# Delete rows in a matrix that contain ONLY a negative number

9 views (last 30 days)
Christina on 14 Jan 2013
Hello
I've got a really large matrix (say 2000 x 3000) and I want to delete the rows that contain ONLY the negative number -999.
For example, I've got A = [5 3 -999 ; -999 -999 -999; 4 1 7 ; -999 -999 -999]
and I want my new matrix to be A = [5 3 -999 ; 4 1 7]
How can I do that?
Thanks!

Walter Roberson on 14 Jan 2013
idx = all(A == -999, 2);
A(idx, :) = [];

Kye Taylor on 14 Jan 2013
Edited: Kye Taylor on 14 Jan 2013
Try
isNegRow = all( A==-999, 2 );
A(isNegRow,:) = [];

Shashank Prasanna on 14 Jan 2013
Edited: Shashank Prasanna on 14 Jan 2013
If you don't want to use loops and keep it simple in a single line, I suggest logical indexing as follows:
A(all(A==-999,2),:)=[]
This should do the trick.

Amith Kamath on 14 Jan 2013
Edited: Amith Kamath on 14 Jan 2013
I'm quite positive that this does the trick. There would probably be more optimal ways to do it!
X = 100.*rand(12,5); %for example.
X([2 5],:) = -999; % artificially create rows to remove.
i = 1;
while i <= size(X,1)
if(sum(X(i,:) == -999) == size(X,2))
X(i,:) = [];
else
i = i+1;
end
end
James Tursa on 15 Jan 2013
Edited: James Tursa on 15 Jan 2013
Using a loop in this manner is very bad for performance. At each row that is deleted, potentially huge amounts of data will need to be copied. The same data can end up being copied many, many times. For a large matrix with many rows being deleted, this can easily result in a performance hit that is one or more orders of magnitude slower than the other answers shown above. (This is basically the same problem as increasing the size of an array in a loop). Best to get the indices of everything to be deleted first and then delete it all at once as shown above.