pathlib glob ignores trailing slash in pattern

[jocabt]

BPO	22276
Nosy	@pitrou, @bitdancer, @emilyemorehouse, @robbuckley, @akulakov
PRs	bpo-22276: Fix pathlib.Path.glob not to ignore trailing path separator #10349

^{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 = 'https://github.com/emilyemorehouse'
closed_at = None
created_at = <Date 2014-08-26.11:37:22.662>
labels = ['3.8', 'type-bug', 'library']
title = 'pathlib glob ignores trailing slash in pattern'
updated_at = <Date 2021-11-21.02:47:05.959>
user = 'https://bugs.python.org/jocabt'

bugs.python.org fields:

activity = <Date 2021-11-21.02:47:05.959>
actor = 'andrei.avk'
assignee = 'emilyemorehouse'
closed = False
closed_date = None
closer = None
components = ['Library (Lib)']
creation = <Date 2014-08-26.11:37:22.662>
creator = 'joca.bt'
dependencies = []
files = []
hgrepos = []
issue_num = 22276
keywords = ['patch']
message_count = 9.0
messages = ['225914', '226033', '226034', '226036', '226095', '226102', '406700', '406701', '406702']
nosy_count = 7.0
nosy_names = ['pitrou', 'r.david.murray', 'cvrebert', 'joca.bt', 'emilyemorehouse', 'robbuckley', 'andrei.avk']
pr_nums = ['10349']
priority = 'normal'
resolution = None
stage = 'patch review'
status = 'open'
superseder = None
type = 'behavior'
url = 'https://bugs.python.org/issue22276'
versions = ['Python 3.8']

[jocabt]

Both fnmatch and glob support the "*/" glob. However, pathlib does not seem to handle this kind of globs correctly.

dir = Path("/a/directory/")
file = Path("/a/file")
print(dir.match("*/")) # True
print(file.match("*/")) # True

The "/" is being discarded by the match, resulting in incorrect matches.

Both the fnmatch and glob libraries support this correct.
print(fnmatch("/a/directory/", "*/")) # True
print(fnmatch("/a/file", "*/")) # False

bpo-21039 may be related to this.

[pitrou]

What is "*/" supposed to do? Only select directories?

[jocabt]

Yes.

[bitdancer]

Heh. I never noticed that about shell globs, but it is logical. Learn something new every day.

[pitrou]

Well, it would be logical if pathlib gave special meaning to trailing slashes, which it (still) doesn't :-) I'm still not fond of encoding path characteristics in the path string itself.
(and what about symlinks? etc.)

[bitdancer]

I'm not sure that a trailing '/' is a "path characteristic" in the same sense that a symlink is. Whether the path has a trailing slash or not has meaning both to the user and to the OS. pathlib isn't just modeling actual path objects on the file system, but the abstract concept of a path, and in the abstract context the presence or absence of a trailing '/' as meaning.

But that's a wider discussion than this issue :)

[akulakov]

I have also run into this when looking into path.glob('dangling_symlink') issue.

I can add a few things (in the examples, *myfile* is a file, not a directory):

This is probably more common / less obscure than '*/':

path.glob('myfile/') => True

This is inconsistent with how shell ls command works and with glob.glob() and looks wrong.

Path('myfile/').exists() => True

Path('myfile/') == Path('myfile') => True

str(Path('myfile/')) => 'myfile'

You can compare this to behavior of ls (tested on MacOS):

ls myfile
myfile

ls myfile/
ls: myfile/: Not a directory

I think many users will expect behavior consistent with ls and glob.glob. I've used ls in this manner before.

[akulakov]

I meant to say:

path.glob('myfile/') => [PosixPath('myfile')]

[akulakov]

Generally if Path is created with a trailing separator, I think it should error out for all methods that apply to files, for example .touch(), read*(), write*(), others.

This is consistent with shell commands:

touch xyz/ touch: xyz/: Not a directory

echo 'blah' > xyz/
zsh: not a directory: xyz/

[barneygale]

This was fixed in #10349. This issue could be closed.

[barneygale]

Fixed in #10349.