How to add a matrix A to another matrix B with a different size and it should maintain rows of B which do not have the first two elements equal to the first two elements of a row in matrix A?
Show older comments
I got the following problem. if I have the following two matrices:
A=[1 2 1
3 4 1
7 8 1]
B=[1 2 0
2 6 0
3 4 0
4 9 0
5 2 0
6 1 0
7 8 0]
I need to add matrix A to matrix B with remaining the rows of matrix B. Only when the first two elements are the same, the row should be replaced. I need a matrix:
C=[1 2 1
2 6 0
3 4 1
4 9 0
5 2 0
6 1 0
7 8 1]
It has to work for a 41120x3 matrix.
1 Comment
Azzi Abdelmalek
on 11 Jun 2013
Glenn, give an example that is closer to your case
Answers (3)
Azzi Abdelmalek
on 11 Jun 2013
C=B
C(A(:,1),:)=A
3 Comments
Glenn Roumen
on 11 Jun 2013
Azzi Abdelmalek
on 11 Jun 2013
Give a sample of your real data
Glenn Roumen
on 11 Jun 2013
Roger Stafford
on 11 Jun 2013
[tf,loc] = ismember(B(:,1:2),A(:,1:2),'rows');
t = loc(tf);
C = B;
C(t,:) = A(t,:);
Note that if B contains more than one row with the same first two elements, this will select the one with the higher index.
5 Comments
Glenn Roumen
on 11 Jun 2013
Roger Stafford
on 11 Jun 2013
Edited: Roger Stafford
on 11 Jun 2013
In what way does it not work? Does it give an error message or does it finish execution but simply make the wrong substitutions? In the latter case, can you provide a simple example with only a few elements in which its substitutions are in error?
Looking at your most recent example I notice that either your arrays have five columns or you are using commas for decimal points. Does your system properly handle commas in that way?
Also with floating point decimal fractions there is always the problem of numbers that look equal in their display but are actually different down in their least significant bits due to differing rounding procedures in producing them. The 'ismember' function demands exact equality.
Glenn Roumen
on 11 Jun 2013
Roger Stafford
on 11 Jun 2013
A thousand pardons! My own code is in error. I should have written this:
[tf,loc] = ismember(B(:,1:2),A(:,1:2),'rows');
C = B;
C(tf,3:end) = A(loc(tf),3:end);
Roger Stafford
on 11 Jun 2013
If I understand you correctly, the reason you are getting a 41120 x 5 array is that your system is NOT handling commas as you expect! It is interpreting them is dividing symbols between separate elements of a five-element row. Try this simple experiment on your machine:
x = [1,2 3,4]
and see how many elements x has and what they are. If there are two and they are fractional, your system is dealing with commas properly. If there are four and they are all integers, it is not using commas properly.
Azzi Abdelmalek
on 11 Jun 2013
for k=1:size(B,1)
ii=find(ismember(A(:,1:2),B(k,1:2),'rows'),1);
if ~isempty(ii)
B(k,:)=A(ii,:);
end
end
1 Comment
Glenn Roumen
on 11 Jun 2013
Categories
Find more on Logical in Help Center and File Exchange
on 11 Jun 2013
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!
