tip¶

astype(int)¶

will does a conversion toward zero, not rounding, try this s1.round().astype(int)

np.where(cond, pd.Timestamp(), pd.Timestamp())¶

will change pd.Timestamp() to inetger, the right way to use where np.where(cond, pd.Timestamp(), np.datetime64())

list of lists to csv¶

csv.writer(open('out.csv', 'w', newline='')).writerows(lls)

set index in each group starting from 0¶

df['idx'] = df.groupby('grp_id').cumcount()