CODEWITHSUNDEEP
pythonpandastuplessubset
EB2127
EB2127

Reputation: 1858

How to subset tuples in a pandas DataFrame, given lists of tuples?

I have the following pandas DataFrame. There are two columns A and B composed of lists of mutltiple tuples. 

import pandas as pd
dictionary_input = {'A' : [5,6,3,4], 
                    'B' : [[('AA', 4, 11), ('ABC', 28, 99), ('ABC', 23, 86)], [('AA', 2, 10)], [('ABC', 56, 76), ('BB', 15, 183)], [('BB', 15, 183)]], 
                    'C': [[('XYZ', 7, 9), ('XX',24, 33), ('BB', 179, 184)], [('XX',72, 75)], [('ABC',25, 45)], [('BB',91, 187)]]}

df = pd.DataFrame(dictionary_input)
print(df)

which results in:

   A                                            B                                            C
0  5  [(AA, 4, 11), (ABC, 28, 99), (ABC, 23, 86)]  [(XYZ, 7, 9), (XX, 24, 33), (BB, 179, 184)]
1  6                                [(AA, 2, 10)]                               [(XX, 72, 75)]
2  3               [(ABC, 56, 76), (BB, 15, 183)]                              [(ABC, 25, 45)]
3  4                              [(BB, 15, 183)]                              [(BB, 91, 187)]

My problem is that I would like to subset this DataFrame based on the values in the lists of tuples, i.e. based on a single tuple.

If I were to subset the dataframe based on B has tuple (BB, 15, 183), then the following would be the output:

   A                                            B                                            C
2  3               [(ABC, 56, 76), (BB, 15, 183)]                              [(ABC, 25, 45)]
3  4                              [(BB, 15, 183)]                              [(BB, 91, 187)]

I tried to accomplish this using 

df[df.B.isin(('BB', 15, 183))]

But this is wrong, as it gives me an empty DataFrame.

How do I subset based on values inside a list in pandas DataFrame, if the values are tuples?

Upvotes: 1

Views: 475

Answers (2)

Quang Hoang
Quang Hoang

Reputation: 150735

If you are working with pandas 0.25+, you can make use of explode, which make a series out of the list in each cell and concatenate them. similar to pd.concat(pd.Series(x) for x in df['B']), but keeps the original index. Then you can compare that series to your triple and groupby:

s = df['B'].explode()

df[(s == ('BB', 15, 183)).groupby(level=0).any()]

Output:

   A                               B                C
2  3  [(ABC, 56, 76), (BB, 15, 183)]  [(ABC, 25, 45)]
3  4                 [(BB, 15, 183)]  [(BB, 91, 187)]

Output (s):

0      (AA, 4, 11)
0    (ABC, 28, 99)
0    (ABC, 23, 86)
1      (AA, 2, 10)
2    (ABC, 56, 76)
2    (BB, 15, 183)
3    (BB, 15, 183)
Name: B, dtype: object

Upvotes: 2

Mahesh Sinha
Mahesh Sinha

Reputation: 91

You can do this by apply method:

df[df['B'].apply(lambda x: ('BB', 15, 183) in x)]

Upvotes: 1

Related Questions