Allow class-based NamedTuple syntax in semantic analyzer for stubs in pre-3.6 mode

Not very important, but would be nice if stubs could use

class A(NamedTuple):
    x: int
    y: str

instead if

A = NamedTuple('A', [('x', int), ('y', str)])

Had this come up in python/typeshed#3278.

This seems to be the relevant code (in mypy/semanal_namedtuple.py):

    def check_namedtuple_classdef(
            self, defn: ClassDef) -> Optional[Tuple[List[str], List[Type], Dict[str, Expression]]]:
        """Parse and validate fields in named tuple class definition.

        Return a three tuple:
          * field names
          * field types
          * field default values
        or None, if any of the types are not ready.
        """
        if self.options.python_version < (3, 6):
            self.fail('NamedTuple class syntax is only supported in Python 3.6', defn)
            return [], [], {}
...

Also, there is is_typeshed_stub_file() in SemanticAnalyzerInterface which isn't quite what we need. A similar property could be added for is_stub_file(), and then the rest should be easy.

Anybody want to write a PR?

Support NamedTuple class syntax in all stubs

PR in #7646.

Support NamedTuple class syntax in all stubs (#7646)

Do we know if this is supported by type checkers other than mypy?

We checked this in python/typeshed#3321 and pytype, pyre, and pyright, PyCharm, and the Microsoft Language Server for Python all support this.

Navigation Menu

Repository navigation

Allow class-based NamedTuple syntax in semantic analyzer for stubs in pre-3.6 mode #3464

Metadata

Assignees

Labels

Projects

Milestone

Relationships

Development

Allow class-based NamedTuple syntax in semantic analyzer for stubs in pre-3.6 mode #3464

Description

Activity

srittau commented on Oct 7, 2019

JukkaL commented on Oct 7, 2019

srittau commented on Oct 7, 2019

gvanrossum commented on Oct 7, 2019

srittau commented on Oct 30, 2019

Metadata

Metadata

Assignees

Labels

Projects

Milestone

Relationships

Development

Issue actions