Parsons problem files repeat details

[nmstoker]

I may be missing some sensible reason, but it seems like the Parsons problem files repeat various details and thus aren't "DRY".

There's a risk that the tests explained in the problem description in the .yaml file end up differing from those included in the .py header.

I wonder if it might be feasible to simply do away completely with the .py header - after all, it isn't even a complete stand-alone Python script - and then simply rely on the details in the .yaml file.

I see there are slight (intentional) differences in the problem description from the .yaml file and the .py docstrings anyway, and the ability to add text/formatting you wouldn't likely put into a docstring seems useful, so those details seem like they should remain in the .yaml file along with the doctest text they contain and then it's simply a matter of using that for feeding the doctest process.

Hope that makes sense and that I haven't completely misinterpreted how the code is intended to be 🙂 Happy to have a go at this with an initial PR to show how this could be achieved, if you like?

[nmstoker]

Also, by using the .py header it adds another limitation:

It means that Parsons problems that incorporate a blank section in the def/class section can end up passing no matter what people enter, since it's ignoring the solutions lines for that part and instead getting lines from the header.

[pamelafox]

Yes, they do! That has to do with our original implementation with the Berkeley OKPy grader. I think we could get away with only the YAML if we were quite strict with how the doctests were declared in the YAML. Perhaps a YAML checker would be in order? Initial PR would be great. So sorry I missed these issues, I apparently wasn't on full notifications for this repo. So s