Inconsistent name for keyword that determines whether new object is generated

[8one6]

If I understand this right, the syntax is:

df.sort_index(..., inplace=False)

but is

df.reindex_axis(..., copy=True)

Any reason why we shouldn't just use the same keyword (either copy or inplace) in all functions to determine whether to modify the original object or make a new one?

[jtratner]

I understand your point. To be clear, right now reindex_axis is written such that it always returns an object. So inplace would actually be new functionality (i.e. keeping data with the original object reference). Also, when copy=False, there's no guarantee that a copy isn't made, copy means "force copy" here.

From the docs on reindex_axis:

copy : boolean, default True
Return a new object, even if the passed indexes are the same

sort_index always makes a copy of the underlying data.

[jtratner]

I believe the rationale for that is that you might use reindex_axis to perform alignment first, then manipulate the object, so you might not want to get a copy if it turns out that both objects are actually indexed with the exact same Index object.

[jreback]

this would cause a lot of API changes..pushing to someday

[mroeschke]

reindex_axis has been deprecated and removed. Closing