clip inplace?

[cnsgsz]

Is it possible to support clip in place, just like numpy does? I assume that would speed up working with large dataframe/series as no extra copies will be needed.

If not, is that a workaround to achieve the same effect? I tried np.clip(df.values,-clip_bounds.values, clip_bounds.values, out=df.values), but realized that df.values is actually an expensive call as it is merging blocks underneath.

[jreback]

virtually all methods in pandas return new objects, exception are indexing. inplace does not provide ANY benefit, and should never have been added as it provides no utility and serves to confuse people.

so no, won't be adding any more inplace ops.

[jorisvandenbossche]

@jreback this clip function is actually one of rare methods where having an inplace arg actually would make sense, IMO (since it is possible to do it 'really' in place, opposed to something like sorting)

And it would also not be difficult to add, as DataFrame.where (which is what clip uses under the hood) already supports inplace.

[jreback]

I am ok with it - @cnsgsz if you would like to do an implementation would be great

[guygoldberg]

Starting to work on it