gh-102542 Refactor the mime audio module

[JosephSBoyle]

Refactor the MIME audio module, the mechanism for attempting to infer a subtype is currently quite verbose.

Notes:

• fakefile is not used in any of the functions it is passed to, hence it has been removed.
• Slicing the header (hrd = data[:512]) isn't necessary so has been removed.

• Issue: Refactor the MIMEAudio module #102542

[terryjreedy]

This changes the behavior if _subtype is '' (or any other non-None null value). Someone other than me needs to agree with this.

[JosephSBoyle]

That's correct. Falsey values other than None such as: 0, False, '' and empty collections will all fall back to the second or operand: _infer_subtype(_audiodata).

[warsaw]

I personally prefer more explicit tests than just "falseiness" when a more explicit type is indicated. In this case _subtype=None defaults to None to mean that this argument wasn't given.

[terryjreedy]

@warsaw This PR mostly replaces the section you committed April '22. I believe that replacement logic is correct.

Joseph, have you checked the coverage of the mime audio test?

[JosephSBoyle]

@warsaw This PR mostly replaces the section you committed April '22. I believe that replacement logic is correct.

Joseph, have you checked the coverage of the mime audio test?

None of the tests execute this path as far as I could tell from putting a breakpoint in and running the tests.

I convinced myself that the logic in _infer_subtype was correct by just running the script in a REPL and evaluating the output of different input strings.

[JosephSBoyle]

Force pushed to correct a mistake in the commit message.

[AlexWaygood]

None of the tests execute this path as far as I could tell from putting a breakpoint in and running the tests.

That implies that we should perhaps first work on improving test coverage, before undertaking a risky refactoring of this module.

("Risky" only in the sense that all refactorings are risky if test coverage is low :)

[JosephSBoyle]

That implies that we should perhaps first work on improving test coverage, before undertaking a risky refactoring of this module.

You're right as usual;) Turns out adding the tests was easier than I first thought.

Apologies for rebasing rather than merging, I'm still working on using that workflow @AlexWaygood

[JosephSBoyle]

Update: I was incorrect about this path not being tested - there are in fact tests for this, see: TestMIMEAudio::test_guess_minor_type.

As such I've deleted the redundant ones.

[JosephSBoyle]

@terryjreedy could you have a re-review when you get a chance please.

I think the minimum set of changes for this PR would be simply removing the unused fakefile that is passed to each function as this isn't used whatsoever. If the other changes don't seem worthwhile let me know and they can be reverted.

[terryjreedy]

I have nothing to add, which was that I will not approve the change as this is outside my expertise.

[JosephSBoyle]

Hia @AlexWaygood. Who I should request a review from on this? I have the feeling it may have slipped through people's todos!

Cheers

[warsaw]

I've provided some additional thoughts on #102542

[JosephSBoyle]

Closing as we've decided to leave the @rule decorator for extensibility reasons; see the issue for discussion.