From a matrix, remove a number from each row (in this case " Inf " )

Hello Matlab practitioner...I have a matrix a =

     3     4   Inf     2
     3     4     1   Inf
     4   Inf     1     2

I want a matrix

b =

     3     4     2
     3     4     1
     4     1     2

where each Inf in every row is removed. There is exactly one Inf in every row.

Thank you.

Answer 1

[Rik]

I tested it with an extra row:

 A=[  3     4   Inf     2
     3     4     1   Inf
     4   Inf     1     2
     4     4     4   Inf];
 temp_A=A';
 temp_A(isinf(temp_A))=[];
 A=reshape(temp_A,size(A,2)-1,size(A,1))';

[Rik]

A speedtest of my method, the method Image Analyst suggested, and another one:

Elapsed time is 0.000539 seconds.
Elapsed time is 0.000616 seconds.
Elapsed time is 0.001997 seconds.

code:

tic
A=[3     4   Inf     2
    3     4     1   Inf
    4   Inf     1     2
    4     4     4   Inf];
temp_A=A';
temp_A(isinf(temp_A))=[];
A=reshape(temp_A,size(A,2)-1,size(A,1))';
toc
tic
m=[3     4   Inf     2
    3     4     1   Inf
    4   Inf     1     2
    4     4     4   Inf];
[rows, columns] = size(m);
infLocations = isinf(m);
out=zeros(rows,columns-1);
for row = 1 : rows
    out(row,:) = m(row, ~infLocations(row, :));
end
toc
tic
mat=[3     4   Inf     2
    3     4     1   Inf
    4   Inf     1     2
    4     4     4   Inf];
mat=mat2cell(mat,ones(size(mat,1),1),size(mat,2));
mat=cellfun(@(x) x(~isinf(x)),mat,'UniformOutput',false);
mat=cell2mat(mat);
toc

Answer 2

[Image Analyst]

Here's another way:

m=[...
  3     4   Inf     2
  3     4     1   Inf
  4   Inf     1     2]
[rows, columns] = size(m);
infLocations = isinf(m)
for row = 1 : rows
  out(row,:) = m(row, ~infLocations(row, :))
end
out % Print to command window.

[Rik]

You forgot (or just chose not) to pre-allocate out (which will cut about 1ms with this example)

[Image Analyst]

I chose not to. Often times I don't preallocate when dealing with such tiny arrays. Preallocation takes time also. I got around 0.000013 to 0.000014 seconds when not preallocating, and tried it hundreds of times with preallocation and always got a longer time of around 0.000014 to 0.000020 when preallocating.

[Anthony Klinkert]

Image Analyst, this looks good too! Thanks!