeX writer: Use `\textasciitilde` for literal `~`.

* HTML writer: Don't escape contents of EQ tags with --gladtex.
This fixes a regression from 1.8.

* Use `<q>` tags for Quoted items for HTML5 output.
The quote style can be changed by modifying the template
or including a css file. A default quote style is included.

* LaTeX reader: Fixed accents (\~{a}, `\c{c}`).
Correctly handle \^{}. Support "minted" as a LaTeX verbatim block.

* Updated LaTeX template for better language support.
Use `polyglossia` instead of `babel` with xetex.
Set `lang` as documentclass option.
`\setmainlanguage` will use the last of a comma-separated
list of languages. Thanks to François Gannaz.

* Fixed default LaTeX template so `\euro` and `€` work. The
`eurosym` package is needed if you are using pdflatex.

* Fixed escaping of period in man writer (thanks to Michael Thompson).

* Fixed list label positions in beamer.

* Set `mainlang` variable in context writer.
This parallels behavior of latex writer. `mainlang` is the last
of a comma-separated list of languages in lang.

* EPUB language metadat: convert e.g. `en_US` from locale to `en-US`.

* Changed `-V` so that you can specify a key without a value.
Such keys get the value `true`.

* Fixed permissions on installed man pages - thanks Magnus Therning.

* Windows installer: require XP or higher. The installer is
now compiled on a Windows 7 machine, which fixes a problem
using citation functions on Windows 7.

* OSX package: Check for 64-bit Intel CPU before installing.

pandoc (

* Better handling of raw latex environments in markdown. Now


turns into a raw latex block as expected.

* Improvements to LaTeX reader:

+ Skip options after block commands.
+ Correctly handle `{\\}` in braced.
+ Added a needed 'try'.
+ Citations: add `, ` to suffix if it doesn't start with space or
punctuation. Otherwise we get no space between the year and the
suffix in author-date styles.

* Added two needed data files for S5. This fixes a problem with
`pandoc -t s5 --self-contained`. Also removed `slides.min.js`,
which was no longer being used.

* Fixed some minor problems in `reference.docx`:
name on "Date" style, `xCs` instead of `xIs`.

* Fixed a problem creating docx files using a reference docx
modified using Word. The problem seems to be that Word
modifies `_rels/.rels`, changing the Type of the Relationship to
`docProps/core.xml`. Pandoc now changes this back to the correct
value if it has been altered, fixing the problem.

* Fixed html5 template so it works properly with highlighting.

pandoc (1.9.1)

* LaTeX reader:

+ Fixed regression in 1.9; properly handle escaped $ in latex math.
+ Put LaTeX verse environments in blockquotes.

* Markdown reader:

+ Limit nesting of strong/emph. This avoids exponential lookahead
in parasitic cases, like `a**a*a**a*a**a*a**a*a**a*a**a*a**a*a**`.
+ Improved attributes syntax (inn code blocks/spans):
(1) Attributes can contain line breaks. (2) Values in key-value
attributes can be surrounded by either double or single quotes, or
left unquoted if they contain no spaces.

* Don't wrap headers in markdown or RST writers.

* Added `stateMaxNestingLevel` to `ParserState`.
We set this to 6, so you can still have `Emph` inside `Emph`,
just not indefinitely.

* More efficient implementation of `nowrap` in `Text.Pandoc.Pretty`.

* `Text.Pandoc.PDF`: Only run latex twice if `\tableofcontents`
is present.

* Require highlighting-kate >=, texmath >=

pandoc (

* Changed cabal file so that build-depends for the test program
are not required unless the tests flag is used.

* LaTeX writer: insert `{}` between adjacent hyphens so they don't
form ligatures (dashes) in code spans.

pandoc (

* Raised version bound on test-framework to avoid problems
compiling tests on GHC 7.4.1.

* LaTeX reader: Use raw LaTeX as fallback inline text for Cites,
so citations don't just disappear unless you process with
citeproc. Ignore `\bibliographystyle`, `\nocite`.

* Simplified tex2pdf; it will always run latex twice to
resolve table of contents and hyperrefs.

pandoc (

* Require Cabal >= 1.10.
* Tweaked cabal file to meet Cabal 1.10 requirements.

pandoc (

* Allow build with json 0.4 or 0.5. Otherwise we can't build with
ghc 6.12.

pandoc (1.9)

[new features]

* Added a Microsoft Word `docx` writer. The writer includes support
for highlighted code and for math (which is converted from TeX to OMML,
Office's native math markup language, using texmath's new OMML module).
A new option `--reference-docx` allows the user to customize the

* Added an `asciidoc` writer (<>).

* Better support for slide shows:

+ Added a `dzslides` writer. DZSlides is a lightweight HTML5/javascript
slide show format due to Paul Rouget (<>).

+ Added a LaTeX `beamer` writer. Beamer is a LaTeX package for creating
slide presentations.

+ New, flexible rules for dividing documents into sections and slides
(see the "Structuring the slide show" in the User's Guide). These
are backward-compatible with the old rules, but they allow slide
shows to be organized into sections and subsections containing
multiple slides.

+ A new `--slide-level` option allows users to override defaults
and select a slide level below the first header level with content.

* A new `--self-contained` option produces HTML output that does not
depend on an internet connection or the presence of any external
files. Linked images, CSS, and javascript is downloaded (or fetched
locally) and encoded in `data:` URIs. This is useful for making portable
`HTML slide shows. The --offline` option has been deprecated and is now
`treated as a synonym or --self-contained`.

* Support for PDF output:

+ Removed the old `markdown2pdf`.
+ `pandoc` can now create PDFs (assuming you have latex and a set of
appropriate packages installed): just specify an output file with the
`.pdf` extension.
+ A new option `--latex-engine` allows you to specify `pdflatex`,
`xelatex`, or `lualatex` as the processor.

* Highlighting changes:

+ Syntax highlighting is now a standard feature; the `highlighting`
flag is no longer needed when compiling.
+ A new `--no-highlight` option allows highlighting to be disabled.
+ Highlighting now works in `docx`, `latex`, and `epub`, as well as
`html`, `html5`, `dzslides`, `s5`, and `slidy`.
+ A new `--highlight-style` option selects between various highlighting
color themes.

* Internal links to sections now work in ConTeXt and LaTeX as well as HTML.

* LaTeX `\include` and `\usepackage` commands are now processed,
provided the files are in the working directory.

* EPUB improvements:

+ Internal and external links now work in EPUB.
+ Raw HTML is allowed.
+ New `--epub-embed-font` option.
+ Customizable templates for EPUB pages offer more control over
formatting: `epub-page.html`, `epub-coverimage.html`,

* `--mathml` now works with DocBook.

* Added support for math in RST reader and writer. Inline math uses the
`` :math:`...` `` construct. Display math uses

.. math:: ...

or if the math is multiline,

.. math::


These constructions are now supported now by ``.

* GitHub syntax for fenced code blocks is supported in pandoc's
markdown. You can now write

