ENH: density of a categorical

[jreback]

I think s.cat.density is appropriate here

In [1]: s = Series(list('aabbcccc')).astype('category')

In [2]: 100*len(s.cat.categories)/len(s)
Out[2]: 37

[bashtage]

density seems like the wrong word since it isn't continuous.

[jreback]

this is to mimic sparse, could be cardinaility as well (but that's kind of long).

[TomAugspurger]

mass is the discrete version of density. I would expect that to be a normalized value_counts though. Not sure about cardinality.

[jorisvandenbossche]

Is this needed? Seems like easy to calculate it yourself if you need it

(is there a specific use case where this came up?)

[kawochen]

Of the suggestions so far I think density sounds the most reasonable. There needn't be a probablistic interpretation at all. It is more akin to a compaction ratio in string interning.

[shoyer]

I'm also not convinced this is necessary -- what's the use case here?

[jreback]

IIRC was having a discussion about the use cases of categoricals with high (k < n), and low (k << n), where k is number of cats and n is length. So density came up.