Python: deprecate most indexing operations.

[ritchie46]

Which ones should we still allow?

[slonik-az]

Please keep the slicing operations. They are standard in Data Science since numpy. The easiest way to take every other row of a series or a dataframe is df[::2].

Moreover, it would be great to allow for negative steps in Series like series[::-1]. Currently, negative steps work fine for DataFrames but result in error for Series (probably a bug):

import polars as pl
sr = pl.Series(name='int', values=np.arange(10,20))
sr[::-1]

OverflowError: can't convert negative int to unsigned

[ritchie46]

I agree, I want to remove predicates in index notation. That should always be a filter imo.

I want to maintain support for slicing and selection by strings/index.

[ghuls]

I think other cases can indeed be removed. Slicing operations were greatly improved in https://github.com/pola-rs/polars/pull/3904

[stinodego]

I agree, I want to remove predicates in index notation. That should always be a filter imo.

I want to maintain support for slicing and selection by strings/index.

This is the status quo, so I think this can be closed.