explode multiple rows

Question

I have to below dataframe:

import pandas as pd

a = pd.DataFrame([{"name": "John", 
                   "item" : "item1||item2||item3", 
                   "itemVal" : "item1Val||item2Val||item3Val"}, 
                  {"name" : "Tom", 
                   "item":"item4", 
                   "itemVal" : "item4Val"
                  }
                 ])

The dataframe is like this:

   name                 item                       itemVal
   John  item1||item2||item3  item1Val||item2Val||item3Val
    Tom                item4                      item4Val

I want to explode the row into multiple rows so that it will be like this (note that the item and its itemVal need to match).

   name                 item                       itemVal
   John                item1                      item1Val
   John                item2                      item2Val
   John                item3                      item3Val
    Tom                item4                      item4Val

I have looked at other answers here:

Split (explode) pandas dataframe string entry to separate rows

pandas: How do I split text in a column into multiple rows?

But the works on only one column. How do I make it work on multiple columns? I'm using Pandas 1.0.1 and Python 3.8

Is it always the case that item and itemVal have the same number of partitions? — Michael Delgado, May 16 '20 at 21:55

Andrej Kesely · Accepted Answer · 2020-05-16T22:19:27.107

4

a = a.apply(lambda x: [v.split('||') for v in x]).apply(pd.Series.explode)
print(a)

Prints:

   name   item   itemVal
0  John  item1  item1Val
0  John  item2  item2Val
0  John  item3  item3Val
1   Tom  item4  item4Val

EDIT: If you want to split only selected columns, you can do:

exploded = a[['item', 'itemVal']].apply(lambda x: [v.split('||') for v in x]).apply(pd.Series.explode)
print( pd.concat([a['name'], exploded], axis=1) )

edited May 16 '20 at 22:19

answered May 16 '20 at 21:56

Andrej Kesely

168,389
15
48
91

1

Thanks for your solution, this works. However, is there a way to specify the only columns I want to split? – hydradon May 16 '20 at 22:13

score 1 · Answer 2 · edited Sep 27 '22 at 20:32

A combination of zip, product and chain can achieve the split into rows. Since this involves strings, and more importantly no numerical computation, you should get faster speeds in Python, than running it in Pandas:

from itertools import product,chain
combine = chain.from_iterable

#pair item and itemval columns
merge = zip(df.item,df.itemVal) 

#pair the entires from the splits of item and itemval
merge = [zip(first.split("||"),last.split("||")) for first, last in merge]

#create a Cartesian product with the name column
merger = [product([ent],cont) for ent, cont in zip(df.name,merge)]

#create your exploded values
res = [(ent,*cont) for ent, cont in combine(merger)]
pd.DataFrame(res,columns=['name','item','itemVal'])

    name    item    itemVal
0   John    item1   item1Val
1   John    item2   item2Val
2   John    item3   item3Val
3   Tom     item4   item4Val

score 0 · Answer 3 · answered Nov 25 '20 at 10:36

This may not be as fast as the answer Sammywemmy suggested, nonetheless here is a generic function which works using Pandas functions. Note that explode function works only on one column at a time. So:

df = pd.DataFrame({'A': [1, 2], 'B': [['a','b'], ['c','d']], 'C': [['z','y'], ['x','w']]})

A    B     C
--------------
1 [a, b] [z, y]
2 [c, d] [x, w]

##Logic for multi-col explode
list_cols = {'B','C'}
other_cols = list(set(df.columns) - set(list_cols))
exploded = [df[col].explode() for col in list_cols]
df2 = pd.DataFrame(dict(zip(list_cols, exploded)))
df2 = df[other_cols].merge(df2, how="right", left_index=True, right_index=True)

A B C
------
1 a z
1 b y
2 c x
2 d w

explode multiple rows

3 Answers3

Linked