How to combine data from two columns based on multipe conditions in pandas?

Question

I have this dataframe where I am trying to combine Email_x and Email_y to be a column email.

If both are NaN then the result should be "" or np.nan.
Email_y has a value and Email_x does not then use Email_y as the result(vice versa).
If both have values then take the first one only.

                Email_x            Email_y Verification status_x Verification status_y
489   [email protected]                NaN                 valid                 valid
975    [email protected]                NaN            accept_all                   NaN
1192                NaN                NaN                   NaN                   NaN
1370  [email protected]                NaN               unknown                   NaN
2001                NaN                NaN                   NaN                   NaN
2565  [email protected]  [email protected]                 valid                   NaN
3900                NaN                NaN                   NaN                   NaN
3998                NaN                NaN                   NaN                   NaN
4192                NaN                NaN                   NaN                   NaN
4757                NaN   [email protected]                   NaN                   NaN

I have tried using this but it fails when both have a value and results in both the values being combined.

df["email"] = df[["Email_x", "Email_y"]].fillna("").agg(" ".join, axis=1)

Result:

                Email_x            Email_y Verification status_x Verification status_y                                email
489   [email protected]                NaN                 valid                 valid                   [email protected] 
975    [email protected]                NaN            accept_all                   NaN                    [email protected] 
1192                NaN                NaN                   NaN                   NaN                                     
1370  [email protected]                NaN               unknown                   NaN                   [email protected] 
2001                NaN                NaN                   NaN                   NaN                                     
2565  [email protected]  [email protected]                 valid                   NaN  [email protected] [email protected]
3900                NaN                NaN                   NaN                   NaN                                     
3998                NaN                NaN                   NaN                   NaN                                     
4192                NaN                NaN                   NaN                   NaN                                     
4757                NaN   [email protected]                   NaN                   NaN                     [email protected]

Expected result:

                Email_x            Email_y Verification status_x Verification status_y                                email
489   [email protected]                NaN                 valid                 valid                   [email protected] 
975    [email protected]                NaN            accept_all                   NaN                    [email protected] 
1192                NaN                NaN                   NaN                   NaN                                     
1370  [email protected]                NaN               unknown                   NaN                   [email protected] 
2001                NaN                NaN                   NaN                   NaN                                     
2565  [email protected]  [email protected]                 valid                   NaN                    [email protected]
3900                NaN                NaN                   NaN                   NaN                                     
3998                NaN                NaN                   NaN                   NaN                                     
4192                NaN                NaN                   NaN                   NaN                                     
4757                NaN   [email protected]                   NaN                   NaN                     [email protected]

Sample dictionary:

import numpy as np

a = {
    "Email_x": {
        489: "[email protected]",
        975: "[email protected]",
        1192: np.nan,
        1370: "[email protected]",
        2001: np.nan,
        2565: "[email protected]",
        3900: np.nan,
        3998: np.nan,
        4192: np.nan,
        4757: np.nan,
    },
    "Email_y": {
        489: np.nan,
        975: np.nan,
        1192: np.nan,
        1370: np.nan,
        2001: np.nan,
        2565: "[email protected]",
        3900: np.nan,
        3998: np.nan,
        4192: np.nan,
        4757: "[email protected]",
    },
    "Verification status_x": {
        489: "valid",
        975: "accept_all",
        1192: np.nan,
        1370: "unknown",
        2001: np.nan,
        2565: "valid",
        3900: np.nan,
        3998: np.nan,
        4192: np.nan,
        4757: np.nan,
    },
    "Verification status_y": {
        489: "valid",
        975: np.nan,
        1192: np.nan,
        1370: np.nan,
        2001: np.nan,
        2565: np.nan,
        3900: np.nan,
        3998: np.nan,
        4192: np.nan,
        4757: np.nan,
    },
}

Why not simply use `fillna`? Check `df['Email_x'].fillna(df['Email_y'])` — Shubham Sharma, May 09 '21 at 07:30
@ShubhamSharma because the value for `email` can come from `Email_y` as well. — yudhiesh, May 09 '21 at 07:34

score 2 · Answer 1 · answered May 09 '21 at 07:23

2

Via np.select

condlist = [
    (df.Email_x.isna()) & (~df.Email_y.isna()), # 1st column NAN but 2nd is not 
    (df.Email_y.isna()) & (~df.Email_x.isna()), # 2nd column NAN but 1st is not
    (~df.Email_x.isna()) & (~df.Email_y.isna()) # both is not NAN
]

choicelist = [
    df.Email_y,
    df.Email_x,
    df.Email_x
]

df['Email'] = np.select(condlist,choicelist, default=‘') # default value ''

answered May 09 '21 at 07:23

Nk03

14,699
2
8
22

This works but is a bit length as I need to do this for 4 more columns. If no one comes up with a better answer I will accept it :) – yudhiesh May 09 '21 at 07:38

score 1 · Accepted Answer · answered May 09 '21 at 07:34

1

Here is a great answer, which can be achieved with combine_first().

df['email'] = df['Email_x'].combine_first(df['Email_y'])

answered May 09 '21 at 07:34

r-beginners

31,170
3
14
32

This doesn't work when the value in `Email_y` is needed. – yudhiesh May 09 '21 at 07:35
Did you see in the question the requirement for Email_y? – r-beginners May 09 '21 at 07:40
If you see the last row of the `expected result` it is shown there and I mentioned it in point 2 but will edit it for better understanding. – yudhiesh May 09 '21 at 07:41
Correction confirmed. Did you run my code? It seems to meet your requirements. – r-beginners May 09 '21 at 07:45
Yes it does my mistake, works very well thanks! – yudhiesh May 09 '21 at 07:48

score 1 · Answer 3 · answered May 09 '21 at 07:46

You can use .mask(), as follows:

df['email'] = df['Email_x'].mask(df['Email_x'].isna(), df['Email_y'])

It will retain the value of df['Email_x'] if the condition is false (i.e. not NaN) and replace with value of df['Email_y'] if the condition is true (i.e. df['Email_x'] is NaN).

How to combine data from two columns based on multipe conditions in pandas?

Result:

Expected result:

Sample dictionary:

3 Answers3