x = 2

instead of

~~~ {.ruby}
x = 2

* Easier scripting: a new `toJsonFilter` function makes it easier to
write Haskell scripts to manipulate the Pandoc AST.

[behavior changes]

* Fixed parsing of consecutive lists in markdown.
Pandoc previously behaved like for consecutive
lists of different styles. Thus, the following would be parsed
as a single ordered list, rather than an ordered list followed
by an unordered list:

1. one
2. two

- one
- two

This change makes pandoc behave more sensibly, parsing this as
two lists. Any change in list type (ordered/unordered) or in
list number style will trigger a new list. Thus, the following
will also be parsed as two lists:

1. one
2. two

a. one
b. two

Since we regard this as a bug in, and not something
anyone would ever rely on, we do not preserve the old behavior
even when `--strict` is selected.

* Dashes work differently with `--smart`: `---` is always em-dash,
and `--` is always en-dash. Pandoc no longer tries to guess when
`-` should be en-dash. *Note:* This may change how existing documents
look when processed with pandoc. A new option, `--old-dashes`,
is provided for legacy documents.

* The markdown writer now uses setext headers for levels 1-2.
The old behavior (ATX headers for all levels) can be restored
using the new `--atx-headers` option.

* Links are now allowed in markdown image captions. They are also
allowed in links, but will appear there as regular text. So,

[link with [link](/url)](/url)

will turn into

<p><a href="/url">link with link</a></p>

* Improved handling of citations using `citeproc-hs-0.3.4`.
Added `--citation-abbreviations` option.

* Citation keys can no longer end with a punctuation character.
This means that `@item1.` will be parsed as a citation with key
'item1', followed by a period, instead of a citation with key
'item1.', as was the case previously.

* In HTML output, citations are now put in a span with class `citation`.

* The markdown reader now recognizes DocBook block and inline tags.
It was always possible to include raw DocBook tags in a markdown
document, but now pandoc will be able to distinguish block from
inline tags and behave accordingly. Thus, for example,


will not be wrapped in `<para>` tags.

* The LaTeX parser has been completely rewritten; it is now much more
accurate, robust, and extensible. However, there are two important
changes in how it treats unknown LaTeX. (1) Previously, unknown
environments became BlockQuote elements; now, they are treated
as "transparent", so `\begin{unknown}xyz\end{unknown}` is the
same as `xyz`. (2) Previously, arguments of unknown commands
were passed through with their braces; now the braces are stripped

* `--smart` is no longer selected automatically with `man` output.

* The deprecated `--xetex` option has been removed.

* The `--html5`/`-5` option has been deprecated. Use `-t html5`
instead. `html5` and `html5+lhs` are now separate output formats.

* Single quotes are no longer escaped in HTML output. They do not
need to be escaped outside of attributes.

* Pandoc will no longer transform leading newlines in code
blocks to `<br/>` tags.

* The ODT writer now sizes images appropriately, using the image
size and DPI information embedded in the image.

* `--standalone` is once again implicitly for a non-text output format
(ODT, EPUB). You can again do `pandoc test.txt -o test.odt`
and get a standalone ODT file.

* The Docbook writer now uses `<sect1>`, `<sect2>`, etc. instead of

* The HTML writer now uses `<del>` for strikeout.

* In HTML output with `--section-divs`, the classes `section` and
`level[1,2,..6]` are put on the `div` tags so they can be styled.
In HTML 5 output with `--section-divs`, the classes
`level[1,2,...6]` are put on `section` tags.

* EPUB writer changes:

+ The `lang` variable now sets the language
in the metadata (if it is not set, we default to the locale).
+ EPUB: UTF-8 is used rather than decimal entities.

* Added `titleslide` class to title slide in S5 template.

* In HTML, EPUB, and docx metadata, the date is normalized into
YYYY-MM-DD format if possible. (This is required for validation.)

* Attributes in highlighted code blocks are now preserved in HTML.
The container element will have the classes, id, and key-value attributes
you specified in the delimited code block. Previously these were stripped

* The reference backlink in the HTML writer no longer has a special
`footnoteBacklink` class.

* The HTML template has been split into `html` and `html5` templates.

* Author and date are treated more consistently in HTML templates.
Authors are now `<h2>`, date `<h3>`.

* URLs are hyphenated in the ConTeXt writer (B. Scott Michel).

* In `Text.Pandoc.Builder`, `+++` has been replaced by `<>`.

[bug fixes]

* Better support for combining characters and East Asian wide characters
in markdown and reST.

* Better handling of single quotes with `--smart`.
Previously `D'oh l'*aide*` would be parsed with left and right single
quotes instead of apostrophes. This kind of error is now fixed.

* Highlighting: Use `reads` instead of `read` for better error handling.
Fixes crash on `startNum="abc"`.

* Added blank comment after directives in rst template.

* Unescape entities in citation `refId`. The `refId`s coming
from citeproc contain XML numeric entities, and these don't match with the
citation keys parsed by pandoc. Solution is to unescape them.

* HTML reader: Fixed bug parsing tables with both thead and tbody.

* Markdown reader:

+ Better handling of escapes in link URLs and titles.
+ Fixed backslash escapes in reference links.
+ Fixed bug in table/hrule parsing, by checking that the top
line of a table is not followed by a blank line. This bug caused
slowdowns on some files with hrules and tables, as pandoc tried to
interpret the hrules as the tops of multiline tables.
+ Fixed bug in code block attribute parser. Previously the ID attribute
got lost if it didn't come first. Now attributes can come in any order.

* RST reader: allow footnotes followed by newline without space characters.

* LaTeX reader:

+ Ignore empty groups {}, { }.
+ LaTeX reader: Handle \@.
+ LaTeX reader: Don't crash on commands like `\itemsep`.
+ LaTeX reader: Better handling of letter environments.

* RST writer: Fixed bug involving empty table cells. isSimple was being
calculated in a way that assumed there were no non-empty cells.

* ConTeXt writer:

+ Made `--toc` work even without `--number-sections`.
+ Escape # in link URLs.
+ Use buffering for footnotes containing code blocks.
+ Changed 'descr' to 'description', fixed alignment.

* LaTeX writer:

+ Escape euro character.
+ Don't escape ~ inside href{...}.
+ Escape # in href URLs.
+ Improved detection of book classes. We now check the
`documentclass` variable, and if that is not set, we look through
the template itself. Also, we have added the KOMA classes scrreprt
and scrbook. You can now make a book using
`pandoc -V documentclass:book mybook.txt -o mybook.pdf`
+ LHS files now set the "listings" variable, so that the definition
of the `code` environment will be included in the template.
+ Links are colored blue by default (this can be changed by
modifying `hyperref` settings in the template).
+ Added `lang` variable to LaTeX template.

* HTML writer:

+ Fixed bug in HTML template with html5 and mathml.
+ Don't use self-closing img, br, hr tags for HTML5.
+ Use `<section>` for footnotes if HTML5.
+ Update HTML templates to use Content-Style-Type meta tag.
+ Use separate variables for meta-date, meta-author.
This makes footnotes work in author and date fields.
+ Use 'vertical-align:middle' in WebTeX math for better alignment.

* S5/slidy writer: Make footnotes appear on separate slide at end.

* MIME: Added 'layout-cache' to getMimeType. This ensures that the
META-INF/manifest.xml for ODT files will have everything it needs, so
that ODT files modified by LibreOffice can be used as `--reference-odt`.

* `Text.Pandoc.Templates`: Return empty string for json template.

* `Text.Pandoc.Biblio`:

+ Expand citations recursively inside nested inlines.
+ Treat `\160` as space when parsing locator and suffix.
This fixes a bug with "p. 33" when `--smart` is used. Previously
the whole "p. 33" would be included in the suffix, with no locator.
+ Put whole author-in-text citation in a Cite. Previously just the
date and other info went in the Cite.
+ Don't add comma+space to prefix if it ends in punctuation.

* Updated chicago-author-date.csl. The old version did not work
properly for edited volumes with no author.

* EPUB writer:

+ Add date to EPUB titlepage and metadata.
+ Added TOC identifier in EPUB page template.
+ Don't generate superfluous file `cover-image.jpg`.

[under the hood improvements]

* Modified `` to use cabal-dev.
Items are no longer installed as root.
Man pages are zipped and given proper permissions.

* Modified windows installer generator to use cabal-dev.

* Setup: Making man pages now works with cabal-dev (at least on OSX). In
Setup.hs we now invoke 'runghc' in a way that points it to the correct
package databases, instead of always falling back to the default user
package db.

* Updated to work with GHC 7.4.1.

* Removed dependency on old-time.

* Removed dependency on dlist.

* New slidy directory for "self-contained."

* TeXMath writer: Use unicode thin spaces for thin spaces.

* Markdown citations: don't strip off initial space in locator.

[API changes]

* Removed `Apostrophe`, `EmDash`, `EnDash`, and `Ellipses`
from the native `Inline` type in pandoc-types. Now we use `Str`
elements with unicode.

* Improvements to `Text.Pandoc.Builder`:

+ `Inlines` and `Blocks` are now newtypes (not synonyms for
+ Instances are defined for `IsString`, `Show`, `Read`, `Monoid`,
and a new `Listable` class, which allows these to be manipulated
to some extent like lists. Monoid append includes automatic
+ `+++` has been replaced by `<>` (mappend).

* Use blaze-html instead of xhtml for HTML generation.
This changes the type of `writeHtml`.

* `Text.Pandoc.Shared`:

+ Added `warn` and `err`.
+ Removed `unescapeURI`, modified `escapeURI`.
(See under [behavior changes], above.)

* Changes in URI escaping: Previously the readers escaped URIs by
converting unicode characters to octets and then percent encoding.
Now unicode characters are left as they are, and `escapeURI` only
percent-encodes space characters. This gives more readable
URIs, and works well with modern user agents. URIs are no longer unescaped
at all on conversion to `markdown`, `asciidoc`, `rst`, `org`.

* New module `Text.Pandoc.SelfContained`.

* New module `Text.Pandoc.Docx`.

* New module `Text.Pandoc.PDF`.

* Added `writerBeamer` to `WriterOptions`.

* Added `normalizeDate` to `Text.Pandoc.Shared`.

* Added `splitStringWithIndices` in `Text.Pandoc.Shared`.
This is like `splitWithIndices`, but it is sensitive to distinctions
between wide, combining, and regular characters.

* `Text.Pandoc.Pretty`:

+ Added `chomp` combinator.
+ Added `beforeNonBreak` combinator. This allows you to include
something conditionally on it being before a nonblank.
Used for RST inline math.
+ Added `charWidth` function. All characters marked W or F in the unicode
spec EastAsianWidth.txt get width 2.
+ Added `realLength`, based on `charWidth`. `realLength` is now
used in calculating offsets.

* New module `Text.Pandoc.Slides`, for common functions for breaking
a document into slides.

* Removed `Text.Pandoc.S5`, which is no longer needed.

* Removed `Text.Pandoc.CharacterReferences`. Moved
`characterReference` to `Text.Pandoc.Parsing`.
`decodeCharacterReferences` is replaced by `fromEntities`
in `Text.Pandoc.XML`.

* Added `Text.Pandoc.ImageSize`. This is intened for use
in `docx` and `odt` writers, so the size and dpi of images
can be calculated.

* Removed `writerAscii` in `WriterOptions`.

* Added `writerHighlight` to `WriterOptions`.

* Added `DZSlides` to `HTMLSlideVariant`.

* `writeEPUB` has a new argument for font files to embed.

* Added `stateLastStrPos` to `ParserState`. This lets us keep track
of whether we're parsing the position immediately after a regular
(non-space, non-symbol) string, which is useful for distinguishing
apostrophes from single quote starts.

* `Text.Pandoc.Parsing`:

+ `escaped` now returns a `Char`.
+ Removed `charsInBalanced'`, added a character parser as
a parameter of `charsInBalanced`. This is needed for
proper handling of escapes, etc.
+ Added `withRaw`.

* Added `toEntities` to `Text.Pandoc.XML`.

* `Text.Pandoc.Readers.LaTeX`:

+ Export `handleIncludes`.
+ Export `rawLaTeXBlock` instead of `rawLaTeXEnvironment'`.

* Added `ToJsonFilter` class and `toJsonFilter` function to
`Text.Pandoc`, deprecating the old `jsonFilter` function.

* `Text.Pandoc.Highlighting`:

+ Removed `highlightHtml`, `defaultHighlightingCss`.
+ Export `formatLaTeXInline`, `formatLaTeXBlock`, and `highlight`, plus
key functions from highlighting-kate.
+ Changed types of highlighting function. `highlight` returns a
`Maybe`, not an `Either`.

pandoc (

* Relaxed cabal consntraints for test-framework (S. Trofimovich).

* Relaxed cabal constraints for pandoc-types.

* Adjusted Arbitrary instance to help avoid timeouts in tests.

* Added `Tests.Writers.Markdown` to cabal file.

pandoc (1.8.2)

* Added script to produce OS X package.

* Made `templates` directory a git submodule. This should make it
easier for people to revise their custom templates when the default
templates change.

* Changed template naming scheme: `FORMAT.template` -> `default.FORMAT`.
**Note:** If you have existing templates in `~/.pandoc/templates`, you
must rename them to conform to the new scheme!

* Fixed smart quotes bug, now handling `'...hi'` properly.

* RST reader:

+ Partial support for labeled footnotes.
+ Improved accuracy of `simpleReferenceName` parser.

* HTML reader:

+ Substitute correct unicode characters for
characters in the 128..159 range, which are often found even in
HTML that purports to be UTF-8.

* LaTeX reader: Handle `\subtitle` command (a subtitle is added
to the title, after a colon and linebreak). Closes #280.

* Leaner `reference.odt`.

* Added unexported module `Text.Pandoc.MIME` for use in
the ODT writer.

* ODT writer: Construct `manifest.xml` based on archive contents.
This fixes a bug in ODTs containing images. Recent versions of
LibreOffice would reject these as corrupt, because `manifest.xml`
did not contain a reference to the image files.

* LaTeX writer:

+ Make verbatim environments flush to avoid spurious
blank lines. Closes #277.
+ Use `\texttt` and escapes insntead of `\verb!...!`, which
is too fragile (doesn't work in command arguments).
+ Use `\enquote{}` for quotes if the template includes
the `csquotes` package. This provides better support for
local quoting styles. (Thanks to Andreas Wagner for the idea.)

* ConTeXt writer: Make `\starttyping`/`\stoptyping` flush with
margin, preventing spurious blank lines.

* Slidy writer:

+ Use non-minimized version of `slidy.css` with `--offline`
option, so users can more easily edit it.
+ Also fixed a bug in the CSS that prevented proper centering
of title (now reported and fixed upstream).

* S5 writer:

+ Replaced `s5/default/slides.js.{comment,packed}` with
new compressed `s5/default/slides.min.js`.
+ Use `data:` protocol to embed S5 CSS in `<link>` tags,
when `--offline` is specified. Using inline CSS didn't
work with Chrome or Safari. This fixes offline
S5 on those browsers.

* HTML writer: Removed English title on footnote backlinks.
This is incongrous in non-English documents.

* Docbook writer:

+ Use CALS tables. (Some older docbook software does not work
well with XHTML tables.) Closes #77.
+ Use `programlisting` tags (instead of `screen`) for code blocks.

* `markdown2pdf`:

+ Calls latex with `-halt-on-error -interaction nonstopmode` instead
of `-interaction=batchmode`, which essentially just ignored errors,
leading to bad results. Better to know when something is wrong.
+ Fixed issues with non-UTF-8 output of `pdflatex`.
+ Better error reporting.

* `--mathjax` now takes an optional URL argument. If it is not
provided, pandoc links directly to the (secure) mathjax CDN,
as now recommended (thanks to dsanson).

* Deprecated `--xetex` option in `pandoc`. It is no longer needed,
since the LaTeX writer now produces a file that can be processed by
`latex`, `pdflatex`, `lualatex`, or `xelatex`.

* Introduced `--luatex` option to `markdown2pdf`. This causes `lualatex`
to be used to create the PDF.

* If a template specified with `--template` is not found, look for it
in `datadir`. Also, if no extension is provided, supply one based
on the writer. So now you can put your `special.latex` template in
`~/.pandoc/templates`, and use it from any directory via
`pandoc -t latex --template special`.

* Default template improvements:

+ HTML: Display author and date after title.
+ HTML: Made table of contents more customizable. The container
for the TOC is now in the template, so users can insert a header
or other styling. (Thanks to Bruce D'Arcus for the suggestion.)
+ HTML, Slidy, S5: Enclose scripts in CDATA tags.
+ Slidy, S5: Added `s5-url` and `slidy-url` variables, instead of
hard-coding. If you want to put your slidy files in the slidy
subdirectory, for example, you can do
`pandoc -t slidy -V slidy-url=slidy -s`.
+ LaTeX: Use `\and` to separate authors in LaTeX documents (reader
& writer). Closes #279.
+ LaTeX: Use different `hyperref` options for `xetex`, fixing
problems with unicode bookmarks (thanks to CircleCode).
+ LaTeX: Removed `ucs` package, use `utf8` rather than `utf8x`
with `inputenc`. This covers fewer characters but is more
robust with other packages, and `ucs` is unmaintained. Users
who need better unicode support should use xelatex or lualatex.

pandoc (

* Added `--epub-cover-image` option.

* Documented `--biblatex` and `--natbib` options.

* Allow `--section-divs` with slidy output. Resolves Issue #296.

* Disallow notes within notes in reST and markdown.
These previously caused infinite looping and stack overflows.
For example:


[^1]: See [^1]

Note references are allowed in reST notes, so this isn't a full
implementation of reST. That can come later. For now we need to
prevent the stack overflows. Partially resolves Issue #297.

* EPUB writer: Allow non-plain math methods.

* Forbid ()s in citation item keys. Resolves Issue #304: problems with
`(@item1; @item2)` because the final paren was being parsed as part of
the item key.

* Changed URI parser so it doesn't include trailing punctuation.
So, in RST, `` should be parsed as a link followed by a
period. The parser is smart enough to recognize balanced parentheses, as
often occur in wikipedia links: ``.

* Markdown+lhs reader: Require space after inverse bird tracks, so that
HTML tags can be used freely at the left margin of a markdown+lhs document.
Thanks to Conal Elliot for the suggestion.

* Markdown reader:

+ Improved emph/strong parsing; fixes bug found by Perry Wagle.
+ Fixed bug in footnote order (reported by CircleCode).

* RST reader:
+ Fixed bug in in field lists with multi-line items at the
end of the list.
+ Added parentheses to RST `specialChars`, so
`(` will be parsed as a link in parens.
Resolves Issue #291.
+ Allow `|` followed by newline in RST line block.

* LaTeX reader:
+ Support `\dots`.
+ Gobble option & space after linebreak `\\[10pt]`.

* Textile reader:
+ Make it possible to have colons after links. (qerub)
+ Make it possible to have colons after links. (Christoffer Sawicki)

* HTML reader:
+ Skip spaces after `<b>`, `<emph>`, etc.
+ Handle tbody, thead in simple tables. Closes #274.
+ Implicit `Para`s instead of `Plains` in some contexts.

* OpenDocument writer: Use special `First paragraph` style for
first paragraph after most non-paragraph blocks. This allows users to
specify e.g. that only paragraphs after the first paragraph of a block are
to be indented. Thanks to Andrea Rossato for the patch. Closes #20.

* LaTeX writer: use `deVerb` on table and picture captions.
Otherwise LaTeX complains about `\verb` inside command argument.
Thanks to bbanier for reporting the bug.

* Markdown writer: Insert HTML comment btw list and indented code block.
This prevents the code block from being interpreted as part of the list.

