A DSL for marking student work https://github.com/DIKU-EDU/remarks#readme
|Latest on Hackage:||0.1.13|
This package is not currently in any snapshots. If you're interested in using it, we recommend adding it to Stackage Nightly. Doing so will make builds more reliable, and allow stackage.org to host generated Haddocks.
remarks — A DSL for marking student work
When judging student performance, it is useful to have both small, composable, quantitative judgements, and qualitative remarks. This makes both spreadsheets and mere text-files ill-suited for marking student work. Although org-mode can solve this problem to a great extent, it becomes a heavy tool in the light of having to mark hundreds of students in a distributed fashion. With org-mode, everything is in one file, while global, intra-student statistics are not needed until all the students have been fully marked.
- One, or several, human-readable/editable file(s) per student.
- git-friendly file format.
- Export options to spreadsheet-formats.
- Synchronization options with Dropbox and/or Google Drive.
Goal 4 is not necessarily related to
remarks, but is related to marking
student work with external examiners, who are not always willing to use more
explicit version-control systems, such as git.
remarks check [<file>]checks that the file system structure is well-formed. For instance, checks that all explicitly stated point sums are correct.
remarks show [<file>]checks the file system structure as above, and shows the overall judgement for each given argument.
remarks summary [--depth <depth>] [<file>]checks and summarizes the points. Depth 0 (default) lists just the top-level judgements.
remarks pending [--depth <depth>] [<file>]shown the corrections that has not been completed. Can be cut at a given depth; depth 0 lists just the top-level judgements.
remarks export [--format "<format>"] [<file>]exports corrections to a semicolon separated list. The format is a semicolon separated string of properties.
remarks exportHTML [<file>]exports all corrections to a dynamic html-table. Developer features:
remarks parse [<file>]parses the given files and shows their ASTs.
$ cabal install remarks
If you clone the repository, or download the sources, you get two further options:
If you are using the purely functional package manager Nix, you can do this:
This will create a symlink
result, pointing to the directory in your Nix store, containing the binary.
Or, if using Stack, you can do this:
$ stack build $ stack install
.mrk file is a list of judgements.
A judgement starts with a header mark (a sequence of
#), a title (followed by
:), given points (followed by
/), and maximum points (followed by a line
break). The number of
# determines the depth of the header, and every file
must start at depth 1, but may have multiple depth 1 judgements. Headings may
be arbitrarily nested, but must sum up correctly. For instance, here is a file
containing only quantitative remarks:
# Theory: 27/50 ## Question 1: 10/10 ## Question 2: 10/20 ## Question 3: 7/20 # Practice: 35/50 ## Task 1: 20/25 ## Task 2: 15/25
The header of a judgement may be followed by qualitative remarks. Remarks begin with an indent (two spaces), and a mood mark:
*for neutral/structural remarks;
+for positive remarks;
-for negative remarks;
?for impartial remarks.
Impartial remarks are good for judgements where the mood is left to be judged by a higher authority. For instance, when a teaching assistant is uncertain, and would like to let the teacher decide, or the solution cannot be used to make a judgement about this point.
Structural remarks are good for listing things to look for. For instance, a (template) judgement for an operating systems exercise might look something like this:
## T1: /15 ### Formal requirements: /5 * Has code * Has XML in a reloadable/testable form * Has an explanation ### Quality assessment: /10 * Understands semaphores * Understands deadlocks * Understands starvation * Avoids deadlocks * Avoids starvation * Avoids race conditions * Has a good degree of multiprogramming * Solves the problem ### Bonus: +0 * Has an accompanying implementation / simulation
Once filled in by a teaching assistant, this may look something like this:
## T1: 9/15 ### Formal requirements: 5/5 * Has code + Yes. * Has XML in a reloadable/testable form + Yes. * Has an explanation + Yes. ### Quality assessment: 2/10 * Understands semaphores - Seems to treat them as counters. Report doesn't talk about procuring or vacating at all. - Checks semaphore value directly. - Seems to have mixed up procure and vacate. * Understands deadlocks - If more than 5 cars arrive, then only 5 cars will be allowed to drive through, and everything will then deadlock. * Understands starvation ? Seems to, but this is hard to judge. * Avoids deadlocks ? See above. * Avoids starvation ? Can't judge this. * Avoids race conditions - Uses a busy loop in attempt to synchronize. - There is a race condition after the busy loop. * Has a good degree of multiprogramming + It's a ticket system, so it could be okay. * Solves the problem + In a complicated way, but yes. ### Bonus: +2 * Has an accompanying implementation / simulation + Yes!
Bonus-judgements must have the title
Bonus, but they don't have to be there
if you don't like them. The presence of bonus judgments also allows points to
sum up to more than the given maximum. This is regarded as a feature.
Points and Sums
You do not need to manually enter sums. Points are only required for the
bottom-most judgements. Given the template above you can do a
to check if all necessary points have been given. This makes a "pointless"
template a rather useful starting point for grading.
Points can be integers or half-points, that is, the number may end in
.5 is exactly representable in an IEEE-754 (the internal
representation of points), so no numerical imprecision occurs when dealing with
half-points. Maximum points may only be integers (the internal representation
of maximum points is still an IEEE-754 double to avoid numerical conversions).
Files and Directories
The file-format is kept "git-friendly" by keeping it comprehensible in plain-text, and allowing for independent marking by splitting the remarks for a student into multiple files.
The simplest setup is to have one
.mrk file per student (e.g.
To support more exotic setups,
remarks can also work with directories:
If supplied with a directory path,
remarkslooks for files ending in
.mrkinside that directory, and comprehends the files as above, in lexicographic filename order (e.g., directory-with-mrk-files).
If there exists a directory
<basename>is recursively searched for further
.mrkfiles. Their contents is appended, in lexicographic filename order, to the last top-level judgement of
See the organization samples for some examples of how judgements may be structured using files and directories.
├── basic.mrk ├── directory-with-mrk-files │ ├── 01-theory.mrk │ └── 02-practice.mrk └── mixed-directory ├── 01-theory.mrk ├── 02-practice │ ├── Task1.mrk │ └── Task2.mrk └── 02-practice.mrk
mixed-directory all parse to the
same judgements. In particular, the output from the following
commands is identical:
$ cd samples/organization/ $ remarks parse basic.mrk $ remarks parse directory-with-mrk-files $ remarks parse mixed-directory