Get category mapping for categoricals

[mcrumiller]

Description

While usually internal implementations should remain hidden, categorical data is universally known to be represented by integers, and this representation is often useful to developers. I propose we expose the rev mapping in the API via the following:

• Series.cat.get_map - get map for categorical series as dictionary
• pl.StringCache.get_map - get global categorical map.

This has also come up on discord a few times. We usually refer people to s.unique(), but that does not always show all of the categories.

[deanm0000]

I like them both. For the first one though you can do:

s=pl.Series("a",['apple','banana', 'apple','carrot','date','eggplant'],dtype=pl.Categorical)
s=s.filter(s!='carrot')

{x[1]:x[0] for x in s.cat.get_categories().to_frame().with_row_index('i').iter_rows()}
{'apple': 0, 'banana': 1, 'carrot': 2, 'date': 3, 'eggplant': 4}

works even though carrot isn't in s anymore.

Actually you don't even need to go through a frame, you can alternatively do

{val:i for i, val in enumerate(s.cat.get_categories())}

[c-peters]

The indices of the cat.get_categories() are equal to the physicals. I'm not convinced that we really need a new method for this as the method by @deanm0000 will provide you the mapping