Unexpected result with `NamedTuple(name, zip(keys, values))`

[randolf-scholz]

Bug report

Bug description:

from typing import NamedTuple

field_names = ["x", "y"]
field_values = [int, int]
Point = NamedTuple("Point", zip(field_names, field_values))
Point(1, 2)  # TypeError: point.__new__() takes 1 positional argument but 3 were given

CPython versions tested on:

3.14

Operating systems tested on:

Linux

Linked PRs

• gh-144321: Fix named tuple bug when input is a non-sequence iterable #144600

[randolf-scholz]

The problem is that if fields is a generator, it gets exhausted in the types assignment.

cpython/Lib/typing.py

Lines 3045 to 3047 in 6543720

	types = {n: _type_check(t, f"field {n} annotation must be a type")
	for n, t in fields}
	field_names = [n for n, _ in fields]

so a fix would be to change this into:

types = {n: _type_check(t, f"field {n} annotation must be a type") 
          for n, t in fields}
field_names = list(types)  # using order preservation of dict keys

[serhiy-storchaka]

cc @rhettinger

[rhettinger]

This would be a nice improvement.

[bkap123]

@randolf-scholz I was wondering whether you’re planning to create a PR, or if you would be okay with me doing it.

[randolf-scholz]

@bkap123 Feel free to go right ahead, I am busy with other things.

[JelleZijlstra]

I'm happy to accept the change but I'd prefer we treat it as a new feature, not a bug.

[AlexWaygood]

I'm happy to accept the change but I'd prefer we treat it as a new feature, not a bug.

The previous behaviour seems surprising to me, and the change seems low-risk to backport, so I'd be happy to backport it personally. But I also don't feel strongly; I'm happy to close this out now that the fix has been merged to main.

[bkap123]

I don’t have a strong preference either; happy to go with whatever you think is appropriate.