Order numbers column by highest number, regardless of positive or negative-CodePudding

I have a small dataframe on synthetic scores and the difference between the two:

score_1    score_2    diff
10         15         5
7          9          2
13         23         10
12         4          -8
...

There are quite a few rows so I want to order the dataframe by the largest difference between score_1 and score_2. I'm currently using:

df.sort_values(by=['diff'], ascending=False)

But this returns:

score_1    score_2    diff
13         23         10
10         15         5
7          9          2
12         4          -8

Whereas what I'm looking for is:

score_1    score_2    diff
13         23         10
12         4          -8
10         15         5
7          9          2

For it to return the biggest difference in descending order regardless of whether it is a negative or positive difference

CodePudding user response：

Use:

df = df.sort_values(by=['diff'], ascending=False, key=lambda x: x.abs())
print (df)
   score_1  score_2  diff
2       13       23    10
3       12        4    -8
0       10       15     5
1        7        9     2

CodePudding user response：

Use:

df2 = df.sort_values(by='diff', ascending=False, key=abs)

or, alternatively:

df2 = df.loc[df['diff'].abs().sort_values(ascending=False).index]

output:

   score_1  score_2  diff
2       13       23    10
3       12        4    -8
0       10       15     5
1        7        9     2

CodePudding user response：

The two other answers are more pertinent but I give you another way with rank:

>>> df.iloc[df['diff'].abs().rank(method='dense', ascending=False).sub(1)]
   score_1  score_2  diff
2       13       23    10
3       12        4    -8
0       10       15     5
1        7        9     2