How add multiple columns with a condition Using np.where()

Question

I know how to use np.where() to add one column by 1 condition:

import pandas as pd
import numpy as np
df=pd.read_csv(file,nrows=5)
df['new_col1']= np.where(df['col1'] < '100', 1,2)
df.head()

output:

   col1  col2  new_col1
0     1     3    1
1     2     4    1

what if I want to add 2 columns by the same condition:

df['new_col1'],df['new_col2']= np.where(df['col1'] < '100', (1,2),(3,4))

I want to add new_col1 and new_col2,the result are (1,2),(3,4)

When I tried this code, I received:

ValueError: too many values to unpack (expected 2)

The output should be:

   col1  col2  new_col1 new_col2
0     1     3    1       3
1     2     4    1       3

`np.where` returns one value. Could you elaborate as to how you want to generate two values to add instead? — ifly6, Jun 16 '21 at 21:14
Thank you for your reply ,what if I want to add 2 columns by 1 condition ,what else I need to use? — William, Jun 16 '21 at 21:14
I don't understand what you mean by 'add 2 columns by 1 condition'. Could you give an example of this? — ifly6, Jun 16 '21 at 21:15
df['column1'],df['column2']= np.where(df['contract'] > '0L000099', 1,2) — William, Jun 16 '21 at 21:16
That isn't an explanation. Can you give an explanation in terms of a single row and a function? — ifly6, Jun 16 '21 at 21:16
What is your expected output? They will have the same values is that fine by you? — Andreas, Jun 16 '21 at 21:17
Just use `df['column2'] = df['column1']` after defining column1 by the np.where above ? — SeaBean, Jun 16 '21 at 21:21
Sorry for the not clear,I have edited the question, please help! — William, Jun 16 '21 at 21:29
@William, updated my answer to provide the result as per your updated question. — Andreas, Jun 16 '21 at 21:43

Andreas · Accepted Answer · 2021-06-16T21:41:22.677

1

You can use the condition multiple times:

mask = df['contract'] > '0L000099'
df['column1'] = np.where(mask, 1, 2)
df['column2'] = np.where(mask, 3, 4)

or even invert the condition:

df['column2'] = np.where(~mask, 1, 2)

Since your question was updated, here the updated answer, however I am not sure thats actually usefull:

import pandas as pd
df = pd.DataFrame({'test':range(0,10)})
mask  = df['test'] > 3
m_len = len(mask)

df['column1'], df['column2'] = np.where([mask, mask], [[1]*m_len, [3]*m_len], [[2]*m_len, [4]*m_len])

   test  column1  column2
0     0        2        4
1     1        2        4
2     2        2        4
3     3        2        4
4     4        1        3
5     5        1        3
6     6        1        3
7     7        1        3
8     8        1        3
9     9        1        3

edited Jun 16 '21 at 21:41

answered Jun 16 '21 at 21:21

Andreas

8,694
3
14
38

Thank you for your answer ,can you please give me some explain about what dose this line mean :[1]*m_len, [3]*m_len], [[2]*m_len, [4]*m_len] – William Jun 16 '21 at 23:21
@William, ofc, it is a numpy specific thing called broadcasting if you want to look it up. Essentially numpy is very efficient because of vectorization, because of this numpy expects certain formats to use that. In this case it needs for example one `1` for each boolean in the condition. Therefore we have to repeat the value, which we can do by multiplying with the length of the condition series. Hope that makes sense. – Andreas Jun 17 '21 at 10:11
Hi friend can you help me with this question?https://stackoverflow.com/questions/68476193/how-to-merge-2-pandas-daataframes-base-on-multiple-conditions-faster – William Jul 21 '21 at 20:39
@William, hey William, see my answer under your question. Let me know how it went. Happy coding! – Andreas Jul 21 '21 at 22:01

How add multiple columns with a condition Using np.where()

1 Answers1