How To: