Type-annotate all function arguments and internal variables

[ShadowJonathan]

Description:

Related to #8347, (partially) blocked till 3.5 is deprecated

The goal is to;

• Convert existing name = ... # type: comments to name: type = .... (see "com2ann progress" below)
• Put type annotations on every function's parameters and return type.
• Put name: type annotations on;
  • Every class variable.
  • Every instance variable (declared in __init__() functions), unless;
    • They're constructing another instance (self.thing = Thing()) (dubious, see Use inline type hints in handlers/ and rest/ #10382 (comment))
    • They're getting assigned a non-container literal (1, b"", "", etc. but not None)
    • The assignment comes from a function/method call (self.thing = hs.get_another_thing()), given that the function signature is known and inferrable.

---

com2ann progress

• tests/ (Use inline type hints in tests/ #10350)
• synapse/
  • Use inline type hints in handlers/ and rest/ #10382
  • Use inline type hints in http/federation/, storage/ and util/ #10381
  • Use inline type hints in various other places (in synapse/) #10380
• Other (One last inline type hint (for the whole repo) #10418)

(com2ann changes comment type annotations to inline type annotations, name = ... # type: to name: type = ..., this is possible since python 3.6, and standard across the python ecosystem.)

[clokep]

This could probably finally allow a mypy linter to run with CI, to catch future inconsistencies.

mypy runs in CI on a subset of the code-base, see https://github.com/matrix-org/synapse/blob/develop/mypy.ini#L9

[clokep]

blocked till 3.5 is deprecated

This isn't blocked? Python 3.5 supports type annotations.

[ShadowJonathan]

blocked till 3.5 is deprecated

This isn't blocked? Python 3.5 supports type annotations.

not variable type annotations, but you're right about it not being completely blocked

[ShadowJonathan]

A large portion of this can be fixed with com2ann, see #9131 (comment)

[ShadowJonathan]

Updated description, the previous one was vague and also implied setting type annotations for every variable inside non-init functions, which i think is overdoing it a little bit, type-annotating all other places noted should already correctly resolve those spots.

[ShadowJonathan]

I found some interesting mypy optional errors that could be enabled to enforce type annotations;

https://mypy.readthedocs.io/en/stable/error_code_list2.html

• Check that type arguments exist [type-arg]
  (Errors on List, requires List[<type>])

• Check that every function has an annotation [no-untyped-def]

• Check that comparisons are overlapping [comparison-overlap]

  def is_magic(x: bytes) -> bool:
    # Error: Non-overlapping equality check 
    #        (left operand type: "bytes", right operand type: "str")  [comparison-overlap]
    return x == 'magic'

• (maybe) Check that no untyped functions are called [no-untyped-call]

• Check that function does not return Any value [no-any-return]

[clokep]

I think this is essentially a duplicate of #11271.