Combining Generic and NamedTuple

[DrPyser]

Currently, in python 3.7, it seems impossible to properly compose typing.NamedTuple and typing.Generic. That is, I want to do this:

import typing
T = typing.TypeVar("T")

class Test(typing.NamedTuple, typing.Generic[T]):
    a: T

So this works, the class can be used as a NamedTuple.
However:

x: Test[int] = Test(a=1) # TypeError: 'type' object is not subscriptable

Right, subclassing NamedTuple generates a class that inherits only from tuple.

Maybe this was discussed before, I've looked a bit in this repo and haven't found much, except #588, which doesn't really address this.

Is there any plan or possibility for this to be resolved? Maybe more generally, allowing NamedTuple to be composable with other types(I know this might be difficult because of metaclass conflicts).

Thanks!

[DrPyser]

Something of a workaround:

import typing
T = typing.TypeVar("T")
class Test(typing.NamedTuple):
    a: T
class Test(Test, typing.Generic[T]): pass

x: Test[int] = Test(a=1) # works fine

But unsurprisingly mypy won't accept this:

$ mypy /tmp/test.py 
/tmp/test.py:6: error: Invalid type "test.T"
/tmp/test.py:8: error: Name 'Test' already defined on line 5
/tmp/test.py:11: error: Generic tuple types not supported
/tmp/test.py:12: error: Generic tuple types not supported

[lemon24]

It seems this is being worked on in python/mypy#685.

[illeatmyhat]

It's also impossible to do

from typing import TypedDict, TypeVar, Generic

T = TypeVar('T')
class Test(TypedDict, Generic[T]):
    a: T

with error
TypedDict cannot inherit from a non-TypedDict base class

[srittau]

If a type checker doesn't support this, this seems to be more of a type checker issue than a general typing issue, so I'm closing this here.

[Gobot1234]

Typing currently bans this in mro entries

[AlexWaygood]

This is now:

• Supported at runtime in 3.11
• Supported by pyright
• and we have Backport generic NamedTuple and TypedDict support typing_extensions#7 to discuss backports via typing_extensions.

So I think this feature request can be closed as "accepted"!