有条件地将单元格的内容合并到列中

Question

Looking for a pandanic way to turn the following df:

    name    desc
0   A       a
1   NaN     aa
2   NaN     aaa
3   B       b
4   NaN     bb

into:

    name    desc
0   A       a
            aa
            aaa
3   B       b
            bb

# strings in desc are concat-ed together with end of line char

I am thinking of the general directions of either itertuple or backfill+groupby, but both of those approaches require some juggling.

here is the starting point:

import pandas as pd
import numpy as np
nan = np.nan

df = pd.DataFrame(
    {'name': ['A', nan, nan, 'B', nan],
    'desc': ['a', 'aa', 'aaa', 'b', 'bb']}
)

Answer 1

you can call ffill directly and agg without using apply and lambda

In [719]: df.ffill().groupby('name').agg('\n'.join).reset_index()
Out[719]:
  name        desc
0    A  a\naa\naaa
1    B       b\nbb

or:

In [729]: df.ffill().groupby('name', as_index=False).agg({'desc': '\n'.join})
Out[729]:
  name        desc
0    A  a\naa\naaa
1    B       b\nbb

Answer 2

I think you want a combination of fillna(method='ffill') and groupby .

How does this look?

import pandas as pd
import numpy as np
nan = np.nan

df = pd.DataFrame(
    {'name': ['A', nan, nan, 'B', nan],
    'desc': ['a', 'aa', 'aaa', 'b', 'bb']}
)

df['name'] = df['name'].fillna(method='ffill')

df = df.groupby('name')['desc'].apply(lambda d: '\n'.join(d)).reset_index()
print df

prints

  name        desc
0    A  a\naa\naaa
1    B       b\nbb