[Python] glob.translate incorrectly matches path separator when using ranges

I am playing around with new Python 3.13 function glob.translate. Docs say that:

By default wildcards do no match path separators, and * pattern segments match precisely one path segment.
​

This can be seen in:

import glob
import re

r = re.compile(glob.translate('a?c', seps=['/'])))
print(r) # (?s:a[^/]c)\Z
print(r.match('abc')) # true
print(r.match('a/c')) # false


I would therefore expect that any range which includes path separator would automatically exclude it, but it is not the case:

import glob
import re

r = re.compile(glob.translate('a[%-0]c', seps=['/'])))
print(r) # (?s:a[%-0]c)\Z
print(r.match('a0c')) # true
print(r.match('a/c')) # true


Are my expectations correct? The glob module is documented to behave as Unix, but the same glob does not match the path separator in my bash shell:

$ mkdir a
$ touch a/c
$ ls -1 a/c
a/c
$ ls -1 [a-c]/[a-c] # check glob support
a/c
$ ls -1 a[%-0]c
ls: cannot access 'a[%-0]c': No such file or directory
$ shopt -s failglob
$ ls -1 a[%-0]c
-bash: no match: a[%-0]b


So it looks to me like a bug in Python, which should instead translate the range [%-0] to [%-.0] or use negative lookahead (?!\/)[%-0].

Continue reading...