Not able to change direction of sort on per column basis for DataFrame

[mattharrison]

The sort method on DataFrame supports sorting by multiple columns, yet only supports sorting in one direction (ascending or descending). It would be great to specify this. Perhaps something like this: df.sort(columns=['Name|descending', 'Age'] which would sort name first in descending order and then where names match use ascending order.

Also I see no mention of whether the sort is stable. If it is, this can be replicated by multiple sort calls.

[jhrhew]

I wish to have this feature too, which can be done in MS Excel.

[wesm]

I was thinking you would do

df.sort_index(by=['a', 'b'], ascending=[False, True])

any opinions?

[jhrhew]

Thanks for the quick response.
That looks quite intuitive and consistent with the current API if the list is short. However, I cannot think of any good idea for a longer list like
df.sort_index(by=['a', 'b', 'c', 'd', 'e'], ascending=[False, True, False, True, True])

[wesm]

About as short as you can make it I guess. ascending=[0, 1, 0, 1, 1] would work too

[jhrhew]

That's clever; it didn't occur to me. Thanks. Then, I think your proposal serves my need.

[wesm]

Got around to this finally:

In [2]: df
Out[2]: 
    A  B         C
0   2  0 -0.696943
1   2  2 -0.304504
2   0  1 -0.686636
3   1  4  0.298355
4   0  3 -1.167454
5   0  0  0.478933
6   2  3 -1.859343
7   2  4  0.040016
8   1  1  0.484016
9   0  4  0.355799
10  0  2 -0.127496
11  1  2  1.078274
12  2  1  1.920544
13  1  3  0.678504
14  1  0  0.210838

In [3]: df.sort
df.sort        df.sort_index  df.sortlevel   

In [3]: df.sort(['A', 'B'])
Out[3]: 
    A  B         C
5   0  0  0.478933
2   0  1 -0.686636
10  0  2 -0.127496
4   0  3 -1.167454
9   0  4  0.355799
14  1  0  0.210838
8   1  1  0.484016
11  1  2  1.078274
13  1  3  0.678504
3   1  4  0.298355
0   2  0 -0.696943
12  2  1  1.920544
1   2  2 -0.304504
6   2  3 -1.859343
7   2  4  0.040016

In [4]: df.sort(['A', 'B'], ascending=[1, 0])
Out[4]: 
    A  B         C
9   0  4  0.355799
4   0  3 -1.167454
10  0  2 -0.127496
2   0  1 -0.686636
5   0  0  0.478933
3   1  4  0.298355
13  1  3  0.678504
11  1  2  1.078274
8   1  1  0.484016
14  1  0  0.210838
7   2  4  0.040016
6   2  3 -1.859343
1   2  2 -0.304504
12  2  1  1.920544
0   2  0 -0.696943

In [5]: df.sort(['A', 'B'], ascending=[1, 1])
Out[5]: 
    A  B         C
5   0  0  0.478933
2   0  1 -0.686636
10  0  2 -0.127496
4   0  3 -1.167454
9   0  4  0.355799
14  1  0  0.210838
8   1  1  0.484016
11  1  2  1.078274
13  1  3  0.678504
3   1  4  0.298355
0   2  0 -0.696943
12  2  1  1.920544
1   2  2 -0.304504
6   2  3 -1.859343
7   2  4  0.040016

In [6]: df.sort(['A', 'B'], ascending=[0, 0])
Out[6]: 
    A  B         C
7   2  4  0.040016
6   2  3 -1.859343
1   2  2 -0.304504
12  2  1  1.920544
0   2  0 -0.696943
3   1  4  0.298355
13  1  3  0.678504
11  1  2  1.078274
8   1  1  0.484016
14  1  0  0.210838
9   0  4  0.355799
4   0  3 -1.167454
10  0  2 -0.127496
2   0  1 -0.686636
5   0  0  0.478933

[jhrhew]

Thanks a lot.