* EPUB writer: Add a meta element specify the cover.
Some EPUB e-readers, such as the Nook, require a meta element inside the
OPF metadata block to ensure the cover image is properly displayed.
(Kelsey Hightower)

* HTML writer: Use embed tag for images with non-image extensions.
(e.g. PDFs). Closes #264.

* LaTeX writer: Improved tables.

+ More space between lines, top-align cells.
+ Use ctable package, which allows footnotes and
provides additional options.
+ Made cell alignments work in multiline tables.
+ Closes #271, #272.

* Un-URI-escape image filenames in LaTeX, ConTeXt, RTF, Texinfo.
Also do this when copying image files into EPUBs and ODTs.
Closes #263.

* Changed to github issue tracker.

* Added failing emph/strong markdown test case due to Perry Wagle.

* Slidy improvements:
+ Updated to use Slidy2.
+ Fixed bug, unclosed div tag.
+ Added `duration` variable in template.
Setting this activates the timer.
+ Use 'titlepage' instead of 'cover' for title div.

pandoc (

* `markdown2pdf`: Removed some debugging lines accidentally included
in the 1.8.1 release. With those lines, the temp directory is created
in the working directory, and it is not deleted. This fix restores
the original behavior.

pandoc (1.8.1)

* Added `--ascii` option. Currently supported only in HTML writer,
which it causes to use numerical entities instead of UTF-8.

* EPUB writer: `--toc` now works to provide a table of contents
at the beginning of each chapter.

* LaTeX writer: Change figure defaults to `htbp`.
This prevents "too many unprocessed floats." Resolves
Issue #285.

* `Text.Pandoc.UTF8`: Encode filenames even when using recent

* `markdown2pdf`: Fixed filename encoding issues. With help from Paulo
Tanimoto. Resolves Issue #286.

* HTML writer: Put line breaks in section divs.

* `Text.Pandoc.Shared`: Make `writerSectionDivs` default to False.

pandoc (

* Fixed Source-repository stanza in cabal file.

pandoc (

* HTML writer:

+ Stringify alt text instead of converting to HTML.
+ Break lines after block elements, not inside tags.
HTML output now closely resembles that of tidy. Resolves Issue #134.

* Markdown reader: Fixed bug in footnote block parser (pointed out
by Jesse Rosenthal). The problem arose when the blank line
at the end of a footnote block contained indenting spaces.

* Shared: Improved 'normalize' function so it normalizes Spaces too.
In normal form, Space elements only occur to separate two non-Space
elements. So, we never have [Space], or [, ..., Space].

* Tests:

+ Improved Arbitrary instance.
+ Added timeout for test instances.


+ Added section on four-space rule for lists. Resolves Issue #283.
+ Clarified optional arguments on math options.

* markdown2pdf: Fixed bug with output file extensions.
Previously `markdown2pdf test.txt -o test.en.pdf` would produce
`test.pdf`, not `test.en.pdf`. Thanks to Paolo Tanimoto for the fix.

pandoc (

* Revised Interact.hs so that it works with the CPP macros
in the UTF8 module.

* Revised Setup.hs so that we don't call MakeManPage.hs unless
the man pages are out of date.

pandoc (1.8)

[new features]

* Support for citations using Andrea Rossato's `citeproc-hs` 0.3.
You can now write, for example,

Water is wet [see @doe99, pp. 33-35; also @smith04, ch. 1].

and, when you process your document using `pandoc`, specifying
a citation style using `--csl` and a bibliography using `--bibliography`,
the citation will be replaced by an appropriately formatted
citation, and a list of works cited will be added to the end
of the document.

This means that you can switch effortlessly between different citation
and bibliography styles, including footnote, numerical, and author-date
formats. The bibliography can be in any of the following formats: MODS,
BibTeX, BibLaTeX, RIS, EndNote, EndNote XML, ISI, MEDLINE, Copac, or JSON.
See the README for further details.

Citations are supported in the markdown reader, using a special
syntax, and in the LaTeX reader, using natbib or biblatex syntax.
(Thanks to Nathan Gass for the natbib and biblatex support.)

* New `textile` reader and writer. Thanks to Paul Rivier for contributing
the `textile` reader, an almost complete implementation of the textile
syntax used by the ruby [RedCloth library](
Resolves Issue #51.

* New `org` writer, for Emacs Org-mode, contributed by Puneeth Chaganti.

* New `json` reader and writer, for reading and writing a JSON
representation of the native Pandoc AST. These are much faster
than the `native` reader and writer, and should be used for
serializing Pandoc to text. To convert between the JSON representation
and native Pandoc, use `encodeJSON` and `decodeJSON` from

* A new `jsonFilter` function in `Text.Pandoc` makes it easy
to write scripts that transform a JSON-encoded pandoc document.
For example:

-- removelinks.hs - removes links from document
import Text.Pandoc
main = interact $ jsonFilter $ bottomUp removeLink
where removeLink (Link xs _) = Emph xs
removeLink x = x

To use this to remove links while translating markdown to LaTeX:

pandoc -t json | runghc removelinks.hs | pandoc -f json -t latex

* Attributes are now allowed in inline `Code` elements, for example:

In this code, `ulist ! [theclass "special"] << elts`{.haskell} is...

The attribute syntax is the same as for delimited code blocks.
`Code` inline has an extra argument place for attributes, just like
`CodeBlock`. Inline code will be highlighted in HTML output, if pandoc
is compiled with highlighting support. Resolves Issue #119.

* New `RawBlock` and `RawInline` elements (replacing `RawHtml`,
`HtmlInline`, and `TeX`) provide lots of flexibility in writing
scripts to transform Pandoc documents. Scripts can now change
how each element is rendered in each output format.

* You can now define LaTeX macros in markdown documents, and pandoc
will apply them to TeX math. For example,

\newcommand{\plus}[2]{#1 + #2}

yields `3+4`. Since the macros are applied in the reader, they
will work in every output format, not just LaTeX.

* LaTeX macros can also be used in LaTeX documents (both in math
and in non-math contexts).

* A new `--mathjax` option has been added for displaying
math in HTML using MathJax. Resolves issue #259.

* Footnotes are now supported in the RST reader. (Note, however,
that unlike docutils, pandoc ignores the numeral or symbol used in
the note; footnotes are put in an auto-numbered ordered list.)
Resolves Issue #258.

* A new `--normalize` option causes pandoc to normalize the AST
before writing the document. This means that, for example,
`*hi**there*` will be rendered as `<em>hithere</em>`
instead of `<em>hi</em><em>there</em>`. This is not the default,
because there is a significant performance penalty.

* A new `--chapters` command-line option causes headers
in DocBook, LaTeX, and ConTeXt to start with "chapter" (level one).
Resolves Issue #265.

* In DocBook output, `<chapter>` is now used for top-level
headers if the template contains `<book>`. Resolves Issue #265.

* A new `--listings` option in `pandoc` and `markdown2pdf` causes
the LaTeX writer to use the listings package for code blocks.
(Thanks to Josef Svennigsson for the pandoc patch, and Etienne
Millon for the markdown2pdf patch.)

* `markdown2pdf` now supports `--data-dir`.

* URLs in autolinks now have class "url" so they can be styled.

* Improved prettyprinting in most formats. Lines will be wrapped
more evenly and duplicate blank lines avoided.

* New `--columns` command-line option sets the column width for
line wrapping and relative width calculations for tables.

* Made `--smart` work in HTML, RST, and Textile readers, as well
as markdown.

* Added `--html5` option for HTML5 output.

* Added support for listings package in LaTeX reader
(Puneeth Chaganti).

* Added support for simple tables in the LaTeX reader.

* Added support for simple tables in the HTML reader.

* Significant performance improvements in many readers and writers.

[API and program changes]

* Moved `Text.Pandoc.Definition` from the `pandoc` package to a new
auxiliary package, `pandoc-types`. This will make it possible for other
programs to supply output in Pandoc format, without depending on the whole
pandoc package.

* Added `Attr` field to `Code`.

* Removed `RawHtml`, `HtmlInline`, and `TeX` elements; added generic
`RawBlock` and `RawInline`.

* Moved generic functions to `Text.Pandoc.Generic`. Deprecated
`processWith`, replacing it with two functions, `bottomUp` and `topDown`.
Removed previously deprecated functions `processPandoc` and `queryPandoc`.

* Added `Text.Pandoc.Builder`, for building `Pandoc` structures.

* `Text.Pandoc` now exports association lists `readers` and `writers`.

* Added `Text.Pandoc.Readers.Native`, which exports `readNative`.
`readNative` can now read full pandoc documents, block lists, blocks,
inline lists, or inlines. It will interpret `Str "hi"`
as if it were `Pandoc (Meta [] [] []) [Plain [Str "hi"]]`.
This should make testing easier.

* Removed deprecated `-C/--custom-header` option.
Use `--template` instead.

* `--biblio-file` has been replaced by `--bibliography`.
`--biblio-format` has been removed; pandoc now guesses the format
from the file extension (see README).

* pandoc will treat an argument as a URI only if it has an
`http(s)` scheme. Previously pandoc would treat some
Windows pathnames beginning with `C:/` as URIs.

* The `--sanitize-html` option and the `stateSanitize` field in
`ParserState` have been removed. Sanitization is better done in the
resulting HTML using `xss-sanitize`, which is based on pandoc's
sanitization, but improved.

* pandoc now adds a newline to the end of its output in fragment
mode (= not `--standalone`).

* Added support for `lang` in `html` tag in the HTML template,
so you can do `pandoc -s -V lang=es`, for example.

* `highlightHtml` in `Text.Pandoc.Highlighting` now takes
a boolean argument that selects between "inline" and
"block" HTML.

* `Text.Pandoc.Writers.RTF` now exports `rtfEmbedImage`.
Images are embedded in RTF output when possible (png, jpeg).
Resolves Issue #275.

* Added `Text.Pandoc.Pretty`. This is better suited for pandoc than the
`pretty` package. Changed all writers that used
`Text.PrettyPrint.HughesPJ` to use `Text.Pandoc.Pretty` instead.

* Rewrote `writeNative` using the new prettyprinting module. It is
now much faster. The output has been made more consistent and compressed.
`writeNative` is also now sensitive to writerStandalone`, and will simply
`print a block list if writerStandalone` is False.

* Removed `Text.Pandoc.Blocks`. `Text.Pandoc.Pretty` allows you to define
blocks and concatenate them, so a separate module is no longer needed.

* `Text.Pandoc.Shared`:

+ Added `writerColumns`, `writerChapters`, and `writerHtml5` to
+ Added `normalize`.
+ Removed unneeded prettyprinting functions:
`wrapped`, `wrapIfNeeded`, `wrappedTeX`, `wrapTeXIfNeeded`, `hang'`,
`BlockWrapper`, `wrappedBlocksToDoc`.
+ Made `splitBy` take a test instead of an element.
+ Added `findDataFile`, refactored `readDataFile`.
+ Added `stringify`. Rewrote `inlineListToIdentifier` using `stringify`.
+ Fixed `inlineListToIdentifier` to treat '\160' as ' '.

* `Text.Pandoc.Readers.HTML`:

+ Removed `rawHtmlBlock`, `anyHtmlBlockTag`, `anyHtmlInlineTag`,
`anyHtmlTag`, `anyHtmlEndTag`, `htmlEndTag`, `extractTagType`,
`htmlBlockElement`, `htmlComment`
+ Added `htmlTag`, `htmlInBalanced`, `isInlineTag`, `isBlockTag`,

* Moved `smartPunctuation` from `Text.Pandoc.Readers.Markdown`
to `Text.Pandoc.Readers.Parsing`, and parameterized it with
an inline parser.

* Added `nonspaceChar` to `Text.Pandoc.Parsing`.

* Ellipses are no longer allowed to contain spaces.
Previously we allowed '. . .', ' . . . ', etc. This caused
too many complications, and removed author's flexibility in
combining ellipses with spaces and periods.

* Allow linebreaks in URLs (treat as spaces). Also, a string of
consecutive spaces or tabs is now parsed as a single space. If you have
multiple spaces in your URL, use `%20%20`.

* `Text.Pandoc.Parsing`:

+ Removed `refsMatch`.
+ Hid `Key` constructor.
+ Removed custom `Ord` and `Eq` instances for `Key`.
+ Added `toKey` and `fromKey` to convert between `Key` and `[Inline]`.
+ Generalized type on `readWith`.

* Small change in calculation of relative widths of table columns.
If the size of the header > the specified column width, use
the header size as 100% for purposes of calculating
relative widths of columns.

* Markdown writer now uses some pandoc-specific features when `--strict`
is not specified: `\` newline is used for a hard linebreak instead of
two spaces then a newline. And delimited code blocks are used when
there are attributes.

* HTML writer: improved gladTeX output by setting ENV appropriately
for display or inline math (Jonathan Daugherty).

* LaTeX writer: Use `\paragraph`, `\subparagraph` for level 4,5 headers.

* LaTeX reader:

+ `\label{foo}` and `\ref{foo}` now become `{foo}` instead of `(foo)`.
+ `\index{}` commands are skipped.

* Added `fontsize` variable to default LaTeX template.
This makes it easy to set the font size using `markdown2pdf`:
`markdown2pdf -V fontsize=12pt input.txt`.

* Fixed problem with strikeout in LaTeX headers when using
hyperref, by adding a command to the default LaTeX template
that disables `\sout` inside pdf strings. Thanks to Joost Kremers
for the fix.

* The `COLUMNS` environment variable no longer has any effect.

[under-the-hood improvements]

* Pandoc now compiles with GHC 7. (This alone leads to a
significant performance improvement, 15-20%.)

* Completely rewrote HTML reader using tagsoup as a lexer. The
new reader is faster and more accurate. Unlike the
old reader, it does not get bogged down on some input
(Issues #277, 255). And it handles namespaces in tags
(Issue #274).

* Replaced `escapeStringAsXML` with a faster version.

* Rewrote `spaceChar` and some other parsers in Text.Pandoc.Parsing
for a significant performance boost.

* Improved performance of all readers by rewriting parsers.

* Simplified Text.Pandoc.CharacterReferences by using
entity lookup functions from TagSoup.

* `Text.Pandoc.UTF8` now uses the unicode-aware IO functions
from `System.IO` if base >= 4.2. This gives support for
windows line endings on windows.

* Remove duplications in documentation by generating the
pandoc man page from README, using `MakeManPage.hs`.

* README now includes a full description of markdown syntax,
including non-pandoc-specific parts. A new `pandoc_markdown`
man page is extracted from this, so you can look up markdown
syntax by doing `man pandoc_markdown`.

* Completely revised test framework (with help from Nathan Gass).
The new test framework is built when the `tests` Cabal flag is set. It
includes the old integration tests, but also some new unit and quickcheck
tests. Test output has been much improved, and you can now specify a glob
pattern after `cabal test` to indicate which tests should be run;
for example `cabal test citations` will run all the citation tests.

* Added a shell script, ``, for filtering ANSI control
sequences from test output: `cabal test | ./ > test.log`.

* Added `Interact.hs` to make it easier to use ghci while developing.
`Interact.hs` loads `ghci` from the `src` directory, specifying
all the options needed to load pandoc modules (including
specific package dependencies, which it gets by parsing

* Added `Benchmark.hs`, testing all readers + writers using criterion.

* Added ``, to make it easier to collect and archive
benchmark and lines-of-code stats.

* Added upper bounds to all cabal dependencies.

* Include man pages in extra-source-files. This allows users to
install pandoc from the tarball without needing to build the man

[bug fixes]

* Filenames are encoded as UTF8. Resolves Issue #252.

* Handle curly quotes better in `--smart` mode. Previously, curly quotes
were just parsed literally, leading to problems in some output formats.
Now they are parsed as `Quoted` inlines, if `--smart` is specified.
Resolves Issue #270.

* Text.Pandoc.Parsing: Fixed bug in grid table parser.
Spaces at end of line were not being stripped properly,
resulting in unintended LineBreaks.

* Markdown reader:

+ Allow HTML comments as inline elements in markdown.
So, `aaa <!-- comment --> bbb` can be a single paragraph.
+ Fixed superscripts with links: `^[link](/foo)^` gets
recognized as a superscripted link, not an inline note followed by
+ Fixed regression, making markdown reference keys case-insensitive again.
Resolves Issue #272.
+ Properly handle abbreviations (like `Mr.`) at the end of a line.
+ Better handling of intraword underscores, avoiding exponential
slowdowns in some cases. Resolves Issue #182.
+ Fixed bug in alignments in tables with blank rows in the header.

* RST reader:

+ Field lists now allow spaces in field names, and
block content in field values. (Thanks to Lachlan Musicman
for pointing out the bug.)
+ Definition list items are now always `Para` instead of
`Plain`, matching behavior of ``.
+ In image blocks, the description is parsed properly and
used for the alt attribute, not also the title.
+ Skip blank lines at beginning of file. Resolves
Debian #611328.

* LaTeX reader:

+ Improved parsing of preamble.
Previously you'd get unexpected behavior on a document that
contained `\begin{document}` in, say, a verbatim block.
+ Allow spaces between `\begin` or `\end` and `{`.
+ Support `\L` and `\l`.
+ Skip comments inside paragraphs.

* LaTeX writer:

+ Escape strings in `\href{..}`.
+ In nonsimple tables, put cells in `\parbox`.

* OpenDocument writer: don't print raw TeX.

* Markdown writer:

+ Fixed bug in `Image`. URI was getting unescaped twice!
+ Avoid printing extra blank lines at the end if there are
no notes or references.

* LaTeX and ConTeXt: Escape `[` and `]` as `{[}` and `{]}`.
This avoids unwanted interpretation as an optional argument.

* ConTeXt writer: Fixed problem with inline code. Previously
`}` would be rendered `\type{}}`. Now we check the string for '}' and '{'.
If it contains neither, use `\type{}`; otherwise use `\mono{}`
with an escaped version of the string.

* `:` now allowed in HTML tags. Resolves Issue #274.

pandoc (1.6)

[ John MacFarlane ]

* New EPUB and HTML Slidy writers. (Issue #122)

- EPUB is a standard ebook format, used in Apple's iBooks for
the iPad and iPhone, Barnes and Noble's nook reader, the Sony
reader, and many other devices.
- Slidy, like S5, is a system for producing HTML+javascript slide

* All input is assumed to be UTF-8, no matter what the locale and ghc
version, and all output is UTF-8. This reverts to pre-1.5 behavior.
Also, a BOM, if present, is stripped from the input.

* Markdown now supports grid tables, whose cells can contain
arbitrary block elements. (Issue #43)

* Sequentially numbered example lists in markdown with `@` marker.

* Markdown table captions can begin with a bare colon and no longer need
to include the English word "table." Also, a caption can now occur
either before or after the table. (Issue #227)

* New command-line options:

- `--epub-stylesheet` allows you to specify a CSS file that will
be used to style your ebook.
- `--epub-metadata` allows you to specify metadata for the ebook.
- `--offline` causes the generated HTML slideshow to include all
needed scripts and stylesheets.
- `--webtex` causes TeX math to be converted to images using the
Google Charts API (unless a different URL is specified).
- `--section-divs` causes div tags to be added around each section
in an HTML document. (Issue #230, 239)

* Default behavior of S5 writer in standalone mode has changed:
previously, it would include all needed scripts and stylesheets
in the generated HTML; now, only links are included unless
the `--offline` option is used.

* Default behavior of HTML writer has changed. Between 1.2 and 1.5,
pandoc would enclose sections in div tags with identifiers on the
div tags, so that the sections can be manipulated in javascript.
This caused undesirable interactions with raw HTML div tags. So,
starting with 1.6, the default is to put the identifiers directly
on the header tags, and not to include the divs. The `--section-divs`
option selects the 1.2-1.5 behavior.

* API changes:

- `HTMLMathMethod`: Added `WebTeX`, removed `MimeTeX`.
- `WriterOptions`: Added `writerUserDataDir`, `writerSourceDirectory`,
