* complete doctest stub
And merge it into 2and3 (the old stubs were virtually empty).
* note things changed in 3.4
2.7 also had an undocumented and deprecated doctest.Tester class. I'm going to leave it out until somebody asks for it to be added.
* fixes and additions to the configparser stub
This broke pytype because of the type ignore. I think it's OK to blacklist
because another instance of this same error (stdlib/3/email/policy.pyi line 95)
is already blacklisted. We can remove it when the pytype parser is fixed to
allow type ignores in this context.
* add SectionProxy.__getattr__
* the __getattr__ returns Any
* don't need type ignore for pytype if we put it all on one line
* dummy_thread improvements
- Complete the Python 3 stub
- Add a Python 2 stub. They're close enough that they should go into 2and3, but
the module name changed so we can't do that.
* fix my bugs
- ThreadError exists (undocumented) on Python 3. It's an alias for _thread.error,
but making it a separate exception seems fine.
- zipfile.error is an alias for BadZipFile on both Python versions.
- zlib.Compress and Decompress are not actually accessible at runtime.
- Removed class definitions from `xml/__init__.pyi` as they were merely outdated duplicates of the definitions from the correct file (`xml/sax/__init__.pyi`)
- Left file intact as it is necessary for the module
- Moves the pdb stub from separate `2`/`3` files to a single file in `2and3`, as they were identical.
- Maintains the comment header, as the stub is incomplete.
- Updates documentation related to previously required comment headers.
- Removes all comment headers from stubs
- Occasionally included a header for stubs that were noted to be incomplete or contained todo's.
* sha3 family of hash functions
* shake variable length hashes that were introduced alongside sha3
* blake2 family of hash functions
* Mark VarLenHash as a private class, since it's not actually exported by hashlib
* Fix typos in my last, and mark hashlib.Hash as a private class since it isn't in the real hashlib
* Add __getattr__ to ModuleType.
Modules can contain anything, so give them a generic "__getattr__(name) -> Any".
This makes code like the following type-check, in pytype:
sys.modules.get("random").randint
* undo Python 3 change
* Revert "undo Python 3 change"
This reverts commit 96cf2d529e374a754f1ba681f0fc81aab25f81ef.
* Revert "Add __getattr__ to ModuleType."
This reverts commit 3ac1cf8a3b25dabfe386fbf34aef680fb1acd112.
* In stdlib/2/, make sys.modules a Dict[str, Any].
(Instead of Dict[str, ModuleType])
Same as stdlib/3/.
* Allow None in concurrent.futures.exception() and set_exception()
* Make Executor.map() signature more precise
* Remove superfluous signatures
* Specify str type for some concurrent.futures constants
* Update concurrent.futures backport
* CR fixes
* Adds Event type to multiprocessing
* Add event and context stub to multiprocessing stub. Updates per recommendations.
* Adding missing newline at EOF.
* Stubbing just BaseContext and not any of the member functions.
* Fix flake8 extra line.
* Fix comment.
* Add todo for BaseContext, make ctx of Any type, and remove multiprocessing.context
From Samuel Freilich:
In Python 2, the predicate parameter in itertools.ifilter and
itertools.ifilterfalse can be None, indicating that true or false values
should be retained (functionally equivalent to passing "bool" as the
predicate). In Python 3, filter and itertools.filterfalse have
the same behavior.
* add typing.ContextManager for 3.6+ only
This fixes the easier part of #655.
Would it make sense to add a generic typing.ContextManager that exists in any Python version?
* update comment
* fix argument types for ContextManager.__exit__
* add AsyncContextManager
* add @asynccontextmanager
* typing.ContextManager now always exists
* back out async-related changes
Will submit those in a separate PR later
* fix import order
* AbstractContextManager only exists in 3.6+
* AbstractContextManager -> ContextManager
