BUG: read_hdf modifying passed columns list (GH7212)

[ajamian]

Closes #7212. If this object is a list, make a copy and then pass to process_axes - otherwise, just pass the object directly (in this case it is most likely none).

A test for this behavior seems complex and somewhat contrived when compared with what exists in test_pytables.py, so I didn't add one --- let me know any thoughts on this.

[shoyer]

Checking for specific types is generally a bad idea -- what if, for example, a user passes in a list subclass, or another sequence type?

A better approach is to cast any provided columns to a list:

if columns is not None:
    columns = list(columns)

However, given that the mutation happens in the process_axes method, it would be a better idea to put this there.

[shoyer]

I know it seems somewhat contrived, but we do try to add tests even for things like this.

[ajamian]

@shoyer thanks for the feedback! much better solution. i'll get a test together as well.

[jreback]

@ajamian can you add a test for this?

[ajamian]

@jreback absolutely, should be getting to it this weekend.

[ajamian]

@shoyer @jreback thanks for the feedback --- let me know if you have any other comments

[shoyer]

@ajamian one last thing -- can you add a note to the release notes for 0.17? otherwise this looks good to me!

[jreback]

yep lgtm as well - pls add a comment on the test with the issue number

[ajamian]

I put the note into doc/source/whatsnew/v0.17.0.txt, I think this is where it belongs...and comment added to test. Thanks guys - let me know if you have any more comments.

[shoyer]

Thanks!

[jreback]

thanks Tom!