Panel.apply() still uses np.apply_along_axis, discarding the indices and labels

[aldanor]

Panel.apply(func, axis) currently just pulls out .values and passes them to numpy.apply_along_axis -- which makes its utility questionable since all the indices and labels get lost in the process.

Wouldn't it make much more sense to pass DataFrames to the func and then try and recombine the result into either Series, DataFrame or a Panel?

By the way, this behaviour is not in the docs and is inconsistent with how apply works for series/frames, so an unwary user might be in for a surprise.

[jreback]

this is a dupe of #1148; I have a working branch of this, but going to happen in 0.14. you are of course welcome too it if you'd like :)

[jreback]

here's the branch: https://github.com/jreback/pandas/tree/panel_apply, need to figure out the alignment issues

[aldanor]

Great, thanks, will take a look later.

[jreback]

@aldanor #5850 is my implementation of this..pls check out and lmk