Meta issue tracker, pt 2

[hauntsaninja]

This is the sequel to #11201. It's awesome to see most of the things in that issue fixed.

Type System Features

• PEP 695 Support for PEP 695 #15238
• PEP 702 Add support for PEP 702 (@deprecated) #16111
• PEP 675 Support PEP 675 (LiteralString) #12554
• PEP 696 Support PEP 696 – Type defaults for TypeVarLikes #14851

Behaviour changes

• Ternary unions Use union instead of join for unifying types #12056

  We should use unions in some places we currently join, see Use union instead of join for unifying types #12056 (comment)

  Update: We now use unions in more places, including ternary expressions. Feel free to investigate this more though!

• Always narrow types Types not added to binder on initial assignment #2008 (comment) (proposal 3)

  Users find it surprising because it seems obvious that mypy should know the variable is of the narrower type. There was a big discussion in typing-sig and all the type checkers agreed this behaviour was probably the best. The change is very easy, but exposes bugs in binder

• Default value of generic parameters Unable to specify a default value for a generic parameter #3737 (comment)

  It's totally reasonable to want to have generic parameters with defaults, we should error at call sites in case of badness

• Support narrowing Literal Narrowing types to Literal using in syntax #12535

  It's hard to really use Literal types without a good way to narrow them. We should also narrow on ==. I think last time I tried this it exposed bugs in binder

• Support property setters with broader types What to do about setters of a different type than their property? #3004

  Should also allow covariant overrides of getter

• Better handling of __new__ Believe more __new__ return types #16020

  This is part of the Python language and comes up weirdly frequently.

• Love for match statement https://github.com/python/mypy/issues?q=is%3Aopen+label%3Atopic-match-statement+sort%3Areactions-%2B1-desc

  I fixed the most popular of these issues, but there's still a long tail of them

• PEP 688 Drop bytes/bytearray and bytes/memoryview promotions (PEP 688) #15313

  Easy, is just a somewhat breaking change

UI

• Per module strict=true bug: strict = true is incompatible with per-module configuration #11401

  There's maybe some other UI stuff we can think about for groups of flags and error codes, as discussed in meetup. I also don't like the new --extra-checks since flag name doesn't mean anything and can't be decomposed.

• Ability to override PEP 561 Allow/disallow pep561 packages #8545

  It would be nice to be able to opt-in to treating packages missing py.typed as having py.typed. mypy can still provide a lot of value just based on function shape.

• Exclude should imply per-module follow_import=skip exclude should imply a per-module follow import skip for matching modules #10377

  Currently --exclude only affects file discovery on command line, not import following. This is a source of frustration for users because the relevant modules can still end up being type checked.

• Provide flag to generate stubs requirements (🎁) Improve usability of --install-types #10600 (comment)

  --install-types is slow, results in unstable type checking, doesn't let users manage their own envs, can fail unclearly if cache isn't written

• Rethink stub suggestion logic Improving and evolving stub package suggestions #10619

  Our current behaviour was needed to help the migration to modular typeshed, but is not behaviour we'd have designed from scratch.

Gradual typing

• Make check_untyped_defs = True the default Give an error if there is a variable annotation within a function but no signature #3948 (comment) / Should we eventually type check all function bodies by default? #6646

  We decided not to do this for mypy 1.0, but could still make sense in the future.

• Infer obvious return types Proposal: treat "obvious" return type as annotated #4409 / Proposal: infer obvious parameters and return types #10149

  As a 1/3-way measure, we could also do (🎁) Flag to make default return type None instead of Any #9413

• Infer method types from superclass untyped defs on derived classes #3903 (comment)

Performance

• Parallel type checking Faster, parallel type checking #933
• Faster caching Faster cache deserialization #3456

Misc

• Member access Refactor all member access to go through checkmember.py (meta issue) #7724

• Merge configuration docs Consider merging command line flags and configuration file options descriptions #12347

See also a similar issue for mypyc here: mypyc/mypyc#785

(other maintainers, feel free to edit)

[JukkaL]

Added performance section with two issues.