fix(cff): enforce valid urls as doi

[cmdoret]

Addresses a crash caused by DOIs being stored as 10.xxxx/abcd in CITATION.cff instead of doi.org/10.xxxx/abcd.
We now post-process the DOI to make it a valid URL before casting to URIRef.

Additionally:

• handle situations where the CITATION.cff is broken (e.g. invalid yaml) without crashing
• minor style tweaks to the docs.

[rmfranken]

Looks good to me, I leave up to you whether regex is smart and worth it :)

[rmfranken]

This is very readable - so I even dare to make a suggestion - though I'm not sure about the performance considerations: Instead of having 2 if statements taking care of multiple possible doi formats - would it not be simpler to regex the "10.xxxx" part of the doi out of whatever format is found in, and always prefix that with https://doi.org?

that way it can also handle any misspelling in the prefix in the cff (e.g. http://doi.org/)

[cmdoret]

Good point, that's a bit more robust :) I've added regex based matching.
The performance cost is negligible (in gimie time-scale):

long_doi = 'https://doi.org/10.123112/abc.def'
short_doi = '10.123112/abc.def'

# With regex 
In [6]: %timeit doi_to_url_regex(long_doi)
1.58 µs ± 64.4 ns per loop (mean ± std. dev. of 7 runs, 1,000,000 loops each)

In [9]: %timeit doi_to_url_regex(short_doi)
1.45 µs ± 18.3 ns per loop (mean ± std. dev. of 7 runs, 1,000,000 loops each)

# With if
In [7]: %timeit doi_to_url_if(long_doi)
230 ns ± 3.92 ns per loop (mean ± std. dev. of 7 runs, 1,000,000 loops each)

In [8]: %timeit doi_to_url_if(short_doi)
358 ns ± 3.32 ns per loop (mean ± std. dev. of 7 runs, 1,000,000 loops each)

[cmdoret]

To avoid crashing the whole program because of malformatted user-entered data, I've also made it a bit more defensive:

• Invalid yaml -> warn user and extract nothing
• Valid yaml but no doi -> successfully extract nothing
• doi value is not a doi -> warn the user and extract nothing

Does that make sense?
In all cases above, the program resumes successfully. warnings are logged to stderr and will not be mixed with the output.

[rmfranken]

I would only doubt a bit on the

Valid yaml but no doi -> successfully extract nothing

It's not mandatory from CFF, but a doi is the only thing we are looking at a CFF for here - maybe we could just produce a warning in this case too? Basically, the only case in which a warning is not produced is either a user has not added a CFF file, or the CFF file they uploaded contains a valid DOI. But maybe that's being too conservative and patriarchal to our users - as if they are not aware that their CFF doesn't contain a DOI at all.

[cmdoret]

That makes sense! I added a warning in 4d43f42

Now the behavior is:

➜ gimie data https://github.com/opencv/cvat > /dev/null
WARNING :: CITATION.cff does not contain a 'doi' key.

EDIT: Added formatting to the log messages to include level (DEBUG/INFO/WARNING/ERROR)