* Use typing.ContextManager for multiprocessing context managers
Prior to this commit, the types for __enter__ and __exit__ were not
fully defined; this addresses that.
* Move Pool class stub to multiprocessing.pool
This is where the class is actually defined in the stdlib.
* Ensure that __enter__ on Pool subclasses returns the subclass
This ensures that:
```py
class MyPool(Pool):
def my_method(self): pass
with MyPool() as pool:
pool.my_method()
```
type-checks correctly.
* Update the signature of BaseContext.Pool to match Pool.__init__
* Restore multiprocessing.Pool as a function
And also add comments to note that it should have an identical signature
to multiprocessing.context.BaseContext.Pool (because it is just that
method partially applied).
This change modifies the PIPE, STDOUT, and DEVNULL constants in the
subprocess module to be of type 'int'.
Note that the Python 2 subprocess module was already typed this way;
this commit is just updating the Python 3 subprocess module in the same
way.
* Make configparser.RawConfigParser.__getitem__ return a SectionProxy
This reflects the code in the cpython tree and makes the following
(valid) code type-check correctly:
```
from configparser import ConfigParser
config = ConfigParser()
config.read_dict({'section': {'key': 'false'}})
assert config['section'].getboolean('key') is False
```
* RawConfigParser.items() returns SectionProxys not mappings
Because .items() uses __getitem__ to produce second item in each tuple.
* section argument to RawConfigParser.items is Optional
* Add comment explaining the status of RawConfigParser.items
TL;DR, we're hitting https://github.com/python/mypy/issues/3805 when
implemented correctly as an override, and
https://github.com/python/mypy/issues/1855 when we try to work around
that with a union.
* Correctly implement RawConfigParser.items overloading
* RawConfigParser.items(str) returns a List not an Iterable
RawConfigParser.read has code explicitly supporting PathLike objects.
Also removed an unused import and widened the accepted type from
Sequence to Iterable.
* Add Protocol and runtime to typing_extensions
* Use private type variable
* Fix typo
* Bound type variable for runtime to only class objects
* Conform to version check specification
In Python 3 (but not Python 2), `object().__dir__()` works and returns a list of strings.
This is relevant when implementing a custom `__dir__` that invokes `super().__dir__()`.
Also change multiprocessing.Queue's put and get timeout arguments to
allow None.
This fixes a problem with logging.handlers.QueueHandler and
QueueListener not accepting a multiprocessing.Queue as the queue
argument.
Declaring the Queue now needs to note what it will be used for. eg.
q = multiprocessing.Queue() # type: multiprocessing.Queue[List[Any]]
E.g. it's legal to call traceback.format_exception(None, None, None). In
particular, this change makes the following idiom type-check:
import traceback
import sys
exc_type, exc_value, exc_traceback = sys.exc_info()
traceback.format_exception(exc_type, exc_value, exc_traceback)
After mypy [started hiding](https://github.com/python/mypy/pull/3706) imported names in stubs unless `from ... import ...` is used, we found an error with stubs of ast module.
It looks like ast module should re-export everything in `_ast` and according to PEP 484, it can do that by doing `from _ast import *`, so this is what this PR does.
This commit disables following flake8 errors:
* F403: `‘from module import *’ used; unable to detect undefined names`
* F405: `name may be undefined, or defined from star imports: module`
We don't need to worry about undefined names since running mypy_test.py
should detect such undefined names.
* Added stub for sre_parse(py3)
* Fixed return type of SubPattern.__getitem__
* Typo
* Fix for issue related to error class
* Added stub for sre_constants(py3)
* Added missing import
