Getting ValueError - non-unique multi-index when using explode() in Pandas-CodePudding

I want to split two columns at their comma and bring them back to the original pandas dataframe. I tried to explode() but I got an error with ValueError: cannot handle a non-unique multi-index! I wonder how I can overcome this error.

import pandas as pd
data = {'fruit_tag': {0: 'apple, organge', 1: 'watermelon', 2: 'banana', 3: 'banana', 4: 'apple, banana'}, 'location': {0: 'Hong Kong , London', 1: 'New York, Tokyo', 2: 'Singapore', 3: 'Singapore, Hong Kong', 4: 'Tokyo'}, 'rating': {0: 'bad', 1: 'good', 2: 'good', 3: 'bad', 4: 'good'}, 'measure_score': {0: 0.9529434442520142, 1: 0.952498733997345, 2: 0.9080725312232971, 3: 0.8847543001174927, 4: 0.8679852485656738}}
dt = pd.DataFrame.from_dict(data)
dt.\
    set_index(['rating', 'measure_score']).\
    apply(lambda x: x.str.split(',').explode())

CodePudding user response：

When you explode, the index are the same for (each) old rows. Pandas doesn't know (or like) to align these indexes, because intention of users can be different from case to case, e.g. align by order, or cross merge. In your case, for example, what do you expect to get from row 1 where you have 2 entries for each column? How about row 2?

If you want a cross merge, you would need to explode manually:

def explode(x, col): return x.assign(**{col:x[col].str.split(', ')}).explode(col)

explode(explode(dt, 'fruit_tag'), 'location')

Output:

    fruit_tag    location rating  measure_score
0       apple  Hong Kong     bad       0.952943
0       apple      London    bad       0.952943
0     organge  Hong Kong     bad       0.952943
0     organge      London    bad       0.952943
1  watermelon    New York   good       0.952499
1  watermelon       Tokyo   good       0.952499
2      banana   Singapore   good       0.908073
3      banana   Singapore    bad       0.884754
3      banana   Hong Kong    bad       0.884754
4       apple       Tokyo   good       0.867985
4      banana       Tokyo   good       0.867985