pandas custom idxmax() function with special handling of equality

Question

I'm working on a csv file full of electoral data. My raw sample could be represented as:

        city      party1     party2     party3
   0    city1     50         107        114
   1    city2     181        323        326
   2    city3     26         28         75
   3    city4     32         47         59
   4    ciy5      8          21         21

I used the idxmax() function of pandas to create a new column, called "winner", like this :

 mydf['winner'] = mydf[['party1','party2','party3']].idxmax(axis=1)

My goal was to determine which party was in first position in each city. Here the result:

        city      party1     party2     party3      winner
   0    city1     50         107        114         party3
   1    city2     181        323        326         party3
   2    city3     26         28         75          party3
   3    city4     32         47         59          party3
   4    ciy5      8          21         21          party2

The winner's value of the last raw is false, because party2 and party3 have the same score.

Is it possible to include one exception with the function idxmax considers equality of two values and gives 'Equality'?

Answer 1

You can use DataFrame.eq for compare subset with DataFrame.max values per row, then sum them and where is value higher as 1 there are duplicates max. So then can be overwrite value of idxmax by mask with mask s > 1:

a = mydf[['party1','party2','party3']]
mydf['winner'] = a.idxmax(axis=1)

s = a.eq(a.max(axis=1), axis=0).sum(axis=1)
print (s)
0    1
1    1
2    1
3    1
4    2
dtype: int64

mydf['winner'] = mydf['winner'].mask(s > 1, 'Equality')
print (mydf)
    city  party1  party2  party3    winner
0  city1      50     107     114    party3
1  city2     181     323     326    party3
2  city3      26      28      75    party3
3  city4      32      47      59    party3
4   ciy5       8      21      21  Equality

If need also values multiple df by values of columns by mul, then apply join and last remove , by strip:

a = mydf[['party1','party2','party3']]
df = a.eq(a.max(axis=1), axis=0)
print (df)
  party1 party2 party3
0  False  False   True
1  False  False   True
2  False  False   True
3  False  False   True
4  False   True   True

mydf['winner'] = df.mul(df.columns.to_series())
                   .apply(','.join, axis=1)
                   .str.strip(',')
print (mydf)
    city  party1  party2  party3         winner
0  city1      50     107     114         party3
1  city2     181     323     326         party3
2  city3      26      28      75         party3
3  city4      32      47      59         party3
4   ciy5       8      21      21  party2,party3