concatenate row values for the same index in pandas

Question

My initial DataFrame looks as follows:

   A    B  quantity
0  1  foo         1
1  1  baz         2
2  1  bar         2
3  1  faz         1
4  2  foo         2
5  2  bar         1
6  3  foo         3

I need to group it by 'A' and make a list of 'B' multiplied by 'quantity':

   A                               B
0  1  [foo, baz, baz, bar, bar, faz]
1  2                 [foo, foo, bar]
2  3                 [foo, foo, foo]

Currently I'm using groupby() and then apply():

def itemsToList(tdf, column):

    collist = []
    for row in tdf[column].iteritems():
        collist = collist + tdf['quantity'][row[0]]*[row[1]]

    return pd.Series({column: collist})

gb = df.groupby('A').apply(itemsToList, 'B')

I doubt it is an efficient way, so I'm looking for a good, "pandaic" method to achieve this.

Answer 1

This could be done in 2 steps, generate a new column that creates the expanded str values, then groupby on 'A' and apply list to this new column:

In [62]:
df['expand'] = df.apply(lambda x: ','.join([x['B']] * x['quantity']), axis=1)
df.groupby('A')['expand'].apply(list)

Out[62]:
A
1    [foo, baz,baz, bar,bar, faz]
2                  [foo,foo, bar]
3                   [foo,foo,foo]
Name: expand, dtype: object

EDIT

OK after taking inspiration from @Jianxun Li's answer:

In [130]:
df.groupby('A').apply(lambda x: np.repeat(x['B'].values, x['quantity']).tolist())

Out[130]:
A
1    [foo, baz, baz, bar, bar, faz]
2                   [foo, foo, bar]
3                   [foo, foo, foo]
dtype: object

Also this works:

In [131]:
df.groupby('A').apply(lambda x: list(np.repeat(x['B'].values, x['quantity'])))

Out[131]:
A
1    [foo, baz, baz, bar, bar, faz]
2                   [foo, foo, bar]
3                   [foo, foo, foo]
dtype: object

Answer 2

Another way to do it. First reshape the df using pivot_table and then apply np.repeat().tolist().

import pandas as pd
import numpy as np

df

Out[52]: 
   A    B  quantity
0  1  foo         1
1  1  baz         2
2  1  bar         2
3  1  faz         1
4  2  foo         2
5  2  bar         1
6  3  foo         3

df.pivot('A','B','quantity').fillna(0).apply(lambda row: np.repeat(row.index.values, row.values.astype(int)).tolist(), axis=1)

Out[53]: 
A
1    [bar, bar, baz, baz, faz, foo]
2                   [bar, foo, foo]
3                   [foo, foo, foo]
dtype: object