TYP use typed version of namedtuple

[MarcoGorelli]

Instead of

OutputKey = collections.namedtuple("OutputKey", ["label", "position"])

we can do

from typing import NamedTuple

class OutputKey(NamedTuple)
    label
    position

The advantage is that like this we can add type hints to the various elements, such as

from typing import NamedTuple

class OutputKey(NamedTuple)
    label: Hashable
    position: int

---

See https://docs.python.org/3/library/typing.html#typing.NamedTuple for reference

[jreback]

better to use a dataclass

[MarcoGorelli]

Is that always true? Immutability is a nice property, I'd have thought we wouldn't want to lose that

[attack68]

Im also waiting for pandas to adopt TypedDict, for similar reasons. Similar to a data class but also with user input args where a dict is expected and a dataclass wouldn't work becuase the user input cant be a data class.

[MarcoGorelli]

Is that always true? Immutability is a nice property, I'd have thought we wouldn't want to lose that

at the time of writing, I wasn't aware that dataclasses could use frozen=True, so that'd do

[franslarsson]

take

[MarcoGorelli]

take

Hi @franslarsson - still working on this?

[franslarsson]

take

Hi @franslarsson - still working on this?

Hi @MarcoGorelli
Sorry for the delay, I'll open a PR for it tomorrow, Wednesday.