Groupby multiple columns and aggregation with dask

Question

dask dataframe looks like this:

A     B     C     D
1     foo   xx    this
1     foo   xx    belongs
1     foo   xx    together
4     bar   xx    blubb

i want to groupy by columns A,B,C and join the strings from D with a blank between, to get

A     B     C     D
1     foo   xx    this belongs together
4     bar   xx    blubb

i see how to do this with pandas:

df_grouped = df.groupby(['A','B','C'])['D'].agg(' '.join).reset_index()

how can this be achieved with dask?

score 2 · Accepted Answer · answered Nov 29 '19 at 18:28

2

ddf = ddf.groupby(['A','B','C'])['D'].apply(lambda row: ' '.join(row)).reset_index()
ddf.compute()

Output:

Out[75]: 
   A    B   C                      D
0  1  foo  xx  this belongs together
0  4  bar  xx                  blubb

answered Nov 29 '19 at 18:28

KRKirov

score 0 · Answer 2 · answered Nov 29 '19 at 17:00

0

You could use a CustomAggregation, where both the per-chunk and aggregation operations are your ' '.join method.

answered Nov 29 '19 at 17:00

MRocklin

2 Answers2