Pandas turn True columns into column value

user_9871643 0
Viewed 162

I have the following table of boolean values:

pd.DataFrame(data={'val1': [True, False, False, True], 
                   'val2': [False, True, False, True], 
                   'val3': [True, True, False, True], 
                   'val4': [True, False, True, False], 
                   'val5': [True, True, False, False],
                   'val6': [False, False, True, True]},
                   index=pd.Series([1, 2, 3, 4], name='index'))
index val1 val2 val3 val4 val5 val6
1 True False True True True False
2 False True True False True False
3 False False False True False True
4 True True True False False True

I want to create a new dataframe with the same indices, but each row has the first three True column names from the previous column.

index TrueVal1 TrueVal2 TrueVal3
1 val1 val3 val4
2 val2 val3 val5
3 val4 val6 NaN
4 val1 val2 val3

If a row has fewer than three True values, the new dataframe will have Null values.

pandas python
2 Answers

Try with dot , then split

#df = df.set_index('index')
out = df.dot(df.columns + ',').str[:-1].str.split(',',expand=True).iloc[:,:3]
out
Out[258]: 
          0     1     2
index                  
1      val1  val3  val4
2      val2  val3  val5
3      val4  val6  None
4      val1  val2  val3
user_79645270

You can do this with numpy + argsort to slice the columns by the position of the True values.

Then use where to NaN rows that were False incase there were too few True values.

import numpy as np
import pandas as pd

# Get the first `N` True columns.
N = 3

arr = df.to_numpy()
data = df.columns.to_numpy()[(~arr).argsort(axis=1, kind='stable')[:, :N]]
mask = np.ones_like(data).cumsum(1) <= arr.sum(1)[:, None]

res = pd.DataFrame(data, columns=[f'TrueVal{i+1}' for i in range(N)],
                   index=df.index).where(mask)

print(res)

      TrueVal1 TrueVal2 TrueVal3
index                           
1         val1     val3     val4
2         val2     val3     val5
3         val4     val6      NaN
4         val1     val2     val3
user_43333590
Related