pathlib should allow converting to absolute paths without resolving symlinks

[mumind]

BPO	25012
Nosy	@pitrou, @serhiy-storchaka, @eryksun
Superseder	bpo-29688: Add support for Path.absolute()

^{Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.}

Show more details

GitHub fields:

assignee = None
closed_at = <Date 2021-03-17.07:10:19.721>
created_at = <Date 2015-09-05.22:05:48.160>
labels = ['type-feature', 'library']
title = 'pathlib should allow converting to absolute paths without resolving symlinks'
updated_at = <Date 2021-03-17.07:10:19.720>
user = 'https://bugs.python.org/mumind'

bugs.python.org fields:

activity = <Date 2021-03-17.07:10:19.720>
actor = 'eryksun'
assignee = 'none'
closed = True
closed_date = <Date 2021-03-17.07:10:19.721>
closer = 'eryksun'
components = ['Library (Lib)']
creation = <Date 2015-09-05.22:05:48.160>
creator = 'mu_mind'
dependencies = []
files = []
hgrepos = []
issue_num = 25012
keywords = []
message_count = 8.0
messages = ['249936', '249941', '249968', '250129', '250139', '250167', '250208', '250209']
nosy_count = 5.0
nosy_names = ['pitrou', 'mu_mind', 'serhiy.storchaka', 'eryksun', 'Socob']
pr_nums = []
priority = 'normal'
resolution = 'duplicate'
stage = 'resolved'
status = 'closed'
superseder = '29688'
type = 'enhancement'
url = 'https://bugs.python.org/issue25012'
versions = ['Python 3.6']

[mumind]

There doesn't seem to be any helper in pathlib to expand a relative path to an absolute path *without* resolving symlinks.

For example, if I want to convert
pathlib.Path('some/path')
to
pathlib.Path('/full/path/to/some/path')
where some/path is a symlink, my best option seems to be
pathlib.Path(os.path.abspath(str(pathlib.Path('some/path'))))

[serhiy-storchaka]

pathlib.Path.cwd() / pathlib.Path('some/path')

[eryksun]

There's a public method that's almost what you want:

    >>> print(pathlib.Path.absolute.__doc__)
    Return an absolute version of this path.  This function works
            even if the path doesn't point to anything.

        No normalization is done, i.e. all '.' and '..' will be kept along.
        Use resolve() to get the canonical path to a file.

However, it appears to be only accidentally public. It isn't tested; it isn't mentioned in the docs; and it doesn't do the [expected] path normalization. Currently it's used as a default in resolve() if self._flavour.resolve(self) returns None.

[pitrou]

Yes, this wasn't meant to be a public method really. I'm still on the fence as to whether we should expose something like this. What is your use case?

[mumind]

The idiom of
pathlib.Path.cwd() / pathlib.Path('some/path')
isn't too bad of an approach if it could just be mentioned in the docs. I would intuitively expected something like
pathlib.Path('some/path').resolve(follow_symlinks=False)

My use case was that I had an equality check like this failing:
expected_path = pathlib.Path.cwd() / pathlib.Path('some/path')
pathlib.Path('some/path') == expected_path
I suppose I should file that as a separate bug because semantically those paths are equal, but I was trying to work around it like
pathlib.Path('some/path').resolve() == expected_path
which in my case still failed because some/path was a symlink.

Even if that's fixed, I would still expect to run into cases where I wanted to print specifically the relative or absolute path in informational messages and would not want to follow symlinks (e.g., in "Requested path X not found", the user would recognize the absolute path as the one they entered but not necessarily the symlink-resolved version).

[pitrou]

Le 08/09/2015 01:21, David Barnett a écrit :

My use case was that I had an equality check like this failing:
expected_path = pathlib.Path.cwd() / pathlib.Path('some/path')
pathlib.Path('some/path') == expected_path
I suppose I should file that as a separate bug because semantically those paths are equal, but I was trying to work around it like
pathlib.Path('some/path').resolve() == expected_path
which in my case still failed because some/path was a symlink.

Semantically, the paths are equal only in that particular system
configuration, with os.getcwd() pointing to a particular directory. Path
comparison is done on "pure" path objects (as explained in the
documentation) and therefore doesn't take into account system specifics.
This makes comparisons much more predictable.

Currently the way to implement "absolute" comparison is to call
.resolve() on both paths. Note you might care about actual file identity
and call os.path.samefile() instead.

Even if that's fixed, I would still expect to run into cases where I
wanted to print specifically the relative or absolute path in
informational messages and would not want to follow symlinks (e.g., in
"Requested path X not found", the user would recognize the absolute path
as the one they entered but not necessarily the symlink-resolved version).

Yes... and yet that may be misleading, since the "absolute" path
obtained may not correspond to the actual one, especially if ".."
fragments are collapsed.

I'll have to think about this one a bit more.

[mumind]

Right, and to clarify a bit further why I didn't just use A.resolve() == B.resolve() from the beginning, this is in a unit test where the equality check wasn't in my code. I wanted to assert I received a certain call on my mock, like
mock_open_method.assert_called_once_with(pathlib.Path('foo'))
so it's much more convenient to be able to create a path object that matches with my target path instead of more complicated matching.

[mumind]

And the symlinks for my paths refer to really cryptic hashes in a virtual filesystem for the tests, so rendering them into the assertion failed errors would really make the failure messages hard to interpret.