Sort Matrix Array and skip zeros.

I have an array as this:

   Array1 = [1  2   3  4   0  0;      
             1  2   3  0   0  0;    
             1  0   0  0   0  0;      
             1  2   0  0   0  0];

Array1 is sorted and is fine as it is. But lets say I type in a mistake like this:

   Array2 = [2  1  3  4  0   0;      
             1  2  3  0  0   0;    
             1  0  0  0  0   0;      
             1  2  0  0  0   0];

Now coulomb 1 in Array2 is not sorted and I would like to sort it. But when I try to sort it with the function sort(Array2(1,:)) the zeros (0) will be listed first. Its easy to understand why it does this because 0 is smaller then 1,2,3,4, etc. It will then look like:

   Array2 = [0  0  1  2  3   4;      
             1  2  3  0  0   0;    
             1  0  0  0  0   0;      
             1  2  0  0  0   0];

But I would like the array to look like Array1:

   Array2 = [1  2  3  4  0   0;      
             1  2  3  0  0   0;    
             1  0  0  0  0   0;      
             1  2  0  0  0   0]

How is this possible? Can I somehow sort an array and skip the zeros?

Your Sincerely.

[Adam]

Do you want the 0's to always be at the end of the rows or to stay where they are while everything else is sorted around them?

e.g. if you have a row

1 3 0 2 4

do you want

1 2 3 4 0

or

1 2 0 3 4?

[Kalle]

I would like the zeros to always be on the end. So if its 1 3 0 2 4 they should be sorted as 1 2 3 4 0.

[Guillaume]

Adam's or the second option in my answer will give you that.

[Kalle]

Ye I noticed. Thanks. :)

Answer 1

[Adam]

Array1( Array1 == 0 ) = NaN; 
Array2 = sort( Array1, 2 ); 
Array2( isnan( Array2 ) ) = 0;

works if you want 0s shuffled to the end, though not if you want them to remain exactly where they are.

[Kalle]

This is exactly what I was looking for. Thanks. :)

[Kalle]

Right now I use the zeros for filling the Matrix and in the end I need to implement it in a c function and test it in a lab. Nice and clean way to sort the Array. Thanks.

Answer 2

[Guillaume]

One possible way:

for row = 1:size(Array2, 1)
   Array2(row, 1:nnz(Array2(row, :))) = nonzeros(sort(Array2(row, :)));
end

Or if zeros are not always at the end in the original matrix:

ncol = size(Array2, 2);
for row = 1:size(Array2, 1)
  Array2(row, :) = [nonzeros(sort(Array2(row, :))); zeros(ncol - nnz(Array2(row, :)), 1)];
end

[Kalle]

Thanks for your time, but Adam answered my question. :)

Answer 3

[Ilham Hardy]

If you know the starting value (say sort start from 1, instead of 0):

tblA = [2  1  3  4  0   0];

vStartSort = 1
tblA = sort(tblA);
tblA = circshift(tblA, [1, -find(tblA == vStartSort, 1) + 1]);

If you just want to skip zero (you don't know the smallest value after zero in array)

tblA = [2  1  3  4  0   0];

tblA = sort(tblA);
minGreaterThanZero = tblA(tblA>0);
vStartSort = minGreaterThanZero(1);
tblA = circshift(tblA, [1, -find(tblA == vStartSort, 1) + 1]);

HTH, IH

[Kalle]

Thanks for your time Ilham Hardy. Adam gave me exactly what I was looking for. Maybe that wasnt so clear after all. But thanks anyways. :) Always nice to learn other methods.

Answer 4

[Image Analyst]

This will give you exactly what you want, with zeros remaining in the original locations.

Array2 = [2  1  3  4  0   0;
  1  2  3  0  0   0;
  1  0  0  0  0   0;
  1  2  0  0  0   0]
% Find locations of non-zeros.
% Transpose array first so we can sort column-major fashion.
nonZeroIndexes = Array2(:).'~=0
% Extract the values from those locations.
nonZeroValues = Array2(nonZeroIndexes)
% Sort them.
sortedValues = sort(nonZeroValues)
% Get an array with the zeros in their original locations.
out = Array2(:); % Initialize
% Assign only those elements in non-zero locations.
% This will still be a column vector.
out(nonZeroIndexes) = sortedValues(:)
% Reshape to original 2D size of Array2.
out = reshape(out, size(Array2))

In command window.

out =
     1     2     3     4     0     0
     1     2     3     0     0     0
     1     0     0     0     0     0
     1     2     0     0     0     0

[Kalle]

This is a great solution if the zeros should be in the original locations. Right now I need the zeros to be placed in the end of the Matrix, since the zeros is only for filling the matrix. Adam gave me a simple answer to this. Thanks for your time though.