/a/** is understood by Path.full_match as /a followed by multiple segments, at least one

Bug report

Bug description:

/a/** is understood by Path.full_match as "/a followed by multiple segments, at least one", instead of "/a followed by multiple segments, including zero".

From https://docs.python.org/3/library/pathlib.html#pathlib-pattern-language

** (entire segment)
Matches any number of file or directory segments, including zero.

# one segment per **
Path('/a/b').full_match('/a/**') # True
# zero segments per **
Path('/a').full_match('/a/**') # False - expected True
# interestingly
Path('/').full_match('/**') # True

CPython versions tested on:

3.13

Operating systems tested on:

Linux

Linked PRs

gh-139580: Clarify pathlib full_match recursive pattern docs #151479

@barneygale

Path('/a').full_match('/a/**') # False - expected True

I was wrong before, and /a/ is also not matched. I don't know if it's a bug in the code or in the docs.

cc @barneygale

Path('/a').full_match('/a/**') # False - expected True

I think this is expected, because the path does not begin with /a/. The docs could be clearer on this point.

I was wrong before, and /a/ is also not matched. I don't know if it's a bug in the code or in the docs.

pathlib.Path removes trailing slashes so Path('/a/') becomes Path('/a') and fails to match.

Oh, so I wasn't wrong before! I don't know how the docs could be improved though.

The ** pattern has different interpretations in glob and full_match:

In Path('/a').glob('**'), it means "all files and directories under '/a', including '/a' itself"
However, Path('/a').full_match('/a/**') interprets it differently, which I find unexpected

from pathlib import Path

from tempfile import gettempdir, mktempi

temp = Path(gettempdir()) / mktemp()
temp_a = temp / 'a'
temp_a.mkdir(parents=True)
temp_a_1 = temp_a / '1.txt'
temp_a_1.touch()
expected = {temp, temp_a, temp_a_1}
actual = {*temp.glob('**')}
print('Actual:', [f"{p}" for p in sorted(actual)])
assert actual == expected, f"\nExpected: {[f"{p}" for p in sorted(expected)]}"

# Actual: ['/tmp/tmpmcdga9my', '/tmp/tmpmcdga9my/a', '/tmp/tmpmcdga9my/a/1.txt']

In Path('/a').glob('**'), it means "all files and directories under '/a', including '/a' itself"

If pathlib preserved trailing slashes, then the glob would yield '/a/', and that would match '/a/**'. The inconsistency comes from pathlib stripping trailing slashes unfortunately...

It's probably too late to change pathlib.Path to preserve trailing slashes - it would cause a lot of breakage.

But I'm open to the idea of a new class, perhaps pathlib.SimplePath, that is merely an object-orientated version of os.path without any path normalization. That would look like:

>>> list(SimplePath('/tmp/a').glob('**'))
[SimplePath('/tmp/a/'), SimplePath('/tmp/a/1.txt')]
>>> SimplePath('/tmp/a/').full_match('/tmp/a/**')
True
>>> SimplePath('/tmp/a/1.txt').full_match('/tmp/a/**')
True

If anyone wants to pursue this, I'd recommend starting a topic in the Ideas forum.

If pathlib preserved trailing slashes, then the glob would yield '/a/', and that would match '/a/**'. The inconsistency comes from pathlib stripping trailing slashes unfortunately...

While this explains the technical reason behind the behavior, it's essentially an implementation detail. From a user's perspective, the functionality shouldn't depend on such nuances; Path('/a').full_match('/a/**') returning False remains counter-intuitive, in my opinion.
If modifying full_match's behavior is impractical, the documentation should at least clearly explain how full_match interprets the ** pattern, since it differs from glob's interpretation.

pythongh-139580: Clarify pathlib full_match recursive pattern docs

There's a docs PR open for this: #151479. It adds a /a/** example and a note on the trailing-slash normalization.

pythongh-139580: Condense pathlib full_match recursive pattern docs

pythongh-139580: Apply review wording suggestions

Navigation Menu

Repository navigation

`/a/**` is understood by `Path.full_match` as `/a` followed by multiple segments, at least one #139580

Bug report

Bug description:

CPython versions tested on:

Operating systems tested on:

Linked PRs

Metadata

Assignees

Labels

Projects

Milestone

Relationships

Development

/a/** is understood by Path.full_match as /a followed by multiple segments, at least one #139580

Description

Bug report

Bug description:

CPython versions tested on:

Operating systems tested on:

Linked PRs

Activity

picnixz commented on Oct 5, 2025

barneygale commented on Oct 5, 2025

picnixz commented on Oct 5, 2025

macmarrum commented on Oct 5, 2025

barneygale commented on Oct 5, 2025

barneygale commented on Oct 5, 2025

macmarrum commented on Oct 5, 2025

harjothkhara commented on Aug 18, 2026

Metadata

Metadata

Assignees

Labels

Projects

Milestone

Relationships

Development

Issue actions

`/a/**` is understood by `Path.full_match` as `/a` followed by multiple segments, at least one #139580