How can I sort the output matrix according to the prescribed order in a column?

Hi all,

I have my output matrix data in nx3 dimension.

Now I wish to sort these data,neither in the ascending nor descending order in a column, but according to the prescribed order in a column.

For example, if my matrix is

1  0.5  0.8 
10 0.6  0.5
11 0.3  0.1
13 0.1  0.8

Now I want to sort the matrix according to the prescribed order of "11 10 1 13" in column 1, that is, the sorted matrix is expected to be

11 0.3  0.1
10 0.6  0.5
1  0.5  0.8  
13 0.1  0.8

What is your suggestion?

Thanks in advance.

Fei

Answer 1

[dpb]

Presuming you have (or can generate) the desired order vector, then it's pretty much trivial--just find where they're located in the original array. This assumes no ties.

Mat
>> ix=[11 10 1 13]';
%
%[erratum -- dpb]
%  following returns the locations in the index vector, NOT in the array
>> [~,ib]=ismember(M(:,1),ix);   % INCORRECT original posting
%
%   It was intended to interchange the order of arguments as follows
>> [~,ib]=ismember(ix,M(:,1));   % CORRECT 
%[ end erratum]
>> M(ib,:)
ans =
 11.0000    0.3000    0.1000
 10.0000    0.6000    0.5000
  1.0000    0.5000    0.8000
 13.0000    0.1000    0.8000
>>

[Fei]

Thanks dpb.

This works efficiently, however I think there might be an error in your code.

[~,ib]=ismember(M(:,1),ix)

returns ib which is the location of M(:,1) found in ix, e.g., '1' is in ix(3), therefore you have ib=3; then you output M(3,:) as the first row but this is not logical because no one tells us M(3,:) should be in the first row, although the final result is right by chance.

Now suppose I add a 5th row in my original matrix, say the matrix is

1  0.5  0.8 
10 0.6  0.5
11 0.3  0.1
13 0.1  0.8
9  0    15

Now I wish to output in the order like [[11 9 10 1 13], using the code above will yield

 13.0000    0.1000    0.8000
   11.0000    0.3000    0.1000
    1.0000    0.5000    0.8000
    9.0000         0   15.0000
   10.0000    0.6000    0.5000

I think the right one is

[~,ib]=ismember(ix,M(:,1))

We output the location of ix found in M(:,1), and then sequentially print the data according to the location. For instance, '11' should be in the first row according to the prescribed order, and it is M(3:1), therefore we have ib=3 and print M(3,:) in the first row.

   M=[1  0.5  0.8 
  10 0.6  0.5
  11 0.3  0.1
  13 0.1  0.8
  9  0 15];

ix=[11 9 10 1 13];

[lia,locb]=ismember(ix,M(:,1))

   M(locb,:)

output is

lia =

       1     1     1     1     1

locb =

       3     5     2     1     4

ans =

     11.0000    0.3000    0.1000
      9.0000         0   15.0000
     10.0000    0.6000    0.5000
      1.0000    0.5000    0.8000
     13.0000    0.1000    0.8000

[dpb]

_This works efficiently, however I think there might be an error in your code.

[~,ib]=ismember(M(:,1),ix)

...the right one is

[~,ib]=ismember(ix,M(:,1)) _

That is correct, sorry--I did inadvertently swap the order of which is in which in the argument list and by chance the original sample case didn't show it up...

Answer 2

[Andrei Bobrov]

a = [1  0.5  0.8
10 0.6  0.5
11 0.3  0.1
13 0.1  0.8
9  0    15];
b = [11 9 10 1 13];

[~,ia] = sort(a(:,1));
[~,ib] = sort(b);
out = a(ia(ib),:);

ADD

[~,ia] = sort(a(:,1));
[~,ib] = sort(b);
ii = sortrows([ia(:),ib(:)],2);
out = a(ii(:,1),:);

[Fei]

Thank you Andrei.

This seems working well also. However I don't understand why

 a = [1  0.5  0.8
      10 0.6  0.5
      11 0.3  0.1
      13 0.1  0.8
      9  0    15];

 [~,ia] = sort(a(:,1))

would give

ia =

       1
       5
       2
       3
       4

[Andrei Bobrov]

see ADD part in my answer