`writerEPUBMetadata` fields. Removed `writerIncludeBefore`,
- Added `headerShift` to `Text.Pandoc.Shared`.
- Moved parsing code and `ParserState` from `Text.Pandoc.Shared`
to a new module, `Text.Pandoc.Parsing`.
- Added `stateHasChapters` to `ParserState`.
- Added `HTMLSlideVariant`.
- Made `KeyTable` a map instead of an association list.
- Added accessors for `Meta` fields (`docTitle`, `docAuthors`,
- `Pandoc`, `Meta`, `Inline`, and `Block` have been given `Ord`
- Reference keys now have a type of their own (`Key`), with its
own `Ord` instance for case-insensitive comparison.
- Added `Text.Pandoc.Writers.EPUB`.
- Added `Text.Pandoc.UUID`.
- Removed `Text.Pandoc.ODT`, added `Text.Pandoc.Writers.ODT`.
Removed `saveOpenDocumentAsODT`, added `writeODT`.
- Added `Text.Pandoc.Writers.Native` and `writeNative`.
Removed `prettyPandoc`.
- Added `Text.Pandoc.UTF8` for portable UTF8 string IO.
- Removed `Text.Pandoc.Writers.S5` and the `writeS5` function.
Moved `s5Includes` to a new module, `Text.Pandoc.S5`.
To write S5, you now use `writeHtml` with `writerSlideVariant`
set to `S5Slides` or `SlidySlides`.

* Template changes. If you use custom templates, please update them,
particularly if you use syntax highlighting with pandoc. The old HTML
templates hardcoded highlighting CSS that will no longer work with
the most recent version of highlighting-kate.

- HTML template: avoid empty meta tag if no date.
- HTML template: Use default highlighting CSS from highlighting-kate
instead of hard-coding the CSS into the template.
- HTML template: insert-before text goes before the title, and
immediately after the <body> tag, as documented. (Issue #241)
- Added slidy and s5 templates.
- Added amssymb to preamble of latex template. (github Issue 1)

* Removed excess newlines at the end of output. Note: because output
will not contain an extra newline, you may need to make adjustments
if you are inserting pandoc's output into a template.

* In S5 and slidy, horizontal rules now cause a new slide, so you
are no longer limited to one slide per section.

* Improved handling of code in man writer. Inline code is now monospace,
not bold, and code blocks now use .nf (no fill) and .IP (indented para).

* HTML reader parses `<tt>` as Code. (Issue #247)

* html+lhs output now contains bird tracks, even when compiled without
highlighting support. (Issue #242)

* Colons are now no longer allowed in autogenerated XML/HTML identifiers,
since they have a special meaning in XML.

* Code improvements in ODT writer. Remote images are now replaced with
their alt text rather than a broken link.

* LaTeX reader improvements:

- Made latex `\section`, `\chapter` parsers more forgiving of
- Parse `\chapter{}` in latex.
- Changed `rawLaTeXInline` to accept `\section`, `\begin`, etc.
- Use new `rawLaTeXInline'` in LaTeX reader, and export `rawLaTeXInline`
for use in markdown reader.
- Fixes bug wherein `\section{foo}` was not recognized as raw TeX
in markdown document.

* LaTeX writer: images are automatically shrunk if they would extend
beyond the page margin.

* Plain, markdown, RST writers now use unicode for smart punctuation.

* Man writer converts math to unicode when possible, as in other writers.

* `markdown2pdf` can now recognize citeproc options.

* Command-line arguments are converted to UTF-8. (Issue #234)

* `Text.Pandoc.TeXMath` has been rewritten to use texmath's parser.
This allows it to handle a wider range of formulas. Also, if a formula
cannot be converted, it is left in raw TeX; formulas are no longer
partially converted.

* Unicode curly quotes are left alone when parsing smart quotes. (Issue

* Cabal file changes:

- Removed parsec < 3 restriction.
- Added 'threaded' flag for architectures where GHC lacks a threaded
- Use 'threaded' only for markdown2pdf; it is not needed for pandoc.
- Require highlighting-kate 0.2.7.

* Use explicit imports from `Data.Generics`. Otherwise we have a
conflict with the 'empty' symbol, introduced in syb >= 0.2. (Issue #237)

* New data files: slidy/slidy.min.js, slidy/slidy.min.css, epub.css.

pandoc (

[ John MacFarlane ]

* Fixed uniqueIdent in Shared so that header identifiers work as
advertized in the README and are are valid XHTML names.

pandoc (1.5.1)

[ John MacFarlane ]

* Fixed treatment of unicode characters in URIs.
+ Shared now exports escapeURI and unescapeURI. These handle
UTF8 encoding and decoding as well as URI escaping/unescaping.
+ Shared: uri and emailAddress now return a pair of the original
parsed text and the escaped URI (in the latter case, with
the mailto: prefix).
+ HTML reader: unsanitaryURI has been modified to allow unicode
high characters in a URI.
+ Readers: All link and image URIs are now escaped using
+ Markdown and RST writers: unescapeURI is used so that URIs
in these formats are human-readable.

* Setup.hs: Don't assume that the build directory is "dist".
Instead, get it from localBuildInfo.

* OpenDocument writer: Use a Map for stTextStyleAttr.
This avoids duplicates (and invalid xml). Resolves Issue #222.

pandoc (

[ John MacFarlane ]

* HTML writer: Fixed error in math writer (with MathML option)
that caused an infinite loop for unparsable MathML.

pandoc (1.5)

[ John MacFarlane ]

* Added --mathml option. When this is selected, pandoc will convert
TeX math into MathML.
+ Added data/MathMLinHTML.js, which is included when no URL is
provided for --mathml. This allows MathML to be displayed (in
better browsers) as text/html.
+ Removed Text.Pandoc.LaTeXMathML. The module was no longer
necessary; it was replaced by two lines in pandoc.hs.
+ Replaced LaTeXMathML.js.comment and LaTeXMathML.js.packed with a
single combined file, LaTeXMathML.js.

* Added --data-dir option.
This specifies a user data directory. If not specified, will default
to ~/.pandoc on unix or Application Data\pandoc on Windows.
Files placed in the user data directory will override system default
data files.

* Added Maybe datadir parameter to readDataFile, saveOpenDocumentAsODT,
latexMathMLScript, s5HeaderIncludes, and getDefaultTemplate. If
Nothing, no user directory is searched for an override.

* Added 'plain' output format. This is similar to markdown, but
removes links, pictures, inline formatting, and most anything that
looks even vaguely markupish. The function writePlain is exported by
Text.Pandoc.Writers.Markdown, with which it shares most of its code.

* Allow multi-line titles and authors in meta block.
Titles may span multiple lines, provided continuation lines
begin with a space character. Separate authors may be put on
multiple lines, provided each line after the first begins with
a space character. Each author must fit on one line. Multiple
authors on a single line may still be separated by a semicolon.
Based on a patch by Justin Bogner.

* When given an absolute URI as parameter, pandoc will try to fetch
the content via HTTP. So you can do:
'pandoc -r html -w markdown'
Adds dependency on HTTP.

* Made HTML reader much more forgiving.
+ Incorporated idea (from HXT) that an element can be closed
by an open tag for another element.
+ Javascript is partially parsed to make sure that a <script>
section is not closed by a </script> in a comment or string.
+ More lenient non-quoted attribute values.
Now we accept anything but a space character, quote, or <>.
This helps in parsing e.g.!
+ Bare & signs are now parsed as a string. This is a common
HTML mistake.
+ Skip a bare < in malformed HTML.

* Removed html2markdown and hsmarkdown.
+ html2markdown is no longer needed, since you can now pass URI
arguments to pandoc and directly convert web pages. (Note,
however, that pandoc assumes the pages are UTF8. html2markdown
made an attempt to guess the encoding and convert them.)
+ hsmarkdown is pointless -- a large executable that could be
replaced by 'pandoc --strict'.

* In most writers, an image in a paragraph by itself is now rendered
as a figure, with the alt text as the caption. (Texinfo, HTML, RST,
MediaWiki, Docbook, LaTeX, ConTeXt, HTML.) Other images are
rendered inline.

* Depend on extensible-exceptions. This allows pandoc to be compiled
on GHC 6.8.

* Added --base-header-level option. For example, --base-header-level=2
will change level 1 headers to level 2, level 2 to level 3, etc.
Closes Debian #563416.

* Incomplete support for RST tables (simple and grid).
Thanks to Eric Kow. Colspans and rowspans not yet supported.

* Added accessors (docTitle, docAuthors, docDate) to Meta type.

* MediaWiki writer: format links with relative URLs as wikilinks.
The new rule: If the link target is an absolute URL, an external
link is created. Otherwise, a wikilink is created.

* Text.Pandoc.Shared: Export uniqueIdent, and don't allow tilde in
identifier. Note: This may break links to sections that involve

* Markdown(+lhs) reader: handle "inverse bird tracks."
Inverse bird tracks (<) are used for haskell example code that is not
part of the literate Haskell program. Resolves Issue #211.

* LaTeX reader:
+ Recognize '\ ' (interword space).
+ Recognize nonbreaking space '~'.
+ Ignore \section, \pdfannot, \pdfstringdef. Ignore alt title in
section headers. Don't treat \section as inline LaTeX.
Resolves Issue #202.
+ LaTeX reader: allow any special character to be escaped.
Resolves Issue #221.
+ LaTeX reader: treat \paragraph and \subparagraph as level 4, 5
headers. Resolves Issue #207.

* Use template variables for include-before/after.
+ These options now imply -s; previously they worked also in fragment
+ Users can now adjust position of include-before and include-after
text in the templates.
+ Default position of include-before moved back (as it was before 1.4)
before table of contents.
+ Resolves Issue #217.

* Don't print an empty table header: (all writers).
Resolves Issue #210.

* HTML, Docbook writer: Use tbody, thead, and cols in tables.

* HTML writer: Don't include TOC div if table of contents is empty.

* Markdown writer: Fixed citations.
Previously the markdown writer printed raw citation codes, e.g.
[geach1970], rather than the expanded citations provided by
citeproc, e.g. (Geach 1970). Now it prints the expanded citations.
This means that the document produced can be processed as a markdown
document without citeproc. Thanks to dsanson for reporting, and
Andrea Rossato for the patch.

* Improved and simplified title block in context template.
Previously it caused an error if there was no title.
This method should also be easier for users to customize.

* Markdown reader:
+ Treat p., pp., sec., ch., as abbreviations in smart mode.
+ Disallow blank lines in inline code span.
+ Allow footnotes to be indented < 4 spaces.
This fixes a regression. A test case has been added.
+ Escape spaces in URLs as %20. Previously they were incorrectly
escaped as +, which is appropriate only for the query part of
a URL. Resolves Issue #220.
+ Require two spaces after capital letter + period for list item.
Otherwise "E. coli" starts a list. This might change the semantics
of some existing documents, since previously the two-space
requirement was only enforced when the second word started
with a capital letter. But it is consistent with the existing
documentation and follows the principle of least surprise.
Resolves Issue #212.

* LaTeX template: redefine labelwidth when using enumerate package.
Otherwise the list labels (numbers) often extend past the left
margin, which looks bad.

* Mediawiki writer: Don't print a "== Notes ==" header before
references. This is too English-centric. Writers can provide their
own header at the end of the document.

* Promoted mediawiki headers. '= head =' is now level 1, '== head =='
level 2, etc. This seems to be correct; it's only by convention
that wikipedia articles have level 2 headers at most.
Patch due to Eric Kow.

* RunTests.hs: Set LANG to a UTF-8 locale. Use 'pandoc --data-dir=' so
data files don't need to have been installed. This removes the need to
set HOME.

* HTML reader:
+ Handle spaces before <html>. Resolves Issue #216.
+ Be forgiving in parsing a bare list within a list.
The following is not valid xhtml, but the intent is clear:
We'll treat the <ol> as if it's in a <li>. Resolves Issue #215.

* Updated INSTALL instructions. cabal method is now promoted.

* Updated markdown2pdf man page. It no longer says all pandoc options
are accepted.

* README/man pages: Removed advice to pipe through tidy before HTML
reader. This is obsolete, now that we have a forgiving HTML parser.

* LaTeX writer: set numbersections template variable, so
the section numbering options work again.

* Removed obsolete Makefile.

* Website: renamed -> index.txt.

* New batch file to make-windows-installer.
+ Removed old
+ Added make-windows-installer.bat
+ Modified default installer name in pandoc-setup.iss

* Removed freebsd and macports directories.
They are no longer up to date.

* Setup.hs:
+ Made man page building sensitive to build verbosity.
+ Improved detection of highlighting support in test hook.
+ Install wrapper scripts into cabal bin directory.
+ Also simplified installManpages.
+ Setup.hs: install manpages to mandir. Code borrowed from darcs.

* Changed default of writerXeTeX to False.

* HTML writer: don't include empty UL if --toc but no sections.
Resolves Issue #199.

* LaTeX writer:

+ If book, report, or memoir documentclass, use \chapter{}
for first-level headers. Otherwise use \section{}.
+ Removed stLink, link template variable. Reason: we now always
include hyperref in the template.

* LaTeX template:

+ Only show \author if there are some.
+ Always include hyperref package. It is used not just for links but
for toc, section heading bookmarks, footnotes, etc. Also added
unicode=true on hyperref options.

* markdown2pdf: always do at least two runs. hyperref bookmarks
require this.

* cabal file: Removed unneeded dependency on template-haskell.

* Windows installer - fixed bug in data file locations.
Resolves Issue #197.

* Deprecated --custom-header in documentation.
Removed old "Custom Headers" section in README.

pandoc (1.4)

[ John MacFarlane ]

* Pandoc will now compile with either GHC 6.10 or 6.12.
+ Don't use System.IO.UTF8 when compiling with 6.12
+ Use -fno-warn-unused-do-bind option when compiling with 6.12

* Replaced old headers with templates. Now users have much more
control over the way documents appear in --standalone mode,
and writer code is simplified. Resolves Issues #59, 147.
Every effort has been made to retain backwards compatibility.
So, the --custom-header option should still work as before.

+ Added Text.Pandoc.Templates. This provides functions for
retrieving default templates and for rendering templates.
+ System templates (in the pandoc data directory) can be
overridden by user templates in $HOME/.pandoc/templates.
+ Removed Text.Pandoc.DefaultHeaders.
+ Removed data/headers directory.
+ Added templates directory.
+ Added writerTemplate and writerVariables fields to WriterOptions.
+ Removed writerTitlePrefix, writerHeader fields from WriterOptions.
+ Changed --print-default-header to --print-default-template.
+ Added --template option.
+ Added -V/--variable option to set custom template variables.

* Pandoc no longer requires Template Haskell. Resolves Issue #186.

+ Removed need for TH in ODT module. Instead get reference.odt from
data file at run time.
+ Removed TH dependency from S5 module. S5 module now exports
s5HeaderIncludes, which pandoc.hs includes if writer is s5 and
+ Refactored LaTeXMathML not to use TH.

* Meta is now Meta [Inline] [[Inline]] [Inline] rather than
Meta [Inline] [String] String. Authors and date in Meta are now lists
of Inline elements rather than raw strings. This means that they can
be formatted and can include footnotes. NOTE: This may be a breaking
change for those using pandoc as a library.

* Added readDataFile to Text.Pandoc.Shared. This retrieves
a data file from the user pandoc data directory (~/.pandoc
on unix), or, if not found there, from the system data
directory ($CABALDIR/shared/pandoc-VERSION/). All data
files, including templates, LaTeXMathML.js, s5 styles,
and reference.odt, can be overridden by the user.

* s5 files moved from data/ui/default to s5/default.

* Use unicode instead of entities in HTML and XML output. Resolves
Issue #163.

* Prettier HTML footnote references: put anchor inside sup,
instead of other way. Resolves Issue #191. Thanks to

* Added --xetex option to pandoc and markdown2pdf.
If --xetex is specified, pandoc produces latex suitable for
processing by xelatex, and markdown2pdf uses xelatex to create
the PDF. Resolves Issue #185.

* RTF writer: multiple authors now occupy multiple paragraphs rather
than using a line break.

* Man writer: now the "--after-body" will come after the "AUTHORS"
section, whereas before it would come before it. This is a
slight break from backwards compatibility.

* Added --reference-odt option, so users may customize the styles
used in pandoc-generated ODT files. Users may also place a
default reference.odt in the ~\.pandoc directory.

* ODT writer:
+ Indented and line-broke styles.xml so it can be modified more easily.
+ Omitted some unnecessary style declarations.
+ Don't wrap text in OpenDocument writer. The tags are too long, making
wrapping ugly and pointless.

* LaTeX reader: use \\ to separate multiple authors.

* Markdown reader: use ; as separator between authors.
This allows you to use ',' within author names: e.g. "John Jones, Jr."

* S5 writer: use linebreak to separate authors in title page.

* RST reader: Allow :: before lhs code block. The RST spec requires the
:: before verbatim blocks. This :: should not be treated as literal
colons. Resolves Issue #189.

* Documented pandoc 1.3's new definition list syntax in README.
(An oversight in the last release.)

* markdown2pdf.hs:
+ interpret ! in a log as an error line.
+ --toc now works properly.

* Changes in RunTests.hs:
+ Use the Diff library rather than a local copy of Diff.hs.
(This vastly increases performance.) This change means that 'cabal
test' presupposes that the Diff library is installed.
+ Removed tests/Diff.hs from cabal file.
+ Changed RunTests to use local environment. We need at least HOME, so
pandoc can find its data directory.

* Updated windows installer to install data files in the app directory.

* Windows installer now installs portable wrappers hsmarkdown and

pandoc (1.3)

[ John MacFarlane ]

* Added --id-prefix option (Issue #41). This adds a prefix to all
automatically generated HTML identifiers, which helps prevent
duplicate identifiers when you're generating a fragment (say a blog

* Added --indented-code-classes option. This specifies classes
to use for indented code blocks. (Patch due to buttock; Issue #87.)

* --number-sections now affects HTML output as well as ConTeXt and LaTeX
(Issue #150).

* Improved syntax for markdown definition lists (Issue #24).
Definition lists are now more compatible with PHP Markdown Extra.
+ You can have multiple definitions for a term (but still not
multiple terms).
+ Multi-block definitions no longer need a column before each block
(indeed, this will now cause multiple definitions).
+ The marker no longer needs to be flush with the left margin,
but can be indented at or two spaces. Also, ~ as well as :
can be used as the marker (this suggestion due to David
+ There can now be a blank line between the term and the

* Better looking simple tables. Resolves Issue #180.
+ Markdown reader: simple tables are now given column widths of 0.
+ Column width of 0 is interpreted as meaning: use default column width.
+ Writers now include explicit column width information only
for multiline tables. (Exception: RTF writer, which requires
column widths. In this case, columns are given equal widths,
adding up to the text width.)
+ Simple tables should now look better in most output formats.

* Allow markdown tables without headers (Issue #50).
The new syntax is described in README. Also allow optional line of
dashes at bottom of simple tables.

* Compensate for width of final table column (Issue #144).

* Treat a backslash followed by a newline as a hard line break
in markdown. Resolves Issue #154. This is a nice alternative
to markdown's "invisible" way of indicating hardline breaks
using lines that end with two spaces.

