[linter] Support for non-function type aliases

[pq]

Follow-up from #44078 and #44951, we need to shore up linter support for non-function type aliases.

• consider existing lints
  • avoid_private_typedef_functions consider nonfunction type alias support in avoid_private_typedef_functions #58354
  • camel_case_types
  • prefer_mixin
  • public_member_api_docs
  • slash_for_doc_comments
• add test cases
• consider new lints

[pq]

I did a quick look through existing lints and besides the fix to prefer_mixin didn't come up with anything else to address. I didn't feel a strong pull to add test cases either though I'm super open to suggestions.

@srawlins, @bwilkerson, @scheglov : curious if anything jumps out at you?

As for new lints, I'm curious if there's any thinking about naming conventions. CamelCase? If yes, then thoughts on extending camel_case_types? Alternatively I guess we could add camel_case_nonfunction_typedefs 😬 ...

[bwilkerson]

Doesn't camel_case_types already cover the names defined via typedef? I would expect it to just continue to work, but a test for that case would be in order.

As for other places that might need fixed up, you could search for places where staticElement is being invoked on a TypeAnnotation, though that's probably going to return lots of places where it's used on something other than a type annotation.

[pq]

Doesn't camel_case_types already cover the names defined via typedef?

Yes it does cover function type aliases.

The question is whether it should flag the following:

class A { }

typedef a = A;

I think yes?

(Currently it does not.)

Thanks for the tip on staticElement greps!

[bwilkerson]

Ah, that's because it's visiting FunctionTypeAliass when I think it ought to be visiting GenericTypeAliass (it might need to visit both until the experiment is enabled). So you can also look for other implementations of visitFunctionTypeAlias.

[scheglov]

FWIW, this means that the lint did not work for "modern" function type aliases as well, like typedef b = void Function();.
FunctionTypeAlias is not going anywhere, I think, because it is a different syntax than GenericTypeAlias.
FunctionTypeAliasElement is deprecated, and will be removed.

[pq]

Looking at visitFunctionTypeAlias, brought up these suspects:

• avoid_private_typedef_functions
• public_member_api_docs
• slash_for_doc_comments

All seem straightforward. Thanks!

[pq]

Marking as done. If ideas for lints present themselves we can open new tracking issues.