Starting with python/mypy#2521 mypy is performing stricter function signature
checks.
This makes the stubs diverge from the actual implementation but makes the stubs
internally consistent. Since this is an actual typing issue in the base
implementation, we need to defer to the original authors to fix it.
Starting with python/mypy#2521 mypy is performing stricter function signature
checks.
This makes the stubs diverge from the actual implementation but makes the stubs
internally consistent. Since this is an actual typing issue in the base
implementation, we need to defer to the original authors to fix it.
Starting with python/mypy#2521 mypy is performing stricter function signature
checks.
This makes the stubs diverge from the actual implementation but makes the stubs
internally consistent. Since this is an actual typing issue in the base
implementation, we need to defer to the original authors to fix it.
This simplifies running flake8 tests and reduces the amount of F821 errors
reported (flake8-pyi enables support for forward references in *.pyi files).
The error code is left disabled until I clean up the remaining issues.
Ran both by Travis and locally. There's some setup required, README updated.
A few important Flake8 checks are still disabled, we're going to enable them as
soon as the stubs are fixed and we can reliably run Flake8 locally with Python
3.6.
This is currently disabled due to the high number of errors that need fixing.
I'll be committing fixes for them and updating the numbers in .flake8 until we
can enable linting for all files by default.
* Define __slots__ for object as Iterable[str] / Iterable[Union[str, unicode]]
* A string as __slots__ value is also valid and represents a single item
Correctly detects calls to `register()` with a function of incompatible return
type. Correctly recognizes the `register()`, `dispatch()`, and
`_clear_cache()` methods on a generic function, as well as the `registry`
mapping.
Possible future improvements: it would be amazing if `register()` checked if
the first argument of the registered callable is indeed of valid type. This
would require Callable[] to support varargs. It would also be great if we
could read the arguments of the remaining arguments during `@singledispatch()`
and cross-check them during `register()` with the currently registered
implementation. Again, this would require Callable[] to become much more
advanced.
Several asyncio AbstractEventLoop methods take a callback as
an argument that is passed *args. The Callable definition was
incorrect for those callbacks.
