feat: better message if method incompatible with base class

[atakiar]

Description

Closes #3379. See PR #3410 for some context

Currently if a method is incompatible with a base class, mypy just states the name of the base class but doesn't actually give more details about the expected signature

This PR makes it so that incompatible methods will print out easily-readable signatures, like follows:

error: Signature of "f" incompatible with supertype "A"
note:      Superclass:
note:          def f(self) -> str
note:      Subclass:
note:          def f(self, x: str) -> None

Test Plan

Updated the relevant test cases. No new test cases were added

[atakiar]

A couple interesting observations with the existing behavior of CallableType and mypy/messages.py's pretty_callable (below)

mypy/mypy/messages.py

Lines 1932 to 1949 in 10ea6ad

	# If we got a "special arg" (i.e: self, cls, etc...), prepend it to the arg list
	if isinstance(tp.definition, FuncDef) and tp.definition.name is not None:
	definition_args = tp.definition.arg_names
	if definition_args and tp.arg_names != definition_args \
	and len(definition_args) > 0:
	if s:
	s = ', ' + s
	s = definition_args[0] + s
	s = '{}({})'.format(tp.definition.name, s)
	elif tp.name:
	first_arg = tp.def_extras.get('first_arg')
	if first_arg:
	if s:
	s = ', ' + s
	s = first_arg + s
	s = '{}({})'.format(tp.name.split()[0], s) # skip "of Class" part
	else:
	s = '({})'.format(s)

First, when printing B.f in testInvalidOverrideArgumentCountWithImplicitSignature2 and I.g in testAbstractClassWithAllDynamicTypes, the output of both did not contain "self". Upon some investigation, isinstance(tp.definition, FuncDef) is False and then tp.def_extras.get('first_arg') is None. I'm guessing this might be since both have implicit return types and that might be interacting somehow with CallableType?

[case testInvalidOverrideArgumentCountWithImplicitSignature2]
...
class B:
    def f(self, x, y): pass
...
main:5: note:      Superclass:
main:5: note:          def f(x: Any, y: Any) -> Any
-- expected this instead? def f(self, x: Any, y: Any) -> Any

Second, I had to split testOverloadedMethodSupertype and testFinalBodyReprocessedAndStillFinalOverloaded into an only_when_cache and only_when_nocache version. Similar to what I described above, in the only_when_nocache versions, isinstance(tp.definition, FuncDef) is False. However, in the only_when_cache versions, it is True and "self" is included in the output

Not sure if this is expected behavior or a bug. Let me know what you think!

[msullivan]

@atakiar Could you look into some of the mypy_primer errors? I'm a little nervous about all of the missing errors, though it wouldn't shock me if some of them had been wrong?

[atakiar]

@msullivan I think new commits made their way to master between my last merge and when the mypy_primer workflow ran (triggered manually for first-time mypy contributors). Trying this again, hopefully it’ll come out looking better this time!

[atakiar]

@msullivan let me know what you think!

By the way, the missing mypy_primer errors are the result of #10579. The new multiline note along with the restriction on number of errors shown, results in the last few errors now being hidden

[github-actions]

Diff from mypy_primer, showing the effect of this PR on open source code:

aioredis (https://github.com/aio-libs/aioredis.git)
+ aioredis/client.py:4286: note:      Superclass:
+ aioredis/client.py:4286: note:          def (self: Redis) -> Any
+ aioredis/client.py:4286: note:      Subclass:
+ aioredis/client.py:4286: note:          def __aenter__(self) -> Coroutine[Any, Any, Pipeline]

[msullivan]

OK, this is absolutely fantastic. Thanks so much for this!