[Passing custom function to agg function in Pandas (with name!)] #pandas #datascience

groupby_agg_example.py

# define a custom function that calculates the percentile using numpy
def perc(n):
    def perc_(x):
        return np.percentile(x, n)
    perc_.__name__ = 'perc_%s' % n
    return perc_
  
# apply this function (and other stuff) to Pandas Groupby
df.groupby(['group_key_1', 'group_key_2']).agg({
    'col_a': np.sum,
    'col_b': [np.average, np.sum],
    'col_c': [
    	perc(25), # the column name will be perc_25
        perc(50), # the column name will be perc_50
        perc(75), # the column name will be perc_75
        perc(99), # the column name will be perc_99
        np.average],
    'col_d' : lambda x: x + 1 # the problem with this is that it won't have a name
})