Improve pypdfium2 integration again

[mara004]

pypdfium2 now implements automatic closing on garbage collection (since v3.3).
Using documents in a context manager has been deprecated to avoid inadvertent closing in wrong order.

Also removed a type check which, I think, is unnecessary. pypdfium2 already checks the input and raises a TypeError accordingly.

[codecov]

Codecov Report

Merging #1096 (8fdce28) into main (3d7c812) will increase coverage by 0.01%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##             main    #1096      +/-   ##
==========================================
+ Coverage   95.19%   95.21%   +0.01%     
==========================================
  Files         142      142              
  Lines        6018     6017       -1     
==========================================
  Hits         5729     5729              
+ Misses        289      288       -1     

Flag	Coverage Δ
unittests	95.21% <100.00%> (+0.01%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
doctr/io/pdf.py	91.66% <100.00%> (-0.65%)	⬇️
doctr/transforms/functional/base.py	95.65% <0.00%> (+1.44%)	⬆️

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

[felixdittrich92]

Thanks @mara004 LGTM 👍

[felixdittrich92]

Could you push an empty commit please to trigger CI again ? :)

[mara004]

Could you push an empty commit please to trigger CI again ? :)

Done. I probably shouldn't have skipped CI in the first place, but thought it kind of a waste of energy to re-run all the checks just because of black formatting...

[felixdittrich92]

Could you push an empty commit please to trigger CI again ? :)

Done. I probably shouldn't have skipped CI in the first place, but thought it kind of a waste of energy to re-run all the checks just because of black formatting...

Yeah i totally agree the thing is that @charlesmindee is the only one currently with repo maintainer rights so i can merge it only if the CI pass 😅

[mara004]

Sure, I see :)

[mara004]

Looks like train-char-classification failed again for unrelated reasons.

[felixdittrich92]

FYI you do not need to change any code if you only want to trigger CI you can use:
git commit --allow-empty

:)

[felixdittrich92]

Looks like train-char-classification failed again for unrelated reasons.

That's ok we know the issue and have it on track this job fails everytime currently (not a required check for merge) 👍

[mara004]

git commit --allow-empty

Thanks, didn't know that yet!

[felixdittrich92]

Thanks for the update 🤗

[mara004]

Thanks for the quick review/merge! Hope I don't annoy you with the frequent changes 🙃

[felixdittrich92]

No, I really appreciate that. 👍
We just have to be careful to keep it backwards compatible to avoid problems like the render_to_pil breaking change we have had. 😅

[mara004]

Yeah, in this case the previous code stills work of course, and should continue to do so until pypdfium2 4.0 at least (if that ever happens).
However, the new code now requires a minimum of pypdfium2 3.3 because of the automatic closing thing.

So what matters is basically

• me not doing backwards incompatible changes within one major version
• doctr restricting the allowed pypdfium2 versions accordingly