CLN: explicit kwargs for select

[jbrockmendel]

Questions for @jreback on follow-ups:

1. Note in read_hdf kwargs is still present in the signature, but i put an assertion that they are empty. In principle we could pass more junk to HDFStore, but we never do in tests. Should we allow it, or can those kwargs be deleted altogether?

2. the current signature for select_column is (self, key: str, column: str, **kwargs). These kwargs are passed to read_column, which would accept "where", "start", "stop". In tests we pass start and stop, but never where. Can we rule out "where"?

[jreback]

Questions for @jreback on follow-ups:

1. Note in read_hdf kwargs is still present in the signature, but i put an assertion that they are empty. In principle we could pass more junk to HDFStore, but we never do in tests. Should we allow it, or can those kwargs be deleted altogether?
2. the current signature for select_column is (self, key: str, column: str, **kwargs). These kwargs are passed to read_column, which would accept "where", "start", "stop". In tests we pass start and stop, but never where. Can we rule out "where"?

IIRC I answered 1) elsewhere. we pass on kwargs in the constructor to .open_file; these are keywords that are used in the construction deep in PyTables for how to open the backends, e.g.https://www.pytables.org/cookbook/inmemory_hdf5_files.html, e.g. you could pass driver= and it would work.

2. yes I think you can remove where

[jreback]

a solution to this would be to remove kwargs and instead add a keyword backend_kwargs={} or driver_kwargs={}

[jbrockmendel]

that would work, will need deprecation cycle, and im all about removing deprecations these days

[jbrockmendel]

for now ill need to make a quick PR reverting the assert not kwargs this introduced

[jreback]

suggestion for a way out to remove the kwargs arg in read_hdf & HDFStore