Here's the example of the custom function being used.
import patsy as ps
import numpy as np
def date_to_num(date_col):
return (date_col - date_col.min()).dt.days
y, X = ps.dmatrices("n_born ~ date_to_num(date) + np.log(yday)", df_clean)
Beware the dangers of stateful transformations though. See the documentation for the long story.