* Improved performance of markdown reader by ~10% by eliminating the
need for a separate parsing pass for notes. Raw notes are now stored
on the first pass (which parses references), then parsed when the
note is inserted into the AST. The stateNotes field in ParserState
is now a list of [(String, String)] pairs instead of [(String,

* In markdown reader, treat 4 or more * or _ in a row as literal
text. (Trying to parse long strings of * or _ as strong or emph
leads to exponential performance problems.)

* Markdown reader: Use + rather than %20 for spaces in URLs.

* Fixed htmlComment parser, adding a needed 'try'.

* Don't print raw HTML in man output.

* Allow . _ and ~ in header identifiers.

* Specially mark code blocks that were "literate" in the input.
They can then be treated differently in the writers. This allows
authors to distinguish bits of the literate program they are writing
from source code examples, even if the examples are marked as
Haskell for highlighting. (Issue #174.)

* Modified html+lhs output to use "haskell" highlighter instead
of "literateHaskell". The highlighting module now adds bird tracks
after highlighting (for HTML output), if the code block has the
"literate" class. This gives better results, because kate's
haskell highlighter is much better than the literateHaskell

* Fixed handling of footnotes in titles (HTML) and headers (LaTeX).
(Issue #137.)

* Support for "..code-block" directive in RST reader. Not core
RST, but used in Sphinx for code blocks annotated with syntax
information. Thanks to Luke Plant for the patch.

* Added "head" to list of block-level HTML tags. Resolves
Issue #108.

* Added stripTags to Text.Pandoc.XML. This is used in the HTML writer.

* Set utf-8 encoding in texinfo headers.

* Docbook writer: add ids to sections. Use link for internal links.
(Issue #60.)

* Blank lines after lists in MediaWiki writer.

* Properly handle commented-out list items in markdown.
Resolves Issue #142. Example:

- a
- b
- c

* Changed heuristic in compactify. compactify has to decide whether a
Para that ends a list is a Para intentionally, or just because of
the blank lines at the end of every list. In the latter case the
Para is turned to a Plain. The old heuristic was: change final Para
to Plain iff the other items all end in Plain. This produces bad
results when, for example, an item contains just a Plain and an HTML
comment, as it does in the list above. The new heuristic: change
final Para to Plain iff the other items don't contain a Para.

* Added % as an rst underline character. Resolves Issue #173.

* Fix inline math parser so that \$ is allowed in math.
Resolves Issue #169.

* Translate \int (integral) into unicode when using unicode math
method. Resolves Issue #177.

* markdown2pdf.hs improvements:
+ Use System.IO.UTF8.
+ Print error messages on last attempt.
+ Do not create a backup when overwriting a PDF (Issue #166).
+ Accept --longopt=val options.
+ Added man/man1/markdown2pdf.1 to extra-tmp-files in cabal, so that
it is properly cleaned.

* Added haddock comments warning that readers assume \n line endings.

* Updated COPYRIGHT file.

* Makefile: Changed EXECSBASE so it doesn't pull in hsmarkdown &
markdown2pdf. Otherwise strip tries to strip shell scripts when you
install using 'make'.

* Changed Makefile so it doesn't build Haskell wrappers.

* Fixed Makefile so it doesn't try to build man pages in build-doc.

* Install pcre3.dll in Windows install script; this allows us to
package a version of pandoc with highlighting support.

pandoc (1.2.1)

[ John MacFarlane ]

* Fixed regression with --preserveTabs. Brought back optPreserveTabs.
The trick of setting tabStop to 0 to mean "preserve tabs" had a bad
side effect: strings of 0 spaces were interpreted as indentation.
So, with --preserve-tabs, unindented paragraphs were treated as
code. Resolves Issue #138.

* HTML writer: wrap sections in divs. Resolves Issue #70.

+ hierarchicalize has been rationalized; it builds a hierarchical
representation of the document from the headers, and simultaneously
gives each section a unique identifier based on the heading title.
+ Identifiers are now attached to the divs rather than
to the headers themselves.
+ Table of content backlinks go to the beginning of the table, rather
than to the section reference that was clicked.
+ Code for constructing identifiers has been moved to Text.Pandoc.Shared
from the HTML writer, since it is now consumed only by
+ In --strict mode, pandoc just prints bare headings, as before
(unless --toc has been specified).
+ In s5 output, it does not wrap sections in divs, as that seems to
confuse the s5 javascript.

* Man writer: break lines at end of each sentence. groff expects this
and treats '.' and '?' differently when followed by line ending as
opposed to ordinary space. Also, don't escape periods. Instead, use
zero-width character \& to avoid unwanted interpretation of periods
at start of line. Resolves Issue #148.

* Markdown writer: Added '#' and '>' to list of characters to be
escaped in markdown output. Removed '<', as it is not an officially
escapable character. This partially resolves Issue #96.

* Make --smart the default for man output format. Otherwise we have
trouble dividing lists of endlines into sentences.

* DocBook writer: Use language attribute to indicate source language
in code blocks.

* RST reader:

+ Allow # to continue list, even if the list was started with an
explicit marker. For example:

A. my list
#. continued

Resolves Issue #140.
+ Allow continuation lines in line blocks. Also added test cases for
line blocks for RST reader. Resolves Issue #149.
+ Allow explicit links with spaces in URL: `link <to this>`_

* Improved LaTeX reader's coverage of math modes. Remove displaymath*
(which is not in LaTeX) and recognize all the amsmath environments
that are alternatives to eqnarray, namely equation, equation*,
gather, gather*, gathered, multline, multline*, align, align*,
alignat, alignat*, aligned, alignedat, split. Resolves Issue #103.
Thanks to shreevatsa.public for the patch.

* Markdown reader:

+ Allow -, _, :, . in markdown attribute names. These are legal in
XML attribute names.
+ Use non-breaking spaces in abbreviations.
+ Markdown reader: improved efficiency of abbreviation parsing.
Instead of a separate abbrev parser, we just check for
abbreviations each time we parse a string. This gives a huge
performance boost with -S. Resolves Issue #141.

* Improved efficiency of shared parsers: hexNum, htmlComment,
whitespace, indentSpaces.

* Export HTMLMathMethod in Text.Pandoc.

* Export languagesByExtension in Text.Pandoc.Highlighting.

* Added new Haskell version of markdown2pdf, due to
Paulo Tanimoto. This should be more portable than the old
shell script.

* Made 'pandoc -v' more explicit about compiler options.
Resolves Issue #139.

* pandoc.hs: Made --strict compatible with --standalone, --toc.

* Use Paths_pandoc to get version number, instead of hard-coding it
into Text/Pandoc.hs.

pandoc (1.2)

[ John MacFarlane ]

* Added support for literate Haskell. lhs support is triggered by
'+lhs' suffixes in formats. For example, 'latex+lhs' is literate
Haskell LaTeX. '.lhs' files are treated by default as literate

+ Added stateLiterateHaskell to parser state.
+ Added parser for lhsCodeBlock to Markdown, RST, LaTeX readers.
+ Added parser for |inline lhs| to LaTeX reader.
+ Added writerLiterateHaskell to WriterOptions.
+ Added lhs support to Markdown, RST, LaTeX, HTML writers.
+ Added definition of code environment to LaTeX header.
+ Added tests (run only if highlighting support compiled in).
+ Documented lhs features in man page and README.

* In Text.Pandoc.Definition, added processWith, processWithM,
and queryWith, and deprecated processPandoc and queryPandoc
for these more general functions.

* Fixed bug in mediawiki writer: improper closing tags in tables.
Thanks to Benct Philip Jonsson for reporting the bug.

* Added --email-obfuscation option.

+ Added writer option for email obfuscation.
+ Implemented email obfuscation options in HTML writer.
+ Added option to option parser.
+ Documented in README and pandoc man page.
+ Resolves Issue #97.

* LaTeX writer: fixed bug with empty table cells.
Resolves Issue #107. Thanks to rodja.trappe for the patch.

* Fixed bug with header spacing in Markdown and RST writers.
A null header (Meta [] [] []) should not cause a blank line
at the beginning of output. But a blank line is needed between
a non-null header and the main text.

* Markdown reader: Relax spacing rules for $$ in display math. Now
space and newlines are allowed after the opening $$ and before the
closing $$. However, the display math cannot contain an entirely
blank line. Resolves Issue #105.

* Markdown reader: Gobble space after Plain blocks containing only
raw html inline. Otherwise following header blocks are not parsed
correctly, since the parser sees blank space before them. Resolves
Issue #124.

* Markdown reader: Allow " as well as '' to end a latex double-quote.

* Conditionally depend on syb and base >= 4 if ghc >= 6.10.
Resolves Issue #109.

* Fixed problems in RST and markdown output due to bug in pretty-

+ Added hang' function to Text.Pandoc.Shared; this will be used instead
of hang, which doesn't work properly in pretty- When pretty
is upgraded, we can go back to hang.
+ Use hang' (and some different techniques) in RST and markdown writers.
Some output is now a bit different.

* Brought citeproc support up to date for citeproc-hs-0.2.
(Patch by Andrea Rossato.)

* Moved all haskell source to src subdirectory. Renamed Main.hs to

* Rewrote hsmarkdown in Haskell for portability (src/hsmarkdown.hs).
For now, keeping the old shell script too.

* Added TemplateHaskell to Extensions for executable, removed
-threaded for library. Thanks to duncan.coutts for the bug report.
Resolves Issue #121.

* Moved some Extra-Source-Files to Data-Files.

* Moved tabFilter to Shared.

* In pandoc.hs, removed optPreserveTabs; instead, tabstop of 0 means
preserve tabs.

* Minor code cleanup based on hlint suggestions.

pandoc (1.1)

[ John MacFarlane ]

* Main.hs:

+ Changed date on copyright message in Main.hs.
+ Have the '-v' option print syntax highlighting languages
separated by commas, and wrapped in lines, instead of in five
columns as before.

* Added --jsmath option. Resolves Issue #68.

+ Added --jsmath option to Main.hs
+ Added JsMath to HTMLMathMethod in Text.Pandoc.Shared.
+ Handle math appropriately in HTML writer when JsMath selected.
+ Documented the option in README and man page.

* Text.Pandoc.Shared: Changed compactify to use a better heuristic
for tight and loose lists. Final Para is changed to Plain if all
other list items *end* with a Plain block. Addresses Issue #99.

* HTML reader:

+ Added colons to protocols in unsanitaryURI. Closes Issue #88.
+ HTML reader: Don't interpret contents of <pre> blocks as markdown.
Added rawVerbatimBlock parser. Resolves Issue #94.

* Markdown reader:

+ Allow URLs with spaces in them in links and references, but escape
them as "%20".
+ Allow blank space at the end of horizontal rules.

* RST reader: Modified 'unknownDirective' parser to handle comment
blocks correctly, and added tests for comment blocks. Resolves Issue
#86. Closes Debian Bug #500662.

* HTML writer:

+ Include classes on tr elements in HTML output:
"header", "odd", "even". This allows tables to be styled with
lines in alternating colors. Resolves Issue #91.
+ Enclose all LaTeXMathML bits in <span class="LaTeX">.
This prevents parts of the document that are not math from being
interpreted as math by LaTeXMathML.js.

* OpenDocument and ODT writers: Added support for HorizontalRule elements,
which were formerly ignored. Resolves Issue #95.

* Text.Pandoc.Shared: Modified wrappedTeX to eliminate the line break
between a footnote and immediately following nonspace characters in
LaTeX and ConTeXt output. (This gets interpreted as a space, which
is not desired in cases like `text^[note]---`.) Resolves Issue #93.

* Windows installer: Don't require admin privileges to run
installer. Modified pandoc-setup.iss, and changed modpath.iss to
modify HKCU path if user lacks admin privileges. Also fixed case
where oldpath is empty (previously this led to the new path
beginning with a semicolon).

* Updated INSTALL instructions for Arch packages and OS X install using

* Removed the (now unneeded) debian directory.
Removed empty Codec and System directories.

* Moved odt-styles/ to data/. Removed unneeded variable in Makefile.

* Modified Setup.hs so that the "test" target returns an error status
when tests fail, and "build" returns a success status if
the build succeeds. Resolves Issue #100.

* Added BUGS to files in tarball.

pandoc (

[ John MacFarlane ]

* Removed spurious reference to pdf output format from pandoc(1) man page.

pandoc (1.0)

[ Andrea Rossato ]

* Added new OpenDocument writer.

* Added support for SmallCaps inline element.

* Added support for integrating pandoc with citeproc-hs.

+ Added Cite element to definition and writers.
+ Added Text.Pandoc.Biblio module
+ Note: This support is included only if the 'citeproc'
Cabal configuration flag is set.

* Made Pandoc data structure an instance of Typeable.
Added new processPandoc and queryPandoc functions, to query
or transform matching elements in a Pandoc structure.

[ Peter Wang ]

* Added new Texinfo writer.

[ John MacFarlane ]

* Changes to Texinfo writer:

+ No space between paragraph and following @verbatim (provides more
pleasing appearance in text formats)
+ Blank line consistently after list environments.
+ Removed deVerb.
+ Use @code instead of @verb for inline code (this solves the character
escaping problem for texi2dvi and texi2pdf).
+ Added news of Texinfo writer to README.
+ Added Texinfo to list of formats in man page, and removed extra 'groff'.
+ Added texi & texinfo extensions to Main.hs, and fixed bug in determining
default output extension.
+ Modified disallowedInNode in Texinfo writer to correct list of disallowed characters.

* Added tests for OpenDocument writer.

* Added ODT writer (using zip-archive library to package output of
OpenDocument writer). Added odt-styles directory with default ODT styles.

* Added new mediawiki writer and tests.

* Markdown reader: Added support for delimited code blocks, with optional
syntax highlighting using highlighting-kate (if the 'highlighting'
configuration option is selected).

+ Currently highlighting is supported only in the HTML writer.
+ Delimited code blocks can have attributes; using the language name as
class triggers highlighting.
+ New Attributes parameter in CodeBlock structure.
+ --version now indicates whether syntax highlighting support is compiled
in, and prints a list of supported languages

* Removed debian directory. Pandoc is no longer a native debian package.

* Changes to build process: pandoc can now be built from the repository
using Cabal. No unix tools are needed (so, pandoc can be built on Windows
without Cygwin).

+ Include shell scripts themselves in repo, rather than generating from wrappers.
Removed wrappers directory and wrappers Makefile target.
+ Text/Pandoc/ASCIIMathML.hs, Text/Pandoc/DefaultHeaders.hs,
and Text/Pandoc/Writers/S5.hs are no longer built in Makefile
from templates in the templates/ directory. Instead, they use template
haskell to read data at compile time from the relevant files in data/.
Template haskell functions go in a new module, Text.Pandoc.TH.
+ man pages are now generated in Setup.hs hook, not by Makefile
+ Makefile 'tarball' target now calls Cabal's 'sdist'
+ Added "Extra-Source-Files" to pandoc.cabal, so sdist contains everything needed
+ Added "Build-Type" field to pandoc.cabal to avoid warning.
+ Added to "Extra-source-files" and "Extra-tmp-files" in pandoc.cabal,
so 'sdist' and 'clean' will work properly.
+ Setup.hs now generates man pages in a postbuild hook.
+ Added dependency-checking to Setup.hs, so it only rebuilds things
that need rebuilding.
+ Added 'library' and 'executable' configuration flags.
Cabal can now be told to build just the library or just the executable.
+ CABALOPTS may now be specified with 'make' to pass Cabal configuration flags.
For example: CABALOPTS=-fhighlighting make

* Rewrote test suite so it doesn't depend on perl or unix tools.

+ Replaced old with a Haskell script RunTests.hs.
+ Added Diff.hs module to be used by RunTests.hs instead of unix 'diff'.
+ Added test hook to Setup.hs, so tests may be run from cabal.
+ Changed Makefile's 'test' target to run tests via cabal.
+ Removed old
+ Since we no longer have 'sed' to filter out raw HTML sections
from the docbook writer test, or raw LaTeX sections from the
context writer test, we now just include these sections.
They can be taken out if it is necessary to process the files.
+ Updated latex and context writer tests to remove extra spaces
after '\\item'
+ Added a markdown table reader test.
+ Added markdown-reader-more.txt to test suite, for additional test cases
for raw ConTeXt environments and more.

* Compatibility fixes for CPP, Cabal, and haddock:

+ Use CPP in "Extensions" field in pandoc.cabal.
+ Removed use of backslash string continuations in source files.

* Removed pandoc.cabal.ghc66. We now require Cabal >= 1.2, GHC >= 6.8,
base >= 3.

* Require parsec < 3.
The compatibility module in parsec 3.0.0 gives far worse performance than
parsec 2.1. Eventually pandoc will be upgraded to use the new bytestring
version of parsec, and then we'll go to parsec 3.0.0.

* Removed Text.Regex dependencies by rewriting using plain Haskell
(Text.Pandoc.Writers.RTF, Text.Pandoc.Writers.HTML, Main.hs)

* Moved Text.Pandoc.Writers.DefaultHeaders -> Text.Pandoc.DefaultHeaders.

* Makefile:

+ Added 'configure' as dependency of 'uninstall-all'.
(It uses the Cabal build program.)
+ Makefile: only use --with-hc-pkg if GHC_PKG is defined.
Note that Cabal will automatically choose the ghc-pkg appropriate
for the compiler selected, so normally specifying GHC by itself
is sufficient.

* Removed Text.Pandoc.UTF8 module; instead, depend on utf8-string and use
its IO and conversion functions.

* Added -Wall to ghc-options in pandoc.cabal. Cleaned up modules so that
everything is -Wall clean.

+ Added pragma to HTML writer to avoid deprecation warning for use of "start" attribute.
+ Added pragma to Text/Pandoc/Shared.hs to get rid of "orphan instance" warnings.
(These are caused by the Lift instance for ByteString.)

* Changed the comment used to replace unsafe HTML if sanitize-html option

* Made -c/--css option repeatable on the command line (like -H, -A, -B).

* Moved XML-formatting functions to new unexported module Text.Pandoc.XML.

* Escape '\160' as "&#160;", not "&nbsp;" in XML.
"nbsp" isn't a predefined XML entity.

* Fixed bug in RST reader, which would choke on: "p. one\ntwo\n".
Added some try's in ordered list parsers.

* Man writer: don't escape " as \".

* Allow newline before URL in markdown link references. Resolves Issue #81.
Added tests for this issue in new "markdown-reader-more" tests.
Changed RunTests.hs to run these tests.

* Support for display math. Resolves Issue #47.

+ Added a DisplayMath/InlineMath selector to Math inlines.
+ Markdown parser yields DisplayMath for $$...$$.
+ LaTeX parser yields DisplayMath when appropriate. Removed
mathBlock parsers, since the same effect is achieved by the math
inline parsers, now that they handle display math.
+ Writers handle DisplayMath as appropriate for the format.
+ Modified tests accordingly; added new tests for display math.

* Use LaTeXMathML instead of ASCIIMathML. LaTeXMathML is closer
to LaTeX in its display of math, and supports many non-math LaTeX environments.

+ Changed -m option to use LaTeXMathML rather than ASCIIMathML.
+ Modified HTML writer to print raw TeX when LaTeXMathML is
being used instead of suppressing it.
+ Removed ASCIIMathML files from data/ and added LaTeXMathML.
+ Replaced ASCIIMathML with LaTeXMathML in source files.
+ Modified README and pandoc man page source.
+ Added --latexmathml option (kept --asciimathml as a synonym
for backwards compatibility)

* Markdown reader: Parse setext headers before atx headers.
Test case:
# hi
parsed by as an H1 header with contents "# hi".

* Markdown reader: Treat "mixed" lists the same way as does.
