Inconsistent storage options for zip filesystem depending on how it was created

[mraspaud]

In an attempt to make use of UPath in our libs, I stumbled upon this problem with fsspec:
When creating a zip filesystem directly using the ZipFilesystem class, it looks like the storage_options are not set in the same way than if we got that filesystem through eg fsspec.open.

See an example here:

>>> from pathlib import Path
>>> local_path = Path("/tmp/myfile.txt")
>>> local_path.touch()
>>> from zipfile import ZipFile
>>> with ZipFile("/tmp/myzipfile.zip", 'w') as myzip:
...     myzip.write("/tmp/myfile.txt")
...     
>>> import fsspec
>>> fsspec_file = fsspec.open("zip://tmp/myfile.txt::/tmp/myzipfile.zip")
>>> fsspec_file.fs.to_dict()
{'cls': 'fsspec.implementations.zip:ZipFileSystem', 'protocol': 'zip', 'args': [], 'target_options': {}, 'target_protocol': 'file', 'fo': '/tmp/myzipfile.zip'}
>>> fsspec.filesystem("zip", **fsspec_file.fs.storage_options).to_dict()
{'cls': 'fsspec.implementations.zip:ZipFileSystem', 'protocol': 'zip', 'args': [], 'target_options': {}, 'target_protocol': 'file', 'fo': '/tmp/myzipfile.zip'}

>>> from fsspec.implementations.zip import ZipFileSystem
>>> zip_fs = ZipFileSystem("/tmp/myzipfile.zip")
>>> zip_fs.to_dict()
{'cls': 'fsspec.implementations.zip:ZipFileSystem', 'protocol': 'zip', 'args': ['/tmp/myzipfile.zip']}
>>> fsspec.filesystem("zip", **zip_fs.storage_options).to_dict()
Traceback (most recent call last):
  File "<python-input-18>", line 1, in <module>
    fsspec.filesystem("zip", **zip_fs.storage_options).to_dict()
    ~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/mraspaud/src/.venv/lib/python3.13/site-packages/fsspec/registry.py", line 322, in filesystem
    return cls(**storage_options)
  File "/home/mraspaud/src/.venv/lib/python3.13/site-packages/fsspec/spec.py", line 81, in __call__
    obj = super().__call__(*args, **kwargs)
  File "/home/mraspaud/src/.venv/lib/python3.13/site-packages/fsspec/implementations/zip.py", line 62, in __init__
    self.fo = fo.__enter__()  # the whole instance is a context
              ~~~~~~~~~~~~^^
  File "/home/mraspaud/src/.venv/lib/python3.13/site-packages/fsspec/core.py", line 105, in __enter__
    f = self.fs.open(self.path, mode=mode)
  File "/home/mraspaud/src/.venv/lib/python3.13/site-packages/fsspec/spec.py", line 1338, in open
    f = self._open(
        path,
    ...<4 lines>...
        **kwargs,
    )
  File "/home/mraspaud/src/.venv/lib/python3.13/site-packages/fsspec/implementations/local.py", line 210, in _open
    return LocalFileOpener(path, mode, fs=self, **kwargs)
  File "/home/mraspaud/src/.venv/lib/python3.13/site-packages/fsspec/implementations/local.py", line 387, in __init__
    self._open()
    ~~~~~~~~~~^^
  File "/home/mraspaud/src/.venv/lib/python3.13/site-packages/fsspec/implementations/local.py", line 392, in _open
    self.f = open(self.path, mode=self.mode)
             ~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^
IsADirectoryError: [Errno 21] Is a directory: '/home/mraspaud/src/'

This is problematic when I then want to create a UPath instance with:

path = UPath("/tmp/myfile", protocol=zip_fs.protocol, storage_options=zip_fs.storage_options)

which crashes eventually with the same error as above when one calls open or fs on it for example.

Some relevant versions:

Python 3.13.2 (main, Feb 12 2025, 14:51:17) [Clang 19.1.6 ] on linux

>>> import fsspec
>>> fsspec.__version__
'2025.7.0'

[martindurant]

It seems that just the args part is not getting traced correctly. More common and recommended use - but not enforced - is to use kwargs always

>>> zip_fs = ZipFileSystem(fo="/tmp/myzipfile.zip")
>>> fsspec.filesystem("zip", **zip_fs.storage_options).to_dict()
{'cls': 'fsspec.implementations.zip:ZipFileSystem',
 'protocol': 'zip',
 'args': [],
 'fo': '/tmp/myzipfile.zip'}

This is still not quite the same, as the path is not analysed to be of type "file" before the instance is created - but it still works.

[mraspaud]

Thanks for the workaround! I'll check it out later this week.