allow repeated extra arguments

[karasikov]

This fixes the error happening when the user passes extra parameters that are also inferred automatically. E.g., happens in lib datasets:

File ./python-default/3.10.14/lib/python3.10/site-packages/datasets/load.py:2692, in load_from_disk(dataset_path, fs, keep_in_memory, storage_options)
   2689     storage_options = fs.storage_options
   2691 fs: fsspec.AbstractFileSystem
-> 2692 fs, *_ = url_to_fs(dataset_path, **(storage_options or {}))
   2693 if not fs.exists(dataset_path):
   2694     raise FileNotFoundError(f"Directory {dataset_path} not found")

File ./3.10.14/lib/python3.10/site-packages/fsspec/core.py:396, in url_to_fs(url, **kwargs)
    385 known_kwargs = {
    386     "compression",
    387     "encoding",
   (...)
    393     "num",
    394 }
    395 kwargs = {k: v for k, v in kwargs.items() if k not in known_kwargs}
--> 396 chain = _un_chain(url, kwargs)
    397 inkwargs = {}
    398 # Reverse iterate the chain, creating a nested target_* structure

File ./3.10.14/lib/python3.10/site-packages/fsspec/core.py:349, in _un_chain(path, kwargs)
    347 if bit is bits[0]:
    348     kws.update(kwargs)
--> 349 kw = dict(**extra_kwargs, **kws)
    350 bit = cls._strip_protocol(bit)
    351 if (
    352     protocol in {"blockcache", "filecache", "simplecache"}
    353     and "target_protocol" not in kw
    354 ):

TypeError: dict() got multiple values for keyword argument 'account_name'

This fix

from fsspec.core import url_to_fs
url_to_fs('az://[email protected]/DATA', **{'anon': False, 'account_name': 'ACCOUNT'})

Out before:

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
Cell In[1], line 2
      1 from fsspec.core import url_to_fs
----> 2 url_to_fs('az://[email protected]/DATA', **{'anon': False, 'account_name': 'ACCOUNT'})

...
...
    347 if bit is bits[0]:
    348     kws.update(kwargs)
--> 349 kw = dict(**extra_kwargs, **kws)
    350 bit = cls._strip_protocol(bit)
    351 if (
    352     protocol in {"blockcache", "filecache", "simplecache"}
    353     and "target_protocol" not in kw
    354 ):

TypeError: dict() got multiple values for keyword argument 'account_name'

Out after:

(<adlfs.spec.AzureBlobFileSystem at 0x10708f460>, 'DIR/DATA')

[martindurant]

Given the update on the line above, this could be repeated updates, and that way we can be a little more explicit about the order of precedence. In your model, user-supplied arguments should always win, overriding inferred ones?

[karasikov]

Exactly, to completely avoid thinking about the priorities, here we simply deduplicate the same key-value pairs (so the priority is irrelevant). If the same key has two different values, they're passed as before, so that will raise the same error, e.g., TypeError: dict() got multiple values for keyword argument 'account_name'

extra_kwargs = {'x': 5, 'y': 4}, kws = {'z': 4} becomes {'x': 5, 'y': 4, 'z': 4}
extra_kwargs = {'x': 5, 'y': 4}, kws = {'x': 5} becomes {'x': 5, 'y': 4}
extra_kwargs = {'x': 5, 'y': 4}, kws = {'x': 4} becomes dict(**{'x': 5, 'y': 4}, **{'x': 4}) and raises TypeError

[martindurant]

so the priority is irrelevant

Checking the values isn't always straight forward, they might not be simple ints and str. We could catch it and pass a useful message to the user?

[martindurant]

Can we please include the two passing examples in some sort of test?