The marker on the first list item determines the type of the whole
list. Thus, a list like
1. one
- two
* three
gets parsed as a single ordered list. (Previous versions of pandoc
treated this as an ordered list with an unordered sublist.)

* Markdown smart typography:

+ Em dashes no longer eat surrounding whitespace. Resolves Issue #69.
+ Use nonbreaking spaces after known abbreviations in markdown parser.
Thus, for example, "Mr. Brown" comes out as "Mr.~Brown" in LaTeX, and does
not produce a sentence-separating space. Resolves Issue #75.

* Markdown writer: Print unicode \160 literally, rather than as &nbsp;.

* Treat '\ ' in (extended) markdown as nonbreaking space.
Print nonbreaking space appropriately in each writer (e.g. ~ in LaTeX).

* The '--sanitize-html' option now examines URIs in markdown links
and images, and in HTML href and src attributes. If the URI scheme
is not on a whitelist of safe schemes, it is rejected. The main point
is to prevent cross-site scripting attacks using 'javascript:' URIs.
and Resolves Issue #62.

* HTML writer:

+ Override Text.XHtml's stringToHtml function,
so that characters below 0xff are not converted to numerical entity
references. Also convert '\160' to "&nbsp;". This should aid readability
and editability of the HTML source. It does presuppose that the HTML
will be served as UTF-8.
+ In code blocks, change leading newlines to <br /> tags.
(Some browsers ignore them.) Resolves Issue #71.
+ Use style attributes rather than css classes for strikethrough
and ordered list styles. This works better when fragments, rather than
standalone documents, are generated.

* HTML reader: Count anything that isn't a known block (HTML) tag as an
inline tag (rather than the other way around). Added "html", "head", and
"body" to list of block tags. Resolves Issue #66, allowing
<lj> to count as an inline tag.

* RTF writer: Fixed bug. Extra spaces were being printed after emphasized,
boldface, and other inline elements. Resolves Issue #64.

* LaTeX reader: improvements in raw LaTeX parsing.

+ "loose punctuation" (like {}) parsed as Space
+ Para elements must contain more than Str "" and Space elements
+ Added parser for "\ignore" command used in literate haskell.
+ Reworked unknownCommand and rawLaTeXInline: when not in "parse raw"
mode, these parsers simply strip off the command part and allow
the arguments to be parsed normally. So, for example,
\blorg{\emph{hi}} will be parsed as Emph "hi" rather than
Str "{\\emph{hi}}".
+ Parse lhs "code" environments as verbatim.
Refactored parsers for verbatim environments.
+ Removed specialEnvironment parser.
+ parse '{}', if present, after \textless, \textgreater,
\textbar, \textbackslash, \ldots.
+ Parse unescaped special characters verbatim rather than
changing them to spaces. This way arguments of unknown
commands will appear in braces.

* Parse raw ConTeXt environments as TeX in markdown reader.
Resolves Issue #73.

* Moved BlockWrapper and wrappedBlocksToDoc from ConTeXt writer to Shared.

* Made some structural changes to parsing of raw LaTeX environments.
Previously there was a special block parser for LaTeX environments.
It returned a Para element containing the raw TeX inline. This has
been removed, and the raw LaTeX environment parser is now used in the
rawLaTeXInline parser. The effect is exactly the same, except that we
can now handle consecutive LaTeX and ConTeXt environments not separated
by spaces. This new flexibility is required by the example in
Issue #73:

\placeformula \startformula
L_{1} = L_{2}

API change: The LaTeX reader now exports rawLaTeXEnvironment' (which
returns a string) rather than rawLaTeXEnvironment (which returns a block
element). This is more likely to be useful in other applications.

* Use \textsubscr instead of \textsubscript for LaTeX subscript macro.
\textsubscript conflicts with a definition in the memoir class.
Resolves Issue #65.

* Removed unneeded space after "\\item" in LaTeX and ConTeXt output.

* Added amsmath package to default LaTeX header. Resolves Issue #48.

* Added \setupitemize[autointro] to ConTeXt header, to prevent orphaned
list introduction lines.

* Changed Float to Double in definition of Table element.
(Double is more efficient in GHC.)

* Fixed bug in Markdown parser: regular $s triggering math mode.
For example: "shoes ($20) and socks ($5)."
The fix consists in two new restrictions:

+ the $ that ends a math span may not be directly followed by a digit.
+ no blank lines may be included within a math span.

Thanks to Joseph Reagle for noticing the bug.

* Use Data.List's 'intercalate' instead of custom 'joinWithSep'.
Removed 'joinWithSep' from Text.Pandoc.Shared.

* Updated README and man pages. Acknowledge contributors in README.
Added paragraph to README about producing S5 with separate CSS/javascript.

* Updated INSTALL to reflect new build system (including configuration
options) and document new dependencies. Added note to INSTALL that
Cabal >= 1.2 is required for build. Resolves Issue #74.

* Fixed some haddock documentation errors.

* Small fix to markdown2pdf man page: only input needs to be piped through iconv.

pandoc (0.46) unstable; urgency=low

[ John MacFarlane ]

* Made -H, -A, and -B options cumulative: if they are specified
multiple times, multiple files will be included.

* Added optional HTML sanitization using a whitelist.
When this option is specified (--sanitize-html on the command line),
unsafe HTML tags will be replaced by HTML comments, and unsafe HTML
attributes will be removed. This option should be especially useful
for those who want to use pandoc libraries in web applications, where
users will provide the input.

+ Main.hs: Added --sanitize-html option.

+ Text.Pandoc.Shared: Added stateSanitizeHTML to ParserState.

+ Text.Pandoc.Readers.HTML:
- Added whitelists of sanitaryTags and sanitaryAttributes.
- Added parsers to check these lists (and state) to see if a given
tag or attribute should be counted unsafe.
- Modified anyHtmlTag and anyHtmlEndTag to replace unsafe tags
with comments.
- Modified htmlAttribute to remove unsafe attributes.
- Modified htmlScript and htmlStyle to remove these elements if

+ Modified README and man pages to document new option.

* Improved handling of email addresses in markdown and reStructuredText.
Consolidated uri and email address parsers. (Resolves Issue #37.)

+ New emailAddress and uri parsers in Text.Pandoc.Shared.
- uri parser uses parseURI from Network.URI.
- emailAddress parser properly handles email addresses with periods
in them.

+ Removed uri and emailAddress parsers from Text.Pandoc.Readers.RST
and Text.Pandoc.Readers.Markdown.

* Markdown reader:

+ Fixed emph parser so that "*hi **there***" is parsed as a Strong
nested in an Emph. (A '*' is only recognized as the end of the
emphasis if it's not the beginning of a strong emphasis.)

+ Moved blockQuote parser before list parsers for performance.

+ Modified 'source' parser to allow backslash-escapes in URLs.
So, for example, [my](/url\(1\)) yields a link to /url(1).
Resolves Issue #34.

+ Disallowed links within links. (Resolves Issue #35.)
- Replaced inlinesInBalanced with inlinesInBalancedBrackets, which
instead of hard-coding the inline parser takes an inline parser
as a parameter.
- Modified reference and inlineNote to use inlinesInBalancedBrackets.
- Removed unneeded inlineString function.
- Added inlineNonLink parser, which is now used in the definition of
- Added inlineParsers list and redefined inline and inlineNonLink parsers
in terms of it.
- Added failIfLink parser.

+ Better handling of parentheses in URLs and quotation marks in titles.
- 'source' parser first tries to parse URL with balanced parentheses;
if that doesn't work, it tries to parse everything beginning with
'(' and ending with ')'.
- source parser now uses an auxiliary function source'.
- linkTitle parser simplified and improved, under assumption that it
will be called in context of source'.

+ Make 'block' conditional on strictness state, instead of using
failIfStrict in block parsers. Use a different ordering of parsers
in strict mode (raw HTML block before paragraph) for performance.
In non-strict mode use rawHtmlBlocks instead of htmlBlock.
Simplified htmlBlock, since we know it's only called in strict

+ Improved handling of raw HTML. (Resolves Issue #36.)
- Tags that can be either block or inline (e.g. <ins>) should
be treated as block when appropriate and as inline when
appropriate. Thus, for example,
should be treated as a paragraph with inline <ins> tags, while
should be treated as a paragraph within <ins> tags.
- Moved htmlBlock after para in list of block parsers. This ensures
that tags that can be either block or inline get parsed as inline
when appropriate.
- Modified rawHtmlInline' so that block elements aren't treated as
- Modified para parser so that paragraphs containing only HTML tags and
blank space are not allowed. Treat these as raw HTML blocks instead.

+ Fixed bug wherein HTML preceding a code block could cause it to
be parsed as a paragraph. The problem is that the HTML parser
used to eat all blank space after an HTML block, including the
indentation of the code block. (Resolves Issue #39.)
- In Text.Pandoc.Readers.HTML, removed parsing of following space
from rawHtmlBlock.
- In Text.Pandoc.Readers.Markdown, modified rawHtmlBlocks so that
indentation is eaten *only* on the first line after the HTML
block. This means that in
the foo won't be treated as a code block, but in


it will. This seems the right approach for least surprise.

* RST reader:

+ Fixed bug in parsing explicit links (resolves Issue #44).
The problem was that we were looking for inlines until a '<' character
signaled the start of the URL; so, if you hit a reference-style link,
it would keep looking til the end of the document. Fix: change
inline => (notFollowedBy (char '`') >> inline). Note that this won't
allow code inlines in links, but these aren't allowed in resT anyway.

+ Cleaned up parsing of reference names in key blocks and links.
Allow nonquoted reference links to contain isolated '.', '-', '_', so
so that strings like 'a_b_' count as links.

+ Removed unnecessary check for following link in str.
This is unnecessary now that link is above str in the definition of

* HTML reader:

+ Modified rawHtmlBlock so it parses </html> and </body> tags.
This allows these tags to be handled correctly in Markdown.
HTML reader now uses rawHtmlBlock', which excludes </html> and </body>,
since these are handled in parseHtml. (Resolves Issue #38.)

+ Fixed bug (emph parser was looking for `<IT>` tag, not `<I>`).

+ Don't interpret contents of style tags as markdown.
(Resolves Issue #40.)
- Added htmlStyle, analogous to htmlScript.
- Use htmlStyle in htmlBlockElement and rawHtmlInline.
- Moved "script" from the list of tags that can be either block or
inline to the list of block tags.

+ Modified rawHtmlBlock to use anyHtmlBlockTag instead of anyHtmlTag
and anyHtmlEndTag. This fixes a bug in markdown parsing, where
inline tags would be included in raw HTML blocks.

+ Modified anyHtmlBlockTag to test for (not inline) rather than
directly for block. This allows us to handle e.g. docbook in
the markdown reader.

* LaTeX reader: Properly recognize --parse-raw in rawLaTeXInline.
Updated LaTeX reader test to use --parse-raw.

* HTML writer:

+ Modified rules for automatic HTML header identifiers to
ensure that identifiers begin with an alphabetic character.
The new rules are described in README. (Resolves Issue #33.)

+ Changed handling of titles in HTML writer so you don't get
"titleprefix - " followed by nothing.

* ConTeXt writer: Use wrappers around Doc elements to ensure proper
spacing. Each block element is wrapped with either Pad or Reg.
Pad'ed elements are guaranteed to have a blank line in between.

* RST writer:

+ Refactored RST writer to use a record instead of a tuple for state,
and to include options in state so it doesn't need to be passed as
a parameter.

+ Use an interpreted text role to render math in restructuredText.
See for the

[ Recai Oktaş ]

* Debian packaging changes:

+ Remove the empty 'include' directory in -dev package, which lintian
complains about.
+ Bump Standarts-Version to 3.7.3.
+ Use new 'Homepage:' field to specify the upstream URL on suggestion of

-- Recai Oktaş <> Tue, 08 Jan 2008 05:13:31 +0200

pandoc (0.45) unstable; urgency=low

[ John MacFarlane ]

* Simplified parsing of reference keys and notes in markdown and RST
readers: The Reference data structure from Text.Pandoc.Shared is no
longer needed, since referenceKey and noteBlock parses return strings
(as many blank lines as were occupied by the key or note) and update
state themselves. getPosition and setPosition are now used to ensure
that error messages will give the correct line number. This yields
cleaner (and slightly faster) code, with more accurate parsing error

* Added new Math inline element:

+ Markdown and LaTeX readers now convert TeX math into Math elements,
not TeX.
+ This allows math to be treated differently from raw TeX in output.
TeX elements are no longer printed in output formats other than
Markdown, LaTeX, and ConTeXt. But Math elements are always printed.

* New default handling of math in writers:

+ New module Text.Pandoc.Readers.TeXMath exports readTeXMath, which
parses raw TeX math and outputs a string of Pandoc inlines that
tries to render it as far as possible using unicode characters,
lapsing into literal TeX when needed.
+ readTeXMath is now used for default HTML output in HTML, S5, RTF,
and Docbook, if no other method for displaying math in HTML is
specified. Enclosing $'s are no longer printed by default.
+ By default, math is put inside `<span class="math">`. This way it can be
distinguished from the surrounding text, e.g. put in a different

* New --gladtex and --mimetex options for display of math in HTML:

+ If --gladtex is specified, math is output between `<eq>` tags, so
it can be processed by gladTeX.
+ If --mimetex is specified, math is put in `<img>` tags with a link
to the mimetex CGI script (or any other script that takes TeX math
as input and outputs an image). The URL of the script may be
specified, but defaults to /cgi-bin/mimetex.cgi.
+ HTMLMathMethod structure in WriterOptions keeps track of how to
display math in HTML output.
+ Updated README with a description of the four options for displaying
math in HTML.

* HTML reader:

+ Fixed bug: parser for minimized attributes should not swallow
trailing spaces.
+ Simplified HTML attribute parsing.
+ Changed parsing of code blocks in HTML reader: `<code>` tag is no
longer needed. `<pre>` suffices. All HTML tags in the code block
(e.g. for syntax highlighting) are skipped, because they are not
portable to other output formats. A `<code>...</code>` block not
surrounded by `<pre>` now counts as inline HTML, not a code block.
+ Remove just one leading and one trailing newline from contents of
`<pre>...</pre>` in codeBlock parser.

* Markdown reader:

+ Removed support for box-style block quotes.
+ Require space before title in links and references.
This fixes a bug in parsing URLs like http://silly/url(withparen).
+ Improved and simplified setextHeader parser.
+ Fixed logic in smart quote parsing, adding some needed 'try'
+ Fixed smart quote parsing so that unicode characters 8216 and 8217
are recognized as single quotes, and 8220 and 8221 as double quotes.

* RST reader:

+ Fixed bug in parsing of code blocks. Previously a full tab indent
was required, but RST allows code to be indented any amount.
Resolves Issue #27.
+ Allow field lists to be indented.
+ Parse the contents of field lists instead of treating as a raw string.
+ Represent field lists as definition lists instead of blockquotes.
+ Fixed bug in which metadata would be overridden if the document
contained multiple field lists.
+ Parse fields associated with '.. image::' blocks, and use 'alt'
field, if given, for image alt and title attributes.

* LaTeX reader:

+ Modified specialChar so that '"' characters are parsed.
+ Fixed a bug in parsing of \[ \] math blocks (thanks to Mark Kalderon).

* HTML writer:

+ Changes in handling of math (see above).
+ Don't produce HTML for table of contents if there are
no headers. (This would be an empty list, which is invalid XHTML.)

* Markdown writer:

+ Don't print title attribute if title is empty. (This differs from
the behavior of, and agrees with PHP Markdown. But John
Gruber has indicated that he prefers this behavior.) Adjusted test
suite accordingly.
+ Fixed incorrect line wrapping in paragraphs including hard line
breaks. Resolves Issue #25.
+ Fixed bug in markdown writer: If an ordered list item began with
a marker greater than 3 characters in width, and the item took more
than one line, it would appear on the line after the list marker,

My list item.

Now it works as follows:

(12) My list item.

* RST writer

+ Fixed bug in RST writer's handling of ordered lists. Previously,
list items with multiple lines would not always line up with
single-line list items. Now, list items are nested the length of
the list marker + 1. This looks better and ensures that list items
all line up. (Note that list markers are padded to the length of
the longest list marker in the series.)
+ Use 3-space indent for unordered lists.
+ If label for a link reference contains a colon, surround it by `
signs so it won't be interpreted as the end of the link label.

* LaTeX writer:

+ Cleaner output for footnotes. Footnotes now always begin on a new
line, and the final } is on a line by itself only when it needs to
be (i.e. only when the note ends with a Verbatim environment).
+ Added writer options to state, so state doesn't need to be passed as
a parameter.
+ Text wrapping now provided, using wrapTeXIfNeeded.

* ConTeXt writer: many improvements for more idiomatic ConTeXt output
(thanks to Idris Samawi Hamid for suggestions).

+ PrettyPrint module now used for output.
+ Writer options are now in state, so they don't have to be passed as
a parameter.
+ Text wrapping now provided, using wrapTeXIfNeeded.
+ Better treatment of footnotes: footnotes are always on lines by
themselves, and the final } is on a line by itself only when
it needs to be (after \stoptyping).
+ Use \subject, \subsubject, ... or \section, \subsection, ... for headings,
depending on whether --number-sections option is selected.
+ Extra blank line inserted after \stopitemize
+ Use new, "official" definition of blockquote environment. Also, use
blank line after \startblockquote to balance blank line at the end.
+ Both itemized and enumerated lists are now generated using
\start-stopitemize, with appropriate options. Removed definitions
of ltxenum and ltxitem, which are no longer needed. Provided
defaults for itemized lists in the preamble. State keeps track of
ordered list level, so that when default numbering is specified,
the appropriate scheme can be used.
+ Changed \useurl to \useURL.
+ Changed link color from red to blue.
+ Use \subsubsubsubsection etc., since these are supported
(up to at least sub x 5).

* Text.Pandoc.Shared:

+ Save and restore position in parseFromString, so that accurate
error messages can be given.
+ Improved efficiency of romanNumeral parser.
+ Added wrappedTeX and wrapTeXIfNeeded functions. These ensure
that footnotes occur on lines by themselves (to make them
easier to see and move) and do not screw up line wrapping.

* Text.Pandoc.UTF8: modified fromUTF8 to strip out the BOM
if present. Windows Notepad and other applications insert a
BOM at the beginning of a UTF8 file.

* Main.hs (tabFilter): Treat '\r' at end of line as newline (in
addition to "\r\n" and '\n').

* Added a writer option for wrapped text and a command-line option
'--no-wrap', which disables text wrapping and minimizes whitespace
in HTML. (Resolves Issue #26.)

+ Added support for '--no-wrap' to Main.hs.
+ Added wrapIfNeeded function to Text.Pandoc.Shared.
+ Use wrapIfNeeded instead of wrapped in the RST, Man, Docbook, and
Markdown writers.
+ Added render and renderFragment helpers to HTML writer.

* Modified html2markdown to run tidy only if the HTML cannot be
parsed. Previously html2markdown piped all input through tidy
before passing it to pandoc. This caused problems on certain pages
