group data by date based on values using pandas

Question

How can I group the following data into by month using pandas:

17/1/2001   800
7/1/2001    1300
2/1/2001    400
1/1/2001    200
25/3/2001   1800
8/3/2001    1300

and then have the following output with first and last days of the month and the corresponding first and last values:

First   Last    First   Last
1/1/2001 17/1/2001  200 800
8/3/2001 25/3/2001  1300 1800

Thanks

Answer 1

try this:

In [102]: res = df.sort_values('date').groupby(df.date.dt.month).agg(['first','last'])

In [104]: res.columns = ['date_first', 'date_last', 'first', 'last']

In [105]: res
Out[105]:
     date_first  date_last  first  last
date
1    2001-01-01 2001-01-17    200   800
3    2001-03-08 2001-03-25   1300  1800

or min , max depending on what you want:

In [95]: res = df.groupby(df.date.dt.month).agg(['min','max'])

In [96]: res.columns = ['date_min', 'date_max', 'min', 'max']

In [97]: res
Out[97]:
       date_min   date_max   min   max
date
1    2001-01-01 2001-01-17   200  1300
3    2001-03-08 2001-03-25  1300  1800

Answer 2

Use idxmin and idxmax to identify indices for which to grab the appropriate rows.

def get_min(x):
    return x.loc[x.date.idxmin(), :]

def get_max(x):
    return x.loc[x.date.idxmax(), :]

def app_by_month(df, f):
    return df.groupby(df.date.dt.month).apply(f)

df2 = pd.concat([app_by_month(df, f) for f in [get_min, get_max]],
                axis=1, keys=['first', 'last']).sort_index(axis=1, level=1)

df2.columns = df2.columns.to_series().str.join('_').values

print df2

     first_date  last_date  first_value  last_value
date                                               
1    2001-01-01 2001-01-17          200         800
3    2001-03-08 2001-03-25         1300        1800