Module documentation for 4.5
Library for parsing epub document metadata (Haskell)
Library for parsing and manipulating epub document metadata. Supports epub versions 2 and 3.
Why was this done?
The motivation for this project grew out of my desire to take charge of missing or incorrect epub metadata in books I have purchased. I started out using the Calibre open source tools for examining this info. Limitations and incomplete implementation of those tools led me here to build a more complete implementation in the programming language that I love beyond all others.
Why didn’t I just use existing solutions?
Calibre ebook-meta utility
I experienced various problems using this software, such as:
Incomplete and in some cases incorrect handling of tags that can exist more than once (creator, contributor), particularly when they are differentiated using attributes according to the spec.
Unable to display many fields in the OPF Package Document metadata specification. Unable to manipulate data that is represented as attributes of tags in the OPF spec.
Astonishingly slow performance. A command-line renaming tool built with epub-metadata is more than 45 times faster at parsing and displaying epub metadata. I’m going to blame Python here for Calibre’s performance. This has had a big impact on projects where I’ve been processing hundreds of epubs in batch operations.
epub on Hackage, epub E-Book construction support library
The focus of this project seems to be with building new documents, not parsing existing files. And there is a specific attempt to do more than the metadata, to gather up the content and other metafiles that make up an epub for creation.
Examining Codec.Ebook.OPF.Types, most of the metadata fields from the OPF Package Document spec are missing or aren’t modeled thoroughly. I felt that to contribute to this project, I would have had to significantly rip up the types and redesign them.
At this time it seemed like a better solution for me to start fresh with modelling the types and code to manipulate them. That said, I would be very interested in combining the epub and epub-metadata projects at some point in some way that makes sense.
Using this library
This file has a working example of using this library.
- Download the cabalized source package from Hackage
- Get the source with darcs:
$ darcs get http://hub.darcs.net/dino/epub-metadata
- Get the source with cabal-install:
$ cabal get epub-metadata
- Get the source with stack:
$ stack unpack epub-metadata
- If you’re just looking, browse the source
And once you have it, building the usual way:
$ stack build
$ stack test
$ stack haddock
Dino Morelli <[email protected]>
- Fix merged for UTF8 issue
- Removed defunct cabal stability field
- Moved copyright dates up to 2016
- Moved tested-with up to GHC 8.0.1
- Removed GHC switch simpl-tick-factor
- Moved cabal-version up to >= 1.10
- Project now builds with stack
- Fixed failing damaged zip file unit test
- Replaced deprecated Control.Monad.Error with Control.Monad.Except
- Updated cabal homepage, tested-with and source-repository
- Cleaned out ui3.info stuff from the README
- Updated boringfile with cabal sandbox filespecs
- TODO is now a Markdown document
- Added many files missing from extra-source-files
- Fix for Simplifier ticks exhausted problem that was exposed after upgrade to GHC 7.8.2
- Added README.md and changelog.md files, better docs!
- Some documentation fixes
- Fixed a bug reading books where the container XML document is malformed
- Added support for epub3 documents. This was done using a single set of datatypes, not specific to either epub2 or epub3.
- Redesigned the book examining API to be an edsl. Actions are to be combined together based on what the developer needs from the document.
- The data structures to contain epub metadata “sections” were redesigned to no longer be nested. Part of this change includes a typeclass-based pretty-print API for displaying this data.
- Documentation rewrites and additions, including a working code example in the API docs.
- Repackaged this code to show that it’s for epub v2.x only at this time
- Fixed a bug where mimetype file not occurring first in the archive on Windows
- Fixed support for all 0-or-more epub2 metadata fields
- Redesigned the Format (metadata display) code
- Modularized the unit tests so that more tests could be added that aren’t specifically about parsing metadata.
- Cosmetic updates to the epub metadata API including some documentation clean-up
- Added code to strip illegal characters from the beginning of the OPF metadata XML file
- Switched the encoding hack to be case-insensitive
- Added a fix for epub zip files with a damaged central directory signature
- Encountered a problem in Windows where using a lazy ByteString to read the entire epub zip file was never closing the file while consumer applications were still running. Changed this library to use a strict ByteString instead for reading the zip archive.
- Changed display output of OPF data to label the textual content of a tag as “text”
- Extensive changes to the cabal build of this project to bring it up to Cabal 1.10
- Unit tests now use the test-suite cabal stanza
- Now includes some API functions for working directly with zip archives and directories full of the loose files which will be made into epubs
- Moved the epubmeta utility from this project. It’s now in epub-tools
- Project now uses zip-archive instead of the libzip library. zip-archive is a pure Haskell solution that’s easier to build on non-UNIX-like platforms
- Other changes to the build and project to make it buildable under Windows
- This library now uses the haskell LibZip library instead of relying on unzip shell command invocation
- Changed command-line utility name from epub-meta to epubmeta
- Now always using namespace qualification for all OPF Package tags. Library would incorrectly fail if the OPF document was created using explicit namespacing for the OPF package elements.
- Hack to avoid problems with UTF-8 encoding in OPF Package XML documents
- Hack to avoid problems with !DOCTYPE declarations. Found some books where a DTD was specified for a completely different type of document. Discarding these declarations for now.
- Changed this code to use System.Process instead of HSH for unzip invocation in order to have more control over interpreting exit codes.
- Added parsing and output of the manifest, spine and guide portions of the spec. Full read capability of OPF Package data is now complete.
- Major redesign of the API. It’s been split and organized into smaller, more focused modules.
- Updated to work with the 9.x version of hxt
- Redesigned epub metadata formatting code to be much more efficient using Writer monad and Data.Sequence
- Added support for new capabilities to command-line utility
- Clarified the Haddock documentation to describe how these data types map to the source XML data
- Updated unit tests to reflect the large API changes
- Initial release