Efficient method for populating a matrix based on its equality with another matrix

Hello. I'm having a problem populating a matrix and would greatly appreciate some help.

If we have:

if true

    A =  10 11 3 4    B =  10 11 NaN NaN
         12 11 4 6         12 11 NaN NaN
         13 13 2 8         13 11 NaN NaN
  end

What I want is the following: if A(i,1:2) and B(i,1:2) are the same, then B(i,3:4) = A(i,3:4). So something like this:

if true

B =  10 11 3 4
     12 11 4 6
     13 11 NaN NaN

end

I've managed to create a loop that achieves this result. Which is the following:

if true

    for i = 1:length(B); 
      for j = 1:length(A);
           if B(i,1) == A(j,1) && B(i,2) == A(j,2);
                  B(i,3) = A(j,3);
           end
      end
    end

end

The problem is, the two arrays I'm working with have close to 3 million rows so a double conditioned loop is extremely slow. Is there a vectorised way to achieve the same result that is significantly quicker?

Any help would be greatly appreciated.

Thanks

Andrew

Answer 1

[Andrei Bobrov]

[lo,ii] = ismember(A(:,1:2),B(:,1:2),'rows');
B(ii(lo),:) = A(lo,:);

or

lo = all(A(:,1:2) == B(:,1:2),2);
B(lo,:) = A(lo,:);

[Andrew Barnden]

Thank you Andrei. That worked perfectly. It's the first time I've worked on a large dataset and I'm quickly realising the impracticality of loops!

Out of interest, can this problem be handled using the 'intersect' function as well? I spent a bit of time looking at it this evening but couldn't figure out a solution.

Thank you once again.

[Andrei Bobrov]

yes, only for case from your example:

[~,ia,ib] = intersect(A(:,1:2),B(:,1:2),'rows');
B(ib,:) = A(ia,:)

best variant - use ismember:

A =  [10 11 3 4    
      12 11 4 6    
      13 13 2 8]         

B = [12 11 NaN NaN
     10 11 NaN NaN
     12 11 NaN NaN
     13 11 NaN NaN]

[lo,ii] = ismember(B(:,1:2),A(:,1:2),'rows');
B(lo,:) = A(ii(lo),:)