[FIXED] np.select join all true values together

Issue

I have a dataframe and would like to check for each row which of my conditions are true. If multiple are true I would like to return all of those choices with np.select. How can I do this?

df = pd.DataFrame({'cond1':[True, True, False, True],
                   'cond2':[False, False, True, True],
                   'cond3':[True, False, False, True],
                   'value': [1, 3, 3, 6]})

conditions = [df['cond1'] & (df['value']>4),
             df['cond2'],
             df['cond2'] & (df['value']>2),
             df['cond3'] & df['cond2']]

choices     = [ '1', '2', '3', '4']

df["class"] = np.select(conditions, choices, default=np.nan)

I get this

   cond1    cond2   cond3   value   class
0   True    False   True    1       nan
1   True    False   False   3       nan
2   False   True    False   3       2
3   True    True    True    6       1

but would like to get this

   cond1    cond2   cond3   value   class
0   True    False   True    1       nan
1   True    False   False   3       nan
2   False   True    False   3       2 and 3
3   True    True    True    6       1 and 2 and 3 and 4

Solution

You can use this trick with np.dot:

df1 = pd.DataFrame(conditions, columns=df.index, index=choices).T
df['class'] = df1.dot(df1.columns + ' and ').str.strip(' and ')

Output:

>>> df
   cond1  cond2  cond3  value                class
0   True  False   True      1                     
1   True  False  False      3                     
2  False   True  False      3              2 and 3
3   True   True   True      6  1 and 2 and 3 and 4

Intermediate result:

>>> df1
       1      2      3      4  <- columns from choices
0  False  False  False  False
1  False  False  False  False
2  False   True   True  False
3   True   True   True   True
^- index from df.index

Answered By - Corralien