TarFileSystem: using wildcards together with `//` in tar

[observingClouds]

What happened

I like to open several tar items within a tar file with fsspec, like:

import fsspec
of = fsspec.open_files("tar://**/*.txt::test.tar")

This works as expected for "normal" tar archives, but unfortunately, the tar entries I need to access include // instead of just /.
While I can access those files directly when mentioning them explicitly, using wildcards fails.

Minimal example

import fsspec
import subprocess
import tarfile
import os

file_path = "test.txt"
subprocess.run(['touch', file_path])

tar_file_path1 = "test_w_extraslash.tar"
with tarfile.open(tar_file_path1, 'w') as tar:
     tar.add(file_path, arcname='path/with/extra/slash//test.txt')

of = fsspec.open_files("tar://**/*::test_w_extraslash.tar")
#of
# <List of 0 OpenFile instances>

tar_file_path2 = "test_wo_extraslash.tar"
with tarfile.open(tar_file_path2, 'w') as tar:
     tar.add(file_path, arcname='path/without/extra/slash/test.txt')

of2 = fsspec.open_files("tar://**/*::test_wo_extraslash.tar")
# of
# <List of 1 OpenFile instances>

Any smart ideas how one can handle this case?

[observingClouds]

Alright, I made a patch by normalizing the look-up name, while keeping the original tar key for the actual extraction. Happy to open a PR for https://github.com/observingClouds/filesystem_spec/tree/fix/tar_normpath

[martindurant]

I don't quite follow what's happening in your commit, but if it solves a use case and doesn't cause any other problems, then including it sounds like a good idea!