如何使用熊猫将concat函数应用于按数据帧分组？

Question

Code 1:

df = pd.read_csv("example.csv", parse_dates=['d'])
df2 = df.set_index(['d', 'c'])
df3 = df2.groupby(level=['c'])

def function(x):
    a = pd.rolling_mean(x, 3).rename(columns = {'b':'rm'})
    c = pd.rolling_std(x, 3).rename(columns = {'b':'rsd'})
    pd.concat([x, a, c], axis=1)

df4 = df3.apply(lambda x: function(x))

Code 2:

df = pd.read_csv("example.csv", parse_dates=['d'])
df2 = df.set_index(['d', 'c'])
df3 = df2.groupby(level=['c'])

def function(x):
    x.assign(rm = lambda x: pd.rolling_mean(x, 3))

df4 = df3.apply(lambda x: function(x))

Output of df4.head() in both of the above code1 AND code 2 is a square in iPython?? I can't figure out why.

Output:

What df3 looks like:

What df looks like:

Answer 1

You're missing a return statement:

In [11]: def function(x):
             a = pd.rolling_mean(x, 3).rename(columns = {'bookings':'rm'})
             c = pd.rolling_std(x, 3).rename(columns = {'bookings':'rsd'})
             return pd.concat([x, a, c], axis=1)

In [12]: df3.apply(lambda x: function(x))
Out[12]:
                   bookings          rm        rsd
ds         city
2013-01-01 City_2        69         NaN        NaN
2013-01-02 City_2       101         NaN        NaN
2013-01-03 City_2       134  101.333333  32.501282
2013-01-04 City_2       155  130.000000  27.221315
2013-01-05 City_2       104  131.000000  25.632011
2013-01-06 City_2       121  126.666667  25.967929
2013-01-07 City_2       143  122.666667  19.553346
2013-01-08 City_2       173  145.666667  26.102363
2013-01-09 City_2       142  152.666667  17.616280
2013-01-10 City_2       154  156.333333  15.631165
2013-01-11 City_2       139  145.000000   7.937254

Without the return function was returning None, hence the empty DataFrame (which was rendered by ipython as a square - which may be a bug).

In [13]: df3.apply(lambda x: None)
Out[13]:
Empty DataFrame
Columns: []
Index: []

---

Note: In some languages (eg Ruby, Julia, Scala) the last line is returned without being explicit with the return. In Python if you miss out the return statement the function returns None.

In [21]: def foo():
             1

In [22]: foo() == None
Out[22]: True