CODEWITHSUNDEEP
pythonpython-2.7pandas
SBad
SBad

Reputation: 1345

conditionally Replace string in a Pandas series with another string

Take the below example. To replace one string in one particular column I have done this and it worked fine:

df = pd.DataFrame({'key': ['A', 'B', 'C', 'A', 'B', 'C'],
                   'data1': range(6),
                   'data2': ['A1', 'B1', 'C1', 'A1', 'B1', 'C1']},
                   columns = ['key', 'data1', 'data2'])

  key  data1 data2
0   A      0    A1
1   B      1    B1
2   C      2    C1
3   A      3    A1
4   B      4    B1
5   C      5    C1



df['data2']= df['data2'].str.strip().str.replace("A1","Bad")

  key  data1 data2
0   A      0    Bad
1   B      1    B1
2   C      2    C1
3   A      3    Bad
4   B      4    B1
5   C      5    C1

Q(1) How can we conditionally replace one string? Meaning that, in column data2, I would like to replace A1 but only if "key==A" and "data1">1. How can I do that?

Q(2) Can the conditional replacement be applied to multiple replacement (i.e., replacing A1 and A2 at the same time with "Bad" but only under similar conditions?

Upvotes: 1

Views: 3484

Answers (3)

SBad
SBad

Reputation: 1345

If we want to extend the example above in the following way:

df = pd.DataFrame({'key': ['A', 'B', 'C', 'A', 'B', 'C'],
                   'data1': range(6),
                   'data2': ['A1', 'B1', 'C1', 'A1', 'B1', 'C1']},
                   columns = ['key', 'data1', 'data2'])  

mask = (df['data1'] > 1)
df.loc[mask, 'data2']= df.loc[mask, 'data2'].str.strip().str.replace("A1",df['key']) 

  key  data1 data2
0   A      0    A1
1   B      1    B1
2   C      2   NaN
3   A      3   NaN
4   B      4   NaN
5   C      5   NaN

I am very surprised by the answer I thought the content of data2 would be replaced by content of column "key" (at the condition data1>1). any idea?

Upvotes: 0

asongtoruin
asongtoruin

Reputation: 10359

You can use numpy and a regex-based replacement to cover A1, A2 and more. if we extend your data to include an example with A3:

import pandas as pd
import numpy as np

df = pd.DataFrame({'key': ['A', 'B', 'C', 'A', 'B', 'C', 'A'],
                   'data1': range(7),
                   'data2': ['A1', 'B1', 'C1', 'A1', 'B1', 'C1', 'A3']},
                   columns=['key', 'data1', 'data2'])

df['data2'] = np.where((df['key'] == 'A') & (df['data1'] > 1),
                       df['data2'].str.replace(r'A\d+','Bad'),
                       df['data2'])

This returns:

  key  data1 data2
0   A      0    A1
1   B      1    B1
2   C      2    C1
3   A      3   Bad
4   B      4    B1
5   C      5    C1
6   A      6   Bad

Upvotes: 3

jezrael
jezrael

Reputation: 862661

I think need filter column in both sides with replace only for filtered rows:

mask = (df['key']=="A") &  (df['data1'] > 1)
df.loc[mask, 'data2']= df.loc[mask, 'data2'].str.strip().str.replace("A1","Bad")  

print (df)
  key  data1 data2
0   A      0    A1
1   B      1    B1
2   C      2    C1
3   A      3   Bad
4   B      4    B1
5   C      5    C1

If need multiple replace use replace with dict:

df = pd.DataFrame({'key': ['A', 'A', 'C', 'A', 'B', 'C'],
                   'data1': range(6),
                   'data2': ['A1', 'A2', 'C1', 'A1', 'B1', 'C1']},
                   columns = ['key', 'data1', 'data2'])

mask = (df['key']=="A") &  (df['data1'] > 0)
df.loc[mask, 'data2']= df.loc[mask, 'data2'].str.strip().replace({"A1":"Bad", "A2":'Bad1'})

Or use regex:

df.loc[mask, 'data2']= df.loc[mask, 'data2'].str.strip().str.replace(r'^A.*',"Bad")


print (df)
  key  data1 data2
0   A      0    A1
1   A      1  Bad1
2   C      2    C1
3   A      3   Bad
4   B      4    B1
5   C      5    C1

Upvotes: 3

Related Questions