Fix namedtuple discrepancies

[elazarg]

• It should be possible to call _make on an instance, as a proper classmethod (?)
• _make and _replace should bind to SelfType, depends on SelfType or another way to spell "type of self" (or, How to define a copy() function) #1212
• _asdict should return closer match to OrderedDict[str, Union[types]]
• _make params new and len should be callable
• Handle locally defined namedtuple

[dlinnemeyer]

I'm working on trying out NamedTuple from typing with mypy, and I ran into an error claiming my tuple has no attribute "_replace" when I try to call _replace.

Is this what you're talking about here, or am I running into a different problem?

[elazarg]

What version of mypy do you use? This sounds like an old bug that should have been fixed. Can you post an example code?

[dlinnemeyer]

0.4.4. I'll work on some example code quickly.

[dlinnemeyer]

from typing import NamedTuple

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

def do_something(foo: Foo) -> Foo:
    return foo._replace(x=7)

I'm getting this error:

test.py: note: In function "do_something":
test.py:9: error: "Foo" has no attribute "_replace"

[gvanrossum]

Can you repro with the latest mypy version from the repro?

[elazarg]

No it was fixed but not released yet.

[gvanrossum]

Our policy is to close bugs when the fix lands in the repo. So if this issue is fixed in the repo it should be closed now. (As the originator you should be able to.)

[dlinnemeyer]

You were right. Installing directly from git doesn't give any errors.

[elazarg]

@gvanrossum please don't close this issue. It is about other problems, not yet solved.

[gvanrossum]

Sorry!

[emmatyping]

namedtuple is no longer an experimental mypy concept, I think this can be closed. Any work left belongs in python/typing or CPython :)