It is simply not implemented. You might want to raise a problem (or, even better, a transfer request). Pragmatically, I would simply type nuniquefor your grouping object
Before
g = df.groupby(df.x + df.y)
result = len(g)
After
result = (df.x + df.y).nunique()
Operationally, this is better because it can be lazy (the result lenin Python must be a specific integer) and because you can choose an option nunique_approxthat will be much faster.
source
share