Make `PurePath` initializer args available as an attribute

[barneygale]

Feature or enhancement

Proposal:

I propose that we add a pathlib.PurePath.segments attribute that stores the original arguments given to the PurePath initializer. If we did this, then:

1. Users would be able to implement their own joinpath() / __truediv__()-like methods without resorting to PurePath.parts (requires normalization - slow!) or the private PurePath._raw_paths attribute.
2. In the pathlib ABCs, we can make JoinablePath.segments abstract, and use it from a default implementation of JoinablePath.__str__().
   • N.B. JoinablePath.__str__() is currently abstract, which is weird given it's non-abstract in object.

Usage would look like:

>>> p = PurePath('/usr', 'bin/python3')
>>> p.segments
('/usr', 'bin/python3')

When a PurePath object is given as an argument, its segments should be merged:

>>> p = PurePath('/usr', PurePath('bin', 'python3'))
>>> p.segments
('/usr', 'bin', 'python3')

Has this already been discussed elsewhere?

This is a minor feature, which does not need previous discussion elsewhere

Links to previous discussion of this feature:

https://discuss.python.org/t/protocol-for-virtual-filesystem-paths/82753/8

Linked PRs

• GH-131916: Add pathlib.PurePath.segments #131917

[encukou]

Hi! Sorry for the delay.

Thinking about it more, I'm not comfortable with exposing non-normalized segments.
It would allow access to superficial details that pathlib currently hides, with unexpected consequences. In particular, I think people would (ab)use it to distinguish between paths with and without a trailing slash.

>>> PurePath('/usr/').segments
('/usr/',)
>>> PurePath('/usr').segments
('/usr',)
>>> PurePath('/usr/') == PurePath('/usr')
True

With that, using parts for unknown path types seems safer. For performance you'd need to use type-specific internals.

Similarly, I think a default __str__ should look at normalized data.

[barneygale]

Thanks - on reflection I agree. Closing this issue and the PR.