refresh sage-fixdoctests using argparse

[chapoton]

see also #31366 and #32332

also full flake8 cleanup of the modified file

[chapoton]

New commits:

​7f4b0c0

refresh sage-fixdoctests using argparse

[tscrim]

One thing that could be considered a regression is that if I do something like

./sage --fixdoctests src/sage/combinat/q_analogues.py src/sage/combinat/rsk.py src/sage/combinat/fqsym.py

then I now get that rsk.py is an empty file, whereas the old behavior left rsk.py alone. So what we get is a new file with no output when the is erroneous input passed. There might not be a way around this because that is how argparse works, but I wanted to ask before giving the go-ahead with this.

[chapoton]

ok, I see.

We could introduce an option -o, --output and try to handle multiple input files.

Or stay with the principle that this is to fix just one single file ?

[tscrim]

I think it is fine to handle just one file, which I believe is documented. It is more if someone gives it bad input (and it realizes this) that it doesn't change any files. However, this isn't really a big deal. I think the current branch is fine, but I wanted to see if there was an easy fix to prevent this from happening.

[chapoton]

I do not see how to distinguish between the incorrect use

sage --fixdoctests a.py b.py

where the intention is to fix both a.py and b.py inplace ; and the correct use

sage --fixdoctests a.py b.py

where the intention is to write a corrected copy of a.py in b.py.

In the second case, it could very well be that b.py is existing already, as in the first case.

I think one can reasonably keep the current behaviour, which is indeed documented.

[tscrim]

I agree that we cannot distinguish between improper use and proper use. What I was saying is if someone does improper use like

sage --fixdoctests a.py b.py c.py

that b.py is not overwritten with a blank file (previously it did not). However, I doubt we can easily fix this, so I am setting this to a positive review.