(e.g. which have well-formed
XHTML that causes tidy to choke. The solution is to pipe through
tidy only if pandoc cannot parse the input by itself. This means
that a temp file is now always used, even when input comes from a
local file or standard input.

* Removed 'version' constant from Main.hs; added 'pandocVersion' to
Text.Pandoc library.

* pandoc.cabal:

+ Modified to work with GHC 6.8 and Cabal configurations. (For GHC
6.8, pretty and containers must be added to Build-Depends, and it
is desirable to use the -O2 compiler option.) Cabal configurations
allows one to select options depending on the compiler version.
For GHC 6.6, the splitBase option can be disabled.
+ pandoc.cabal.ghc66 is provided for users with older versions of
Cabal, which do not support configurations.
+ Use Ghc-Prof-Options to ensure that '-auto-all' is used when
'--enable-(executable|library)-profiling' is specified. Updated
PROFILING instructions accordingly.

* Makefile:

+ Makefile now checks GHC version. If GHC is 6.6, pandoc.cabal.ghc66
is copied to pandoc.cabal, and the old pandoc.cabal is copied
to pandoc.cabal.orig. Otherwise, pandoc.cabal is copied to
pandoc.cabal.orig but otherwise unmodified. This way, the Makefile
will work properly with either GHC 6.6 or 6.8.
+ Changed BUILDCONF to point to dist/setup-config, not .setup-config.
This is where current versions of Cabal put it.
+ Added $(BUILDCMD) target, so setup doesn't get compiled every time.
+ Removed dependency of templates on ./templates, which is circular
now that templates is a subdirectory of the top-level.

* MacPorts Portfile:

+ Modified to install the pandoc library in addition to programs.
+ Installation must be done manually rather than using Makefile's
+ Note that the library must be registered in the activate phase,
after the library files have been copied out of the destroot.
Cabal generates a '' script that will do this.

* debian/control: Added libghc6-network-dev, libghc6-xhtml-dev, and
libghc6-mtl-dev as dependencies for libghc6-pandoc-dev.
Closes: #445235

* debian/rules: Converted to UTF-8.

* Changed pandoc home page to

* Updated ASCIIMathML.js to latest version.

* Directory structure:

+ Moved everything from src into the top-level directory.
+ Changed references to source directory in Makefile and
+ Moved ASCIIMathML.js, headers, and ui into templates directory.
+ Modified to reflect new paths.

[ Recai Oktaş ]

* Makefile: Fixed the issue of having two copies of the library
documentation under some usage scenarios.

* Replaced 'ghc' with '$(GHC)' in Makefile, and made GHC
and GHC_PKG configurable through the environment, to support
unusual ghc installations. For example:
GHC=/opt/ghc/bin/ghc GHC_PKG=/opt/ghc/bin/ghc-pkg make

-- Recai Oktaş <> Sun, 07 Oct 2007 20:51:43 +0300

pandoc (0.44) unstable; urgency=low

[ John MacFarlane ]

* Fixed bug in HTML writer: when --toc was used, anchors were put around
headers, which is invalid XHTML (block content within inline element).
Now the anchors are put inside the header tags. Resolves Issue #23.

* Added xmlns attribute to html element in html writer tests.
This attribute is added by more recent versions of the
xhtml library (>= 3000), and is required for valid XHTML.

[ Recai Oktaş ]

* On configure, compile 'Setup.hs' to 'setup' and use 'setup' as the build
command instead of 'runhaskell', which, on some platforms (such as s390,
alpha, m68k), throws the following error:

runhaskell Setup.hs configure --prefix=/usr
ghc-6.6.1: not built for interactive use

This causes a serious FTBFS bug. Closes: #440668.

-- Recai Oktaş <> Mon, 03 Sep 2007 18:24:02 +0300

pandoc (0.43) unstable; urgency=low

[ John MacFarlane ]

* The focus of this release is performance. The markdown parser
is about five times faster than in 0.42, based on benchmarks
with the TextMate manual.

* Main.hs: Replaced CRFilter and tabFilter with single function
tabFilter, which operates on the whole string rather than breaking
it into lines, and handles dos-style line-endings as well as tabs.

* Added separate LaTeX reader and native reader tests; removed
round-trip tests.

* Text.Pandoc.Shared:

+ Removed tabsToSpaces and tabsInLine (they were used only in Main.hs.)
+ General code cleanup (to elimante warnings when compiling with -Wall.)
+ Added 'wrapped' function, which helps wrap text into paragraphs,
using the prettyprinting library.
+ Rewrote charsInBalanced and charsInBalanced'.
- Documented restriction: open and close must be distinct characters.
- Rearranged options for greater efficiency.
- Bug fix: Changed inner call to charsInBalanced inside
charsInBalanced' to charsInBalanced'.
+ anyLine now requires that the line end with a newline (not eof).
This is a harmless assumption, since we always add newlines to the
end of a block before parsing with anyLine, and it yields a 10% speed
+ Removed unnecessary 'try' in anyLine.
+ Removed unneeded 'try' from romanNumeral parser.
+ Use notFollowedBy instead of notFollowedBy' in charsInBalanced.
+ Removed unneeded 'try' in parseFromString.
+ Removed unneeded 'try' from stringAnyCase. (Now it behaves
like 'string'.)
+ Changed definition of 'enclosed' in Text.Pandoc.Shared so that
'try' is not automatically applied to the 'end' parser. Added
'try' in calls to 'enclosed' where needed. Slight speed increase.

* Writers:

+ Replaced individual wrapping routines in RST, Man, and Markdown
writers with 'wrapped' from Text.Pandoc.Shared.
+ Rewrote LaTeX writer to use the prettyprinting library,
so we get word wrapping, etc.
+ Modified latex writer tests for new latex writer using prettyprinter.
+ Fixed bug in LaTeX writer: autolinks would not cause
'\usepackage{url}' to be put in the document header. Also, changes
to state in enumerated list items would be overwritten.
+ In Markdown writer, escape paragraphs that begin with ordered list
markers, so they don't get interpreted as ordered lists.

* Text.Pandoc.Reades.LaTeX:

+ Fixed bug in LaTeX reader, which wrongly assumed that the roman
numeral after "enum" in "setcounter" would consist entirely of
"i"s. 'enumiv' is legitimate.
+ LaTeX command and environment names can't contain numbers.
+ Rearranged order of parsers in inline for slight speed improvement.
+ Added '`' to special characters and 'unescapedChar'.

* Text.Pandoc.Readers.RST:

+ Removed unneeded try's in RST reader; also minor code cleanup.
+ Removed tabchar.
+ Rearranged parsers in inline (doubled speed).

* Text.Pandoc.Readers.Markdown:

+ Skip notes parsing if running in strict mode. (This yields a nice
speed improvement in strict mode.)
+ Simplify autolink parsing code, using Network.URI to test for
URIs. Added dependency on network library to debian/control and
+ More perspicuous definition of nonindentSpaces.
+ Removed unneeded 'try' in 'rawLine'.
+ Combined linebreak and whitespace into a new whitespace parser, to
avoid unnecessary reparsing of space characters.
+ Removed unnecessary 'try' in 'codeBlock', 'ellipses', 'noteMarker',
'multilineRow', 'dashedLine', 'rawHtmlBlocks'.
+ Use lookAhead in parsers for setext headers and definition lists
to see if the next line begins appropriately; if not, don't waste
any more time parsing.
+ Don't require blank lines after code block. (It's sufficient to
end code block with a nonindented line.)
+ Changed definition of 'emph': italics with '_' must not
be followed by an alphanumeric character. This is to help
prevent interpretation of e.g. `[LC_TYPE]: my_type` as
+ Improved in referenceLink: the two parts
of a reference-style link may be separated by one space, but not
more... [a] [link], [not] [a link].
+ Fixed markdown inline code parsing so it better accords with the marker for the end of the code section is a clump
of the same number of `'s with which the section began, followed
by a non-` character. So, for example,
` h ``` i ` -> `<code>h ``` i</code>`.
+ Split 'title' into 'linkTitle' and 'referenceTitle', since the
rules are slightly different.
+ Rewrote 'para' for greater efficiency.
+ Rewrote link parsers for greater efficiency.
+ Removed redundant 'referenceLink' in definition of inline (it's
already in 'link').
+ Refactored escapeChar so it doesn't need 'try'.
+ Refactored hrule for performance in Markdown reader.
+ More intelligent rearranging of 'inline' so that most frequently
used parsers are tried first.
+ Removed tabchar parser, as whitespace handles tabs anyway.

* Text.Pandoc.CharacterReferences:

+ Refactored.
+ Removed unnecessary 'try's for a speed improvement.
+ Removed unnecessary '&' and ';' from the entity table.

* Build process:

+ Makefile: Get VERSION from cabal file, not Main.hs.
+ Modified MacPorts Portfile:
- Depend on haddock
- Build and install libraries and library documentation in
addition to pandoc executable
- Added template item for md5 sum in
- Incorporated changes from MacPorts repository (r28278).
+ FreeBSD port: Don't try to generate distinfo in Makefile.
It can be made using 'make makesum' in FreeBSD.
+ Make both freebsd and macports targets depend on tarball.

* Website and documentation:

+ Updated INSTALL instructions.
+ Added pandocwiki demo to website.
+ Removed local references to Portfile, since pandoc is now in the
MacPorts repository.

-- Recai Oktaş <> Sun, 02 Sep 2007 15:50:11 +0300

pandoc (0.42) unstable; urgency=low

[ John MacFarlane ]

* Main.hs: Use utf8 conversion on the extra files loaded with
the -H, -C, -B, and -A options. This fixes problems with unicode
characters in these files.

* Exposed Text.Pandoc.ASCIIMathML, since it is imported in
Text.Pandoc.Readers.HTML and without it we get a linking error when
using the library.

* Markdown reader:

+ Added new rule for enhanced markdown ordered lists: if the list
marker is a capital letter followed by a period (including a
single-letter capital roman numeral), then it must be followed by
at least two spaces. The point of this is to avoid accidentally
treating people's initials as list markers: a paragraph might begin,
"B. Russell was an English philosopher," and this shouldn't be
treated as a list. Documented change in README.
+ Blocks that start with "p. " and a digit are no longer treated
as ordered lists (it's a page number).
+ Added a needed 'try' to listItem.
+ Removed check for a following setext header in endline.
A full test is too inefficient (doubles benchmark time), and the
substitute we had before is not 100% accurate.
+ Don't use Code elements for autolinks if --strict specified.

* LaTeX writer: When a footnote ends with a Verbatim environment, the
close } of the footnote cannot occur on the same line or an error occurs.
Fixed this by adding a newline before the closing } of every footnote.

* HTML writer:
+ Removed incorrect "{}" around style information in HTML tables.
Column widths now work properly in HTML.
+ If --strict option is specified (and --toc is not), don't include
identifiers in headers, for better Markdown compatibility.

* Build process:

+ Separated $(web_dest) and website targets.
+ In website, index.txt is now constructed from template
+ Added freebsd target to Markefile. This creates the freebsd Makefile
from, and creates distinfo. Removed Makefile and distinfo
from the repository.
+ Added macport target to Makefile. Portfile is built from template
+ Removed OSX package targets. (Too many difficulties involving
dependencies on dynamic libraries.)
+ More complete INSTALL instructions for all architectures.

* Website:
+ Added a programming demo, pandocwiki.

[ Recai Oktaş ]

* Do not forget to close pandoc's ITP. Closes: #391666

-- Recai Oktaş <> Sun, 26 Aug 2007 22:51:32 +0300

pandoc (0.41) unstable; urgency=low

[ John MacFarlane ]

* Fixed bugs in HTML reader:
+ Skip material at end *only if* `</html>` is present (previously,
only part of the document would be parsed if an error was
found; now a proper error message is given).
+ Added new constant eitherBlockOrInline with elements that may
count either as block-level or as inline. Modified isInline and
isBlock to take this into account.
+ Modified rawHtmlBlock to accept any tag (even an inline tag):
this is innocuous, because rawHtmlBlock is tried only if a regular
inline element can't be parsed.
+ Added a necessary 'try' in definition of 'para'.

* Fixed bug in markdown ordered list parsing. The problem was that
anyOrderedListStart did not check for a space following the
ordered list marker. So in 'A.B. 2007' the parser would be
expecting a list item, but would not find one, causing an error.
Fixed a similar bug in the RST reader. Resolves Issue #22.

* Refactored RST and Markdown readers using parseFromString.

* LaTeX reader will now skip anything after \end{document}.

* Fixed blockquote output in markdown writer: previously, block
quotes in indented contexts would be indented only in the first

* Added note to INSTALL about variations in versions of the xhtml
library that can lead to failed tests (thanks to Leif LeBaron).

-- Recai Oktaş <> Sun, 19 Aug 2007 23:26:07 +0300

pandoc (0.4) unstable; urgency=low

[ John MacFarlane ]

* Added two new output formats: groff man pages and ConTeXt. By
default, output files with extensions ".ctx" and ".context" are
assumed to be ConTeXt, and output files with single-digit extensions
are assumed to be man pages.

* Enhanced ordered lists (documented in README, under Lists):
+ The OrderedList block element now stores information about
list number style, list number delimiter, and starting number.
+ The readers parse this information when possible.
+ The writers use this information to style ordered lists.
+ The enhancement can be disabled using the --strict option.

* Added support for tables (with a new Table block element). Two kinds
of tables are supported: a simple table with one-line rows, and a
more complex variety with multiline rows. All output formats are
supported, but only markdown tables are parsed at the moment. The
syntax is documented in README.

* Added support for definition lists (with a new DefinitionList block
element). All output and input formats are supported. The syntax is
documented in README.

* Added support for superscripts and subscripts (with new Superscript
and Subscript inline elements). All input and output
formats. The syntax is documented in README.

* Added support for strikeout (with a new Strikeout inline element).
All input and output formats are supported. Thanks to Bradley Kuhn,
who contributed a patch. The syntax is documented in README. Resolves
Issue #18.

* Added a --toc|--table-of-contents option. This causes an automatically
generated table of contents (or an instruction that creates one) to
be inserted at the beginning of the document. Not supported in S5,
DocBook, or man page writers.

* Modified the -m|--asciimathml option:

+ If an optional URL argument is provided, a link is inserted
instead of the contents of the ASCIIMathML.js script.
+ Nothing is inserted unless the document actually contains
LaTeX math.

* Removed Blank block element as unnecessary.

* Removed Key and Note blocks from the Pandoc data structure. All
links are now stored as explicit links, and note contents are
stored with the (inline) notes.

+ All link Targets are now explicit (URL, title) pairs; there
is no longer a 'Ref' target.
+ Markdown and RST parsers now need to extract data from key and
note blocks and insert them into the relevant inline elements.
Other parsers have been simplified, since there is no longer any need
to construct separate key and note blocks.
+ Markdown, RST, and HTML writers need to construct lists of
notes; Markdown and RST writers need to construct lists of link
references (when the --reference-links option is specified); and
the RST writer needs to construct a list of image substitution
references. All writers have been rewritten to use the State monad
when state is required.
+ Several functions (generateReference, keyTable,
replaceReferenceLinks, replaceRefLinksBlockList, and some auxiliaries
used by them) have been removed from Text.Pandoc.Shared, since
they are no longer needed. New functions and data structures
(Reference, isNoteBlock, isKeyBlock, isLineClump) have been
added. The functions inTags, selfClosingTag, inTagsSimple, and
inTagsIndented have been moved to the DocBook writer, since that
is now the only module that uses them. NoteTable is now exported
in Text.Pandoc.Shared.
+ Added stateKeys and stateNotes to ParserState; removed stateKeyBlocks,
stateKeysUsed, stateNoteBlocks, stateNoteIdentifiers, stateInlineLinks.
+ Added writerNotes and writerReferenceLinks to WriterOptions.

* Added Text.Pandoc module that exports basic readers, writers,
definitions, and utility functions. This should export everything
needed for most uses of Pandoc libraries. The haddock documentation
includes a short example program.

* Text.Pandoc.ASCIIMathML is no longer an exported module.

* Added Text.Pandoc.Blocks module to help in printing markdown
and RST tables. This module provides functions for working with
fixed-width blocks of text--e.g., placing them side by side, as
in a table row.

* Refactored to avoid reliance on Haskell's Text.Regex library, which
(a) is slow, and (b) does not properly handle unicode. This fixed
some strange bugs, e.g. in parsing S-cedilla, and improved performance.

+ Replaced 'gsub' with a general list function 'substitute'
that does not rely on Text.Regex.
+ Rewrote extractTagType in HTML reader so that it doesn't use
+ In Markdown reader, replaced email regex test with a custom email
autolink parser (autoLinkEmail). Also replaced selfClosingTag regex
with a custom function isSelfClosingTag.
+ Modified Docbook writer so that it doesn't rely on Text.Regex for
detecting 'mailto' links.
+ Removed escapePreservingRegex and reamped entity-handling
functions in Text.Pandoc.Shared and Text.Pandoc.CharacterReferences to
avoid reliance on Text.Regex (see below on character reference
handling changes).

* Renamed Text.Pandoc.Entities as Text.Pandoc.CharacterReferences.

* Changed handling of XML entities. Entities are now parsed (and unicode
characters returned) in the Markdown and HTML readers, rather than being
handled in the writers. In HTML and Docbook writers, UTF-8 is now used
instead of entities for characters above 128. This makes the HTML and
DocBook output much more readable and more easily editable.

+ Removed sgmlHexEntity, sgmlDecimalEntity, sgmlNamedEntity, and
sgmlCharacterEntity regexes from Text.Pandoc.Shared.
+ Renamed escapeSGMLChar to escapeCharForXML. Added escapeStringForXML.
Moved both functions to Text.Pandoc.Writers.Docbook.
+ Added characterReference parser to Text.Pandoc.CharacterReferences.
This parses a string and return a unicode character.
+ Rewrote decodeCharacterReferences to use the new parser instead of
+ Added new charRef parser for Markdown and HTML, which replaces the
old 'entity' parser. Added '&' as a special character in Markdown reader.
+ Modified HTML and Markdown readers to call decodeEntities on all raw
strings (e.g. authors, dates, link titles), to ensure that no
unprocessed entities are included in the native representation of
the document. (In the HTML reader, most of this work is done by a
change in extractAttributeName.)
+ In XML and Markdown output, escape unicode nonbreaking space as '&nbsp;',
since a unicode non-breaking space is impossible to distinguish visually
from a regular space. (Resolves Issue #3.)
+ Removed encodeEntitiesNumerical.
+ Use Data.Map for entityTable and (new) reverseEntityTable, for a
slight performance boost over the old association list.
+ Removed unneeded decodeEntities from 'str' parser in HTML and
Markdown readers.

* Text.Pandoc.UTF8: Renamed encodeUTF8 to toUTF8, decodeUTF8 to
fromUTF8, for clarity.

* Replaced old haskell98 module names replaced by hierarchical module
names, e.g. List by Data.List. Removed haskell98 from dependencies
in pandoc.cabal, and added mtl (needed for state monad). Substituted
xhtml for html.

* Refactored and cleaned up character escaping in writers, using
backslashEscapes and escapeStringUsing functions.

* Instead of adding "\n\n" to the end of an input string in Main.hs,
this is now done in the readers. This makes the libraries behave
the way you'd expect from the pandoc program. Resolves Issue #10.

