[add] seekable for ArrowFile

[nyamashi]

Fix #1153

ref:
https://arrow.apache.org/docs/python/generated/pyarrow.fs.HadoopFileSystem.html#pyarrow.fs.HadoopFileSystem

• open_*_streams and open_input_file of pyarrow.fs.HadoopFileSystem return pyarrow.NativeFile.
• pyarrow.NativeFile supports seekable.

[martindurant]

Can we perhaps add a test in test_arrow.py which checks seekable works?

[nyamashi]

Got it! I'll try to add a test in test_arrow.py.

[nyamashi]

I'd added a test case but I couldn't checked via pytest fsspec. 🤔
However the following code looks working on my local environment.

def test_seekable(fs, hdfs_path):
    for seekable in [True, False]:                                                                                                                                                   
        with fs.open(hdfs_path, "rb", seekable=seekable) as file:                                                                                                        
            assert file.seekable() == seekable 

my_hdfs_url = 'hdfs://foo/bar/'
fs, my_hdfs_path = fsspec.core.url_to_fs(my_hdfs_url)
test_seekable(fs, my_hdfs_path + "/baz")

Could you run the test code on your environment? 🙇

[martindurant]

Thanks!

[f4hy]

This looks like it makes it possible to fix pandas-dev/pandas#50639 but it is unclear how pandas could pass the seekable option.