Narrow input type annotation for GLib.Bytes.new

[phistep]

The type annotation for GLib.Bytes.new() input argument data is confusingly narrow: Sequence[int].

pygobject-stubs/src/gi-stubs/repository/GLib.pyi

Line 1425 in a576d99

def new(cls, data: typing.Optional[typing.Sequence[int]] = None) -> Bytes: ...

Yet the following code works without a problem:

import io

import gi
from gi.repository import GLib

data = b"hi mom."
array = GLib.Bytes.new(data)
array = GLib.Bytes.new(bytearray(data))
array = GLib.Bytes.new(memoryview(data))
array = GLib.Bytes.new(io.BytesIO(data).getvalue())

I think the type annotation should reflect this.

I made a PR that changes the type to typing.Union[typing.Sequence[int], bytes] but I'm neither sure that this will not have unwanted consequences for other classes nor that this is exhaustive for all the possible values shown above or what else might be legal to pass. I hope someone more knowledgeable can figure out what the maximal type annotation would be.

[phistep]

It is possible that I am not familiar enough with the Python type system, since bytes() seems to be a compatible type to typing.Sequence[int]

$ mypy <(cat <<EOF
from typing import Sequence
def f(x: Sequence[int]): ...
f(bytes([0xff]))
EOF
)
Success: no issues found in 1 source file

I just find that very confusing?

[lovetox]

bytes is defined as sequence of integers

https://github.com/python/typeshed/blob/main/stdlib/builtins.pyi#L624

Still your question is valid if we cannot simply type it as bytes, i will think about that. Currently this may be confusing, but it will not yield any type errors.

[phistep]

Thanks for your consideration!