gzip compression options not visible via dataset.compression_opts with v2 filter pipelines

[bnlawrence]

This issue is exposed by adding a small test to test_filter_pipeline_v2

def test_filter_pipeline_compression_opts_v2():

     with pyfive.File(FILTER_PIPELINE_V2_FILE) as hfile:
        assert 'data' in hfile
        d = hfile['data']
        # the point of this test is to ensure we can actually retrieve the compression opts
        x = d.compression_opts

This will fail with

  @property
    def compression_opts(self):
        """ Compression filter options, None is no options/compression. """
        if self._filter_ids is None:
            return None
        if GZIP_DEFLATE_FILTER in self._filter_ids:
            gzip_entry = [d for d in self.filter_pipeline
                          if d['filter_id'] == GZIP_DEFLATE_FILTER][0]
>           return gzip_entry['client_data'][0]
E           KeyError: 'client_data'

pyfive/dataobjects.py:324: KeyError

which I think happens because the v2 spec has only client_data_values not client_data in this dictionary (to be clear, I do not know what the spec has, cause I can't find the details of this bit, but by inspection, this is the difference.)

[bnlawrence]

So for example, this is what gzip_entry looks like in test_compressed_dataset:

odict_keys(['filter_id', 'name_length', 'flags', 'client_data_values', 'filter_name', 'client_data'])

and this is what it looks like in the test above:

odict_keys(['filter_id', 'optional', 'name', 'client_data_values'])

I would just "fix" this to use client_data_values in both, but I am not sure what is actually going on.

[bnlawrence]

Well, for now, this seems to work, but I don't really know what is going on (well, I know what the option is, but the spec here implies that the filter description should not differ between v1 and v2.

key = {0:'client_data_values',1:'client_data'}['client_data' in gzip_entry]
return gzip_entry[key][0]