core: Let inferrable constraints solve variables

[alexarice]

Allows an inferrable type to solve other variables, and further allows range variable constraints to be solved if the underlying constraint is inferrable. Includes a test with the motivating example.

Currently this is restricted to when the underlying constraint can be inferred using no variables, and so is likely only useful for the scenario in the test. Anything more involved would be a fair bit more complex.

I also have a branch somewhere with a draft implementation of "length variables", which would probably achieve a similar thing to this PR (and could potentially be extended to be more powerful). Would be good to hear people's thoughts on whether this would be preferable instead of this (slightly hacky) solution.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 91.30%. Comparing base (b83ddb2) to head (6ab3d93).
Report is 28 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #3698      +/-   ##
==========================================
- Coverage   91.31%   91.30%   -0.01%     
==========================================
  Files         467      467              
  Lines       58487    58527      +40     
  Branches     5640     5645       +5     
==========================================
+ Hits        53406    53441      +35     
- Misses       3632     3634       +2     
- Partials     1449     1452       +3     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[superlopuh]

Suggested change

	class _OperandResultInferenceExtractor(VarExtractor[ParsingState]):
	idx: int
	is_operand: bool
	inner: VarExtractor[Sequence[Attribute]]
	constr: RangeConstraint
	class _OperandResultInferenceExtractor(VarExtractor[ParsingState]):
	idx: int
	KW_ONLY
	is_operand: bool
	inner: VarExtractor[Sequence[Attribute]]
	constr: RangeConstraint

[superlopuh]

should help read the code below

[alexarice]

I've never seen this before

[superlopuh]

https://docs.python.org/3/library/dataclasses.html#dataclasses.KW_ONLY

[superlopuh]

I misremembered the API

[alexarice]

I truly hate that API but agree it would make the code better

[superlopuh]

Overall I think it looks good, and it feels like it would be worth having the support. It feels like it would be worth adding some unit tests with explicit NotImplemented catches with corresponding issues to highlight what's missing.

[alexarice]

Overall I think it looks good, and it feels like it would be worth having the support. It feels like it would be worth adding some unit tests with explicit NotImplemented catches with corresponding issues to highlight what's missing.

I'm going to get the length constraint version into shape and can compare to that.