Directory-level regex matching in MOSS

[RONNCC]

Can you add support for file or directory level matching with full regex's in the Moss submission script in the style of a .gitignore.

E.g. if you have (after extracting) /a/b/c/d.txt you can use a regex to say /c/d.txt or */e/ only . Similarly you could filter on the files as a result. Also the ability to do negative searches - e.g. regex does not match, or support for negative lookaheads would be useful.

Hence you could exclude files_given_to_student/*.py

[dlbucci]

I'm confused. We have a filename field that appears under the assessments you choose to run moss on, and I think that already does exactly what you're describing (minus the negative regex stuff). That filename is what is passed to the mossnet call, so the files that get fixed are really up to that program.

[RONNCC]

sorry i wasn't sure if it was bash style pattern globs - or full regex's.

I'm not sure how to 'exclude' files without a full regex :/ also this works on a per file level I believe rather than being able to exclude whole directories

[dlbucci]

I'm pretty sure moss is only useful if the files you send are all the same/serving the same purpose. Like, if you have an assessment with two files, foo.py and bar.py. It's not gonna make sense to compare foo.py files to bar.py files.

[RONNCC]

yes but if they turn in a directory of files, and we gave 2 of those to them (hence everyone has it), that's not very useful with respect to cheat comparisons and causes false positives

[dlbucci]

Right, but you shouldn't really be sending entire directories of files anyway, right? Is there a specific answer.py file that everyone is filling out? Just send that.

[RONNCC]

uhh.. no Moss seems to be fine with it / has it as an example - and nope just like 112 they submit a bunch of files - putting everything in one file being bad style

[dlbucci]

Do you know the file names? Like will there always be a problem1.py, problem2.py, etc?

[RONNCC]

nope ~ there's no regularity to it

[dlbucci]

I'll look into ways we can accommodate this, but I'd suggest putting course files into different directories/not having students submit those files for future assessments. How do you guys autograde assessments if there's no regularity to the files that are submitted?

[RONNCC]

Yes I think this uses the built in generic bash globbing (correct me if i'm wrong) - i would prefer full regex support - e.g. negative lookaheads to ignore all something_test.py files

[zyx-billy]

Yes, the user input is passed as argument directly, so standard bash globbing is performed on it. It doesn't seem hard to implement regex if that turns out to be useful. We can just perform the check as we extract files from the student's submitted archive (in CoursesController::extract_asmt_for_moss).

One odd thing though: the directory structure is flattened when the archive is extracted ('/' replaced with '-'). This isn't mentioned anywhere on the moss page, which means the paths users entered likely won't do what they'd expect. Would be great if someone knows why this was done in the first place when implemented in #357.