glob does not work on Python < 3.7.6

[remi-braun]

Hello,

I tried your promising 0.7.0 version this morning, however, the glob function breaks for me now:

directory = S3Path('s3://sertit-extracteo-ci/water')
directory.glob('*S2*_MSIL2A*')

throws:

self = S3Path('s3://sertit-extracteo-ci/water'), pattern = '*S2*_MSIL2A*'

    def glob(self, pattern):
        self._glob_checks(pattern)

        pattern_parts = PurePosixPath(pattern).parts
>       selector = _make_selector(tuple(pattern_parts), _posix_flavour)
E       TypeError: _make_selector() takes 1 positional argument but 2 were given

/usr/local/lib/python3.7/dist-packages/cloudpathlib/cloudpath.py:344: TypeError

I'm on docker, debian and cloudpathlib is installed through pip.

[jayqi]

Hi @remi-braun, what version of Python 3.7 are you using? I suspect that there was a change to pathlib internals, and this error is caused by compatibility. I'm not certain from a quick look at the python/cpython source code, but I think maybe 3.7.5+ or 3.7.6+ may resolve this? Our latest tests ran 3.7.12, so I would expect at least that should work.

---

Noting for future debugging that this is the commit that changed the _make_selector signature: https://github.com/python/cpython/commit/175abccbbfccb2f6489dc5c73f4630c1b25ce504#

[remi-braun]

Yes you're right, I am using the version 3.7.3, provided by the command line apt-get python3.7 Do you know how to download a newer version of Python 3.7 on Debian ?

[jayqi]

I'm not a power user of Linux package managers, but some things to try or consider:

• Did you run apt-get update? I believe this updates the package repository information, so maybe your apt-get's index of packages is out of date.
• Have you taken a look at the official Python Docker images? They are all on Debian and come with Python installed already, so maybe you be able to swap it in for your current base image.

Otherwise, I don't really know how to help off of the top of my head. In any case, thanks for bringing this error to our attention. We'll definitely either want to support these older versions or put in version minimums.

[remi-braun]

Yes sadly :( 3.7.3 is the last maintained 3.7 version for Debian. I discarded the official Python images for paths reasons, but I think I will try them again.

[remi-braun]

Switching to 3.7.12 solves this 👍

[cedricdonie]

I also get the same error also for GSPath with version 0.6.5 on Python 3.6.15 (Ubuntu 20.04.4 LTS) through the following minimum working example. Interestingly, only iterating through the generator yields the error.

>>> from cloudpathlib import GSPath
>>> p = GSPath("gs://blablabla/bla")
>>> p
GSPath('gs://blablabla/bla')
>>> p.glob("*.txt")
<generator object CloudPath.glob at 0x7f1ed54af888>
>>> list(p.glob("*.txt"))
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/cedric/.venvs/mjff-ldopa/lib/python3.6/site-packages/cloudpathlib/cloudpath.py", line 344, in glob
    selector = _make_selector(tuple(pattern_parts), _posix_flavour)
TypeError: _make_selector() takes 1 positional argument but 2 were given

This is fixed if I pip install git+https://github.com/drivendataorg/cloudpathlib.git@cb031d3987ffdd9b751169b01f44ac793f1d847b instead to get the fix from https://github.com/drivendataorg/cloudpathlib/pull/202.

[pjbull]

Thanks for reporting @cedricdonie. Did you try the latest version on PyPI as well (0.9.0)? That also has the fix.

That said, we no longer support Python 3.6 since it is end-of-life.

[pjbull]

This is a won't fix. The glob logic is too complicated to try to fix this just for these old versions of Python. Plus, it looks like current Debian stable and Ubuntu LTS releases have compatible Python versions.

[anishazaveri]

Hi,

I have the same issue but I'm using Python 3.8

cp = CloudPath("s3://compteam/anisha_zaveri/2022_09_06_ab_coupled_vae/out/fitted_models/04ffd59445a14ab0952aa4dcfdc0c6e6/", client=client)
list(cp.glob('models/*.ckpt'))

yields

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
Input In [37], in <cell line: 2>()
      1 cp = CloudPath("s3://compteam/anisha_zaveri/2022_09_06_ab_coupled_vae/out/fitted_models/04ffd59445a14ab0952aa4dcfdc0c6e6/", client=client)
----> 2 list(cp.glob('models/*.ckpt'))

File ~/anaconda3/envs/coupled_vae/lib/python3.8/site-packages/cloudpathlib/cloudpath.py:353, in CloudPath.glob(self, pattern)
    350 self._glob_checks(pattern)
    352 pattern_parts = PurePosixPath(pattern).parts
--> 353 selector = _make_selector(tuple(pattern_parts), _posix_flavour)
    355 yield from self._glob(selector)

TypeError: _make_selector() takes 1 positional argument but 2 were given

Cloudpathlib version = 0.10.0 Python version = 3.8.0

[pjbull]

@anishazaveri Can you upgrade to a more recent release of 3.8.x? I believe 3.8.16 is the latest, but anything reasonably recent should work (I believe even 3.8.1 of 3.8.2 will work, but you likely want the security patches in more recent versions).