* URLs and email addresses in autolinks are now typeset as Code.

* In Main.hs, changed putStr to putStrLn -- mainly because MacOS X
doesn't display the whole output unless there's a line ending.

* Major code cleanup in all modules, for greater consistency, concision,
and readability.

* HTML reader:

+ Fixed several bugs (extractTagType, attribute parsing).
+ Remove Null blocks in lists of blocks when possible.
+ Allow HTML comments as raw HTML inline.

* Markdown reader:

+ Ordered list items may no longer begin with uppercase letters, or
letters greater than 'n'. (This prevents first initials and page
reference, e.g. 'p. 400', from being parsed as beginning lists.)
Also, numbers beginning list items may no longer end with ')',
which is now allowed only after letters. Note: These changes
may cause documents to be parsed differently. Users should take
care in upgrading.
+ Changed autoLink parsing to conform better to's
behavior. `<>` is not treated as a link, but
`<>`, `<>`, and
`<>` are.
+ Cleaned up handling of embedded quotes in link titles. Now these are
stored as a '"' character, not as '&quot;'.
+ Use lookAhead parser for the 'first pass' (looking for reference keys),
instead of parsing normally, then using setInput to reset input. This
yields a slight performance boost.
+ Fixed several bugs in smart quote recognition.
+ Fixed bug in indentSpaces (which didn't properly handle
cases with mixed spaces and tabs).
+ Consolidated 'text', 'special', and 'inline' into 'inline'.
+ Fixed bug which allowed URL and title to be separated by multiple blank
lines in links and reference keys. They can be on separate lines but
can't have blank lines between them.
+ Correctly handle bracketed text inside inline footnotes and links,using
new function inlinesInBalanced. Resolves Issue #14.
+ Fixed bug in footnotes: links in footnotes were not being
processed. Solution: three-stage parse. First, get all the
reference keys and add information to state. Next, get all the
notes and add information to state. (Reference keys may be needed
at this stage.) Finally, parse everything else.
+ Replaced named constants like 'emphStart' with literals.
+ Removed an extra occurrence of escapedChar in definition of inline.

* RST reader:

+ Allow the URI in a RST hyperlink target to start on the line
after the reference key.
+ Added 'try' in front of 'string', where needed, or used a different
parser. This fixes a bug where ````` would not be correctly parsed as
a verbatim `.
+ Fixed slow performance in parsing inline literals in RST reader. The
problem was that ``#`` was seen by 'inline' as a potential link or image.
Fix: inserted 'notFollowedBy (char '`')' in link parsers.
Resolves Issue #8.
+ Use lookAhead instead of getInput/setInput in RST reader. Removed
unneeded getState call, since lookAhead automatically saves and
restores the parser state.
+ Allow hyperlink target URIs to be split over multiple lines, and
to start on the line after the reference. Resolves Issue #7.
+ Fixed handling of autolinks.

* LaTeX reader:

+ Replaced 'choice [(try (string ...), ...]' idiom with 'oneOfStrings',
for clarity.
+ Added clauses for tilde and caret. Tilde is \ensuremath{\sim}, and
caret is \^{}, not \^ as before.
+ Added parsing for \url.
+ Parse \texttt{} as code, provided there's nothing fancy inside.

* HTML writer:

+ Modified HTML writer to use the Text.XHtml library. This results
in cleaner, faster code, and it makes it easier to use Pandoc in
other projects, like wikis, which use Text.XHtml. Two functions are
now provided, writeHtml and writeHtmlString: the former outputs an
Html structure, the latter a rendered string. The S5 writer is also
changed, in parallel ways (writeS5, writeS5String).
+ The Html header is now written programmatically, so it has been
removed from the 'headers' directory. The S5 header is still
needed, but the doctype and some of the meta declarations have
been removed, since they are written programmatically. This change
introduces a new dependency on the xhtml package.
+ Fixed two bugs in email obfuscation involving improper escaping
of '&' in the `<noscript>` section and in `--strict` mode. Resolves
Issue #9.
+ Fixed another bug in email obfuscation: If the text to be obfuscated
contains an entity, this needs to be decoded before obfuscation.
Thanks to thsutton for the patch. Resolves Issue #15.
+ Changed the way the backlink is displayed in HTML footnotes.
Instead of appearing on a line by itself, it now generally
appears on the last line of the note. (Exception: when the
note does not end with a Plain or Para block.) This saves space
and looks better.
+ Added automatic unique identifiers to headers:
- The identifier is derived from the header via a scheme
documented in README.
- WriterState now includes a list of header identifiers and a table
of contents in addition to notes.
- The function uniqueIdentifiers creates a list of unique identifiers
from a list of inline lists (e.g. headers).
- This list is part of WriterState and gets consumed by blockToHtml
each time a header is encountered.
+ Include CSS for .strikethrough class in header only if strikethrough
text appears in the document.
+ If the 'strict' option is specified, elements that do not appear in
standard markdown (like definition lists) are passed through as
raw HTML.
+ Simplified treatment of autolinks, using pattern matching instead of

* Markdown writer:

+ Links in markdown output are now printed as inline links by default,
rather than reference links. A --reference-links option has been added
that forces links to be printed as reference links. Resolves Issue #4.
+ Use autolinks when possible. Instead of `[](`,
use `<>`.

* LaTeX writer:

+ Rewrote to use the State monad. The preamble now includes only those
packages that are actually required, given the document's content.
Thus, for example, if strikeout is not used, ulem is not required.
Modified LaTeXHeader accordingly.
+ Modified LaTeX writer to insert '\,' between consecutive quotes.
+ Removed unused function tableRowColumnWidths.
+ Simplified code for escaping special characters.
+ Leave extra blank line after \maketitle.
+ Include empty '\author{}' when no author specified to avoid LaTeX
+ Include fancyvrb code in header only if needed -- that is, only
if there is actually code in a footnote.
+ Use \url{} for autolinks.
+ Include [mathletters] option in ucs package, so that basic unicode
Greek letters will work correctly.

* RST writer: Force blank line before lists, so that sublists will
be handled correctly.

* Docbook writer: Fixed a bug: email links with text, like
[foo](me@bar.baz), were being incorrectly treated as autolinks.

* Removed Text.ParserCombinators.Pandoc and moved all its functions to

* Text.Pandoc.Shared:

+ Added defaultWriterOptions.
+ Added writerTableOfContents to WriterOptions.
+ Added writerIgnoreNotes option to WriterOptions. This is needed
for processing header blocks for a table of contents, since notes on
headers should not appear in the TOC.
+ Added prettyprinting for native Table format.
+ Removed some unneeded imports.
+ Moved escape and nullBlock parsers from
Text.ParserCombinators.Pandoc, since the latter is for
general-purpose parsers that don't depend on Text.Pandoc.Definition.
+ Moved isHeaderBlock from Text.Pandoc.Writers.HTML.
+ Moved Element, headerAtLeast, and hierarchicalize from Docbook
writer, because HTML writer now uses these in constructing a table
of contents.
+ Added clauses for new inline elements (Strikeout, Superscript,
Subscript) to refsMatch.
+ Removed backslashEscape; added new functions escapeStringUsing and
+ Moved failIfStrict from markdown reader, since it is now used also
by the HTML reader.
+ Added a 'try' to the definition of indentSpaces.
+ In definition of 'reference', added check to make sure it's not a note
+ Added functions: camelCaseToHyphenated, toRomanNumeral,
anyOrderedListMarker, orderedListmarker, orderedListMarkers,
charsInBalanced', withHorizDisplacement, romanNumeral
+ Fixed a bug in the anyLine parser. Previously it would parse an empty
string "", but it should fail on an empty string, or we get an error
when it is used inside "many" combinators.
+ Removed followedBy' parser, replacing it with the lookAhead parser from
+ Added some needed 'try's before multicharacter parsers, especially in
'option' contexts.
+ Removed the 'try' from the 'end' parser in 'enclosed', so that
'enclosed' behaves like 'option', 'manyTill', etc.
+ Added lineClump parser, which parses a raw line block up to and
including any following blank lines.
+ Renamed parseFromStr to parseFromString.
+ Added a 'try' to the 'end' parser in 'enclosed'. This makes errors in
the use of 'enclosed' less likely. Removed some now-unnecessary 'try's
in calling code.
+ Removed unneeded 'try' in blanklines.
+ Removed endsWith function and rewrote calling functions to use
isSuffixOf instead.
+ Added >>~ combinator.
+ Fixed bug in normalizeSpaces: Space:Str "":Space should compress to

* Refactored; added separate tests for tables.

* Shell scripts:

+ Added -asxhtml flag to tidy in html2markdown. This will
perhaps help the parser, which expects closing tags.
+ Modified markdown2pdf to run pdflatex a second time if --toc or
--table-of-contents was specified; otherwise the table of
contents won't appear.
+ Modified markdown2pdf to print a helpful message if the 'ulem'
LaTeX package is required and not found.

* Changes to build process:

+ Dropped support for compilation with GHC 6.4. GHC 6.6 or higher
is now required.
+ Removed cabalize and The repository now contains
pandoc.cabal itself.
+ Pandoc.cabal has been changed to pandoc.cabal, because HackageDB
likes the cabal file to have the same name as the tarball.
+ Expanded and revised the package description in pandoc.cabal.
Revised the package synopsis.
+ The tarball built by 'make tarball' now contains files built from
templates (including man pages and shell scripts), so pandoc can
be built directly using Cabal tools, without preprocessing.
+ Executable binaries are now stripped before installing.
+ Man pages are now generated from markdown sources, using pandoc's
man page writer.
+ Use HTML version of README (instead of RTF) in Mac OS X installer.
+ Instead of testing for the existence of a pandoc symlink in build-exec,
use ln -f.

* Documentation:

+ Updated README and man pages with information on new features.
+ Updated INSTALL instructions with some useful clarifications and
+ Updated web content.

* Added FreeBSD port.

[ Recai Oktaş ]

* debian/control:

+ Changed pandoc's Build-Depends to include libghc6-mtl-dev and
libghc6-xhtml-dev. Removed libghc6-html-dev.
+ Suggest texlive-latex-recommended | tetex-extra instead of
tetex-bin. This brings in fancyvrb and unicode support.

-- Recai Oktaş <> Tue, 16 Jan 2007 00:37:21 +0200

pandoc (0.3) unstable; urgency=low

[ John MacFarlane ]

* Changes in pandoc options:

+ Allow options to follow or precede arguments.
+ Changed '--smartypants' to '--smart' and adjusted symbols accordingly.
+ Added '--strict' option.
+ Added '-o/--output' option.
+ Added '--dump-args' and '--ignore-args' options (for use in wrappers).
+ Modified '-v' and '-h' output to go to STDERR, not STDOUT, and return
error conditions. This is helpful for writing wrappers.
+ Added copyright message to '-v' output, modeled after FSF messages.
+ Reformatted usage message so that it doesn't wrap illegibly.
+ Removed extra blanks after '-h' and '-D' output.

* Added docbook writer.

* Added implicit setting of default input and output format based
on input and output filename extensions. These defaults are
overridden if explicit input and output formats are specified using
'-t', '-f', '-r', or '-w' options. Documented in pandoc(1) man page

* Allow ordered list items to begin with (single) letters, as well
as numbers. The list item marker may now be terminated either by
'.' or by ')'. This extension to standard markdown is documented

* Revised footnote syntax. (See README for full details.) The
'[^1]' format now standard in markdown extensions is supported,
as are inline footnotes with this syntax: `^[My note.]`.
The earlier footnote syntax `^(1)` is no longer supported.

* Improved HTML representation of footnotes. All footnotes
are now auto-numbered and appear in an ordered list at the
end of the HTML document. Since the default appearance is now
acceptable, the old footnote styles have been removed from the
HTML header.

* Bug fixes:

+ Fixed a serious bug in the markdown, LaTeX, and RST readers.
These readers ran 'runParser' on processed chunks of text to handle
embedded block lists in lists and quotation blocks. But then
any changes made to the parser state in these chunks was lost,
as the state is local to the parser. So, for example, footnotes
didn't work in quotes or list items. The fix: instead of calling
runParser on some raw text, use setInput to make it the input, then
parse it, then use setInput to restore the input to what it was
before. This is shorter and more elegant, and it fixes the problem.
+ Fixed bug in notFollowedBy' combinator (adding 'try' before
'parser'). Adjusted code that uses this combinator accordingly.
+ Fixed bug in RTF writer that caused improper indentation on
footnotes occurring in indented blocks like lists.
+ Fixed parsing of metadata in LaTeX reader. Now the title, author,
and date are parsed correctly. Everything else in the preamble
is skipped.
+ Modified escapedChar in LaTeX reader to allow a '\' at the end of a
line to count as escaped whitespace.
+ Modified LaTeX reader to produce inline links rather than reference
links. Otherwise, links in footnotes aren't handled properly.
+ Fixed handling of titles in links in Markdown reader, so that
embedded quotation marks are now handled properly.
+ Fixed Markdown reader's handling of embedded brackets in links.
+ Fixed Markdown reader so that it only parses bracketed material
as a reference link if there is actually a corresponding key.
+ Revised inline code parsing in Markdown reader to conform to
markdown standard. Now any number of `s can begin inline code,
which will end with the same number of `s. For example, to
have two backticks as code, write ``` `` ```. Modified Markdown
writer accordingly.
+ Fixed bug in text-wrapping routine in Markdown and RST writers.
Now LineBreaks no longer cause wrapping problems.
+ Supported hexadecimal numerical entity references as well as
decimal ones.
+ Fixed bug in Markdown reader's handling of underscores and other
inline formatting markers inside reference labels: for example,
in '[A_B]: /url/a_b', the material between underscores was being
parsed as emphasized inlines.
+ Changed Markdown reader's handling of backslash escapes so that
only non-alphanumeric characters can be escaped. Strict mode
follows in only allowing a select group of punctuation
characters to be escaped.
+ Modified HTML reader to skip a newline following a `<br>` tag.
Otherwise the newline will be treated as a space at the beginning
of the next line.

* Made handling of code blocks more consistent. Previously, some
readers allowed trailing newlines, while others stripped them.
Now, all readers strip trailing newlines in code blocks. Writers
insert a newline at the end of code blocks as needed.

* Modified readers to make spacing at the end of output more consistent.

* Minor improvements to LaTeX reader:

+ '\thanks' now treated like a footnote.
+ Simplified parsing of LaTeX command arguments and options.
commandArgs now returns a list of arguments OR options (in
whatever order they appear). The brackets are included, and
a new stripFirstAndLast function is provided to strip them off
when needed. This fixes a problem in dealing with \newcommand
and \newenvironment.

* Revised RTF writer:

+ Default font is now Helvetica.
+ An '\f0' is added to each '\pard', so that font resizing works

* Moved handling of "smart typography" from the writers to the Markdown
and LaTeX readers. This allows great simplification of the writers
and more accurate smart quotes, dashes, and ellipses. DocBook can
now use `<quote>`. The '--smart' option now toggles an option in
the parser state rather than a writer option. Several new kinds
of inline elements have been added: Quoted, Ellipses, Apostrophe,
EmDash, EnDash.

* Changes in HTML writer:

+ Include title block in header even when title is null.
+ Made javascript obfuscation of emails even more obfuscatory,
by combining it with entity obfuscation.

* Changed default ASCIIMathML text color to black.

* Test suite:

+ Added --strip-trailing-cr option to diff in, for
compatibility with Windows.
+ Added regression tests with footnotes in quote blocks and lists.

* Makefile changes:

+ osx-pkg target creates a Mac OS X package (directory). New osx
directory contains files needed for construction of the package.
+ osx-dmg target creates a compressed disk image containing the package.
+ win-pkg target creates Windows binary package.
+ tarball target creates distribution source tarball.
+ website target generates pandoc's website automatically, including
demos. New 'web' directory contains files needed for construction
of the website (which will be created as the 'pandoc' subdirectory
of 'web').
+ Makefile checks to see if we're running Windows/Cygwin; if so,
a '.exe' extension is added to each executable in EXECS.

* Removed all wrappers except markdown2pdf and html2markdown.

* Added new wrapper hsmarkdown, to be used as a drop-in replacement
for hsmarkdown calls pandoc with the '--strict'
option and disables other options.

* Added code to html2markdown that tries to determine the character
encoding of an HTML file, by parsing the "Content-type" meta tag.

+ If the encoding can't be determined, then if the content is local,
the local encoding is used; if it comes from a URL, UTF-8 is used
by default.
+ If input is from STDIN, don't try to determine character encoding.
+ Encoding can be specified explicitly using '-e' option.

* Improved warning messages in wrappers:

+ Print warning if iconv not available
+ More user-friendly error messages in markdown2pdf, when
pdflatex fails.

* Code cleanup:

+ Renamed 'Text/Pandoc/HtmlEntities' module to
'Text/Pandoc/Entities'. Also changed function names so as
not to be HTML-specific.
+ Refactored SGML string escaping functions from HTML and Docbook
writers into Text/Pandoc/Shared. (escapeSGML, stringToSGML)
+ Removed 'BlockQuoteContext' from ParserContext, as it isn't
used anywhere.
+ Removed splitBySpace and replaced it with a general, polymorphic
splitBy function.
+ Refactored LaTeX reader for clarity (added isArg function).
+ Converted some CR's to LF's in src/ui/default/print.css.
+ Added license text to top of source files.
+ Added module data for haddock to source files.
+ Reformatted code for consistency.

* Rewrote documentation and man pages. Split README into INSTALL


* Removed TODO, since we now maintain ToDo on the wiki.

* Made COPYRIGHT in top level a symlink to debian/copyright, to avoid

[ Recai Oktaş ]

* Revamped build process to conform to debian standards and created
a proper debian package. Closes: #391666.

* Modified build process to support GHC 6.6.

+ The package can still be compiled using GHC 6.4.2, though because
of dependencies the "make deb" target works only with GHC 6.6+.
+ The script 'cabalize' is used to create an appropriate
'Pandoc.cabal' from '', depending on the GHC and
Cabal versions.

* Refactored template processing (

* Modified wrapper scripts to make them more robust and portable.
To avoid code duplication and ensure consistency, wrappers are
generated via a templating system from templates in src/wrappers.

+ Wrappers now accept multiple filenames, when appropriate.
+ Spaces and tabs allowed in filenames.
+ getopts shell builtin is used for portable option parsing.
+ Improved html2markdown's web grabber code, making it more robust,
configurable and verbose. Added '-e', '-g' options.

-- Recai Oktaş <> Fri, 05 Jan 2007 09:41:19 +0200

pandoc (0.2) unstable; urgency=low

* Fixed unicode/utf-8 translation

-- John MacFarlane <> Mon, 14 Aug 2006 00:00:00 -0400

pandoc (0.1) unstable; urgency=low

* Initial creation of debian package

-- John MacFarlane <> Mon, 14 Aug 2006 00:00:00 -0400
comments powered byDisqus