Markdown support #274
Markdown support #274
Conversation
|
@pehrsoderman want to take a look? I think this PR is ready for a review. Here are the statements i tested it on: https://github.com/Matistjati/kattis-markdown-examples In general, most statements don't look perfect right off the bat, mostly because they come from their own markdown dialect. However, I believe that all ugliness can be fixed by small tweaks. |
There was a problem hiding this comment.
I took a pass over this. As mentioned on slack, it seems like pandoc makes no security guarantees on html output and recommends using a sanitizer (https://pandoc.org/chunkedhtml-demo/19-a-note-on-security.html). That's gonna get messy, I'm not a huge fan of any of the python options. If anyone happens to look at this PR and has any recommendations, feel free to chime in.
Looking through, I think I spotted a number of places where we need to be more careful with escaping stuff we inject (problem name, sample, and just as an extra precaution, problem id). I'm not quite sure how to do that cleanly, t.b.h. If I understand the code correctly, it seems like when pandoc goes from markdown to pdf, it goes via LaTeX, which means that inserting plain LaTeX in the markdown works. That dual-step translation makes me wonder how to sanely escape stuff like \ or $ when we inject it (or, as an extreme example, the sample of a problem about parsing latex :)).
problemtools/problem2pdf.py
Outdated
| # Add problem name and id to the top | ||
| problem_id = os.path.basename(problem_root) | ||
| statement_md = r'\centerline{\large %s}' % f"Problem id: {problem_id}" + statement_md | ||
| statement_md = r'\centerline{\huge %s}' % problem_name + statement_md |
There was a problem hiding this comment.
We need to escape LaTeX command characters in strings we inject this way. This seems to break if the name contains characters like }, \, or $.
problemtools/statement_common.py
Outdated
| config = verifyproblem.ProblemConfig(prob) | ||
| if not config.check(None): | ||
| raise Exception("Invalid problem.yaml") | ||
| names = config.get("name") |
There was a problem hiding this comment.
This will break after #286, but you're probably aware of that. :)
There was a problem hiding this comment.
Yup, talked with Rasmus and Hugo to hopefully work something out
problemtools/statement_common.py
Outdated
| temp_file.write(sample) | ||
| temp_file.flush() | ||
| command = ["pandoc", temp_file.name, "-t" , "markdown"] | ||
| return subprocess.run(command, capture_output=True, text=True, |
There was a problem hiding this comment.
A bit of a corner case, but what happens if the sample contains latex (or, more likely, latex control characters like $)? I'm assuming pandoc won't escape that when converting HTML to markdown, so won't that then later mess up the PDF rendering once you've injected this markdown into the document and are trying to pandoc into a pdf?
There was a problem hiding this comment.
Painstakingly, this has been fixed.
|
TODO:
Known issues (that will either be wontfix or fixed in follow-up pr):
|
supported for both pdf and html:
$a+b$and$$a+b$$A PR of the spec can be found here Kattis/problem-package-format#386.
TODO testing:
not planned to be added in this PR (maybe ever):