根据条件进行数据帧过滤

Question

I have a data-frame like

colA  colB  colC
 A      B    C 
 A      D    C 
 B      B    E 
 A      D    C 
 C      B    C 

I want to filter them in a priority like this: If colC == E then return E, after that check colB == D return D otherwise return colA The output is

colA  colB  colC   final
 A      B    C      A 
 A      D    C      D
 B      B    E      E 
 A      D    C      D
 C      B    C      C

Answer 1

Create the condition Series , the chain with bfill and fillna

s=pd.Series({'colB':'D','colC':'E'})
df['New']=df.where(df.eq(s)).bfill(1).iloc[:,0].fillna(df.colA)

>>> df
  colA colB colC New
0    A    B    C   A
1    A    D    C   D
2    B    B    E   E
3    A    D    C   D
4    C    B    C   C

Answer 2

You could use np.select , which allows you to select among multiple values depending on a list of conditions:

m1 = df.colC =='E'
m2 = df.colB =='D'
df.loc[:,'final'] = np.select([m1,m2], ['E', 'D'], default=df.colA)

   colA colB colC final
0    A    B    C     A
1    A    D    C     D
2    B    B    E     E
3    A    D    C     D
4    C    B    C     C

Answer 3

My favorite is to use a chained mask() , like this:

df["final"] = df["colA"] \
              .mask(df["colB"].eq("D"), "D") \
              .mask(df["colC"].eq("E"), "E")

This is to present your if-then-elif sequence in exact reverse order of checking, but otherwise very readable.

Answer 4

Don't take this seriously

I'm just experimenting

a = df.colA.values.copy()                # Set lowest priority first
a[np.flatnonzero(df.colB == 'D')] = 'D'  # And on down the line
a[np.flatnonzero(df.colC == 'E')] = 'E'  # Highest priority last

df.assign(New=a)

  colA colB colC New
0    A    B    C   A
1    A    D    C   D
2    B    B    E   E
3    A    D    C   D
4    C    B    C   C

Answer 5

using np.where

t['final'] = np.where(t['colC'] == 'E', 'E', (np.where(t['colB'] == 'D', 'D', t['colA'])))

Output

  colA colB colC final
0    A    B    C     A
1    A    D    C     D
2    B    B    E     E
3    A    D    C     D
4    C    B    C     C