Fix type hints

[benoit74]

Fix #109 Fix #112

Rationale

Type hints were too restrictive / incorrect

Changes

• Item
  • for **kwargs, the type specified must be the one of individual param values, not the whole dictionary
  • the Union[str, bool, bytes] type hint was replaced by Any, because the goal of this param is precisely to pass any type which could be further used
  • FileItem and URLItem were modified as well because they were very close to StaticItem
  • no change was necessary in StaticItem in fact
  • references to callback in docstring has been removed because it has been moved to the add_item method quite few years ago
• convert_image
  • src and dst can be either a Pathlib.Path or an io.BytesIO (maybe even typing.BinaryIO or typing.IO[bytes], but I did not find a convenient way to test this to confirm that adherence to typing.IO[bytes] contract is sufficient or if we need something additional from io.BytesIO)
  • for **params, the type specified must be the one of individual param values, not the whole dictionary ; based on existing codebase across all scrapers I assumed that str is sufficient.
  • save_image had to be changed as well to reflect changes made in convert_image

[codecov[bot]]

Codecov Report

All modified and coverable lines are covered by tests :white_check_mark:

Comparison is base (5d95d59) 100.00% compared to head (cd25285) 100.00%.

:exclamation: Current head cd25285 differs from pull request most recent head 824504f. Consider uploading reports for the commit 824504f to get more accurate results

Additional details and impacted files

```diff @@ Coverage Diff @@ ## main #132 +/- ## ========================================= Coverage 100.00% 100.00% ========================================= Files 32 32 Lines 1345 1370 +25 Branches 229 237 +8 ========================================= + Hits 1345 1370 +25 ```

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.

[benoit74]

Changes done

Nota: please do not merge, I will rebase/push-force before merge first since another PR has been merged in the mean time.

[benoit74]

I don't know ... you are right, but somehow this is the price to pay to have proper typing in the whole class, otherwise we will still rely on get_attr and mostly untyped values.

Should we give it a try or rollback the change?

[benoit74]

By rollbacking the change I mean changing only the method signatures, it is already a step forward.

[rgaudin]

Yes the method signature is what's most important because that's the public API.

Could you use the getattr way for the optional stuff in an independent commit so we have a better view at both options?

[benoit74]

I pushed one more commit to "rollback" what is necessary to keep the get_attr approach while still enhancing the public API.

I tend to admit I prefer this "middle-ground" approach.

[benoit74]

Nota: last push is just a rebase on main last commit.