Using a pandas dataframe how to aggregate and groupby and bring in non aggregated/groupby columns

Matthew Doering :

Given:

import pandas as pd

d = {'month': pd.Series(['jan', 'jan', 'feb', 'feb']),
     'week' : pd.Series(['wk1', 'wk2', 'wk1', 'wk2']),
     'high_temp' : pd.Series([10, 20, 30, 20]), 
     'low_temp' : pd.Series([4, 5, 23, 40])} 

df = pd.DataFrame(d)
df

    high_temp  low_temp month week
0         10         4   jan  wk1
1         20         5   jan  wk2
2         30        23   feb  wk1
3         20        40   feb  wk2

I would like to get is a new dataframe with this data

    month    high_temp   high_temp_week  low_temp  low_temp_week
0   Jan         20            wk2           4          wk1
1   Feb         30            wk1           23         wk1

I can easily get the max of the temps grouped by month but I can't figure out how to bring along the week column from the row with the max value.

Ben.T :

you can do it by sort_values depending on the case, drop_duplicates and keep last of first, then merge. You do the merge only on month and you specify suffixes to rename the column week that is in both dataframe.

new_df = df[['month', 'high_temp', 'week']].sort_values('high_temp').drop_duplicates('month', keep='last')\
            .merge(df[['month', 'low_temp', 'week']].sort_values('low_temp').drop_duplicates('month', keep='first'), 
                   on='month', suffixes=('_high_temp', '_low_temp'))

print (new_df)
  month  high_temp week_high_temp  low_temp week_low_temp
0   jan         20            wk2         4           wk1
1   feb         30            wk1        23           wk1