ENH: resample(..., how='range', start=a, stop=b)

[danielballan]

Carrying over from SO comments

Is it worth adding kwargs start and stop to enable this succinct usage?

df.resample('D', how='range', start=0, stop=3)

[jreback]

which could be done faster than, as this can be done in cython

def f(frame):
       l = len(frame)
       if start > l:
          start = l
       if stop > l:
           stop = l
       return frame.iloc[start:stop,:]
df.resample('D',how=f)

[jreback]

@danielballan fyi..i think the perf difference of using groupby(lambda x: x.date()) and resample is really inconsequential to calling the user function

[danielballan]

Where would I put the Cython? lib.pyx?

[jreback]

look in pandas/algos.pyx, you would create something like group_nth_object and in pandas/src/generated.pyx, where group_nth_float64 and other dtypes are handled (the generated are code generateors)

[jreback]

sorry I meant pandas/src/generate_code.py, which actually creates the generated.py (but to test just create the cython directly)

[jreback]

this is xref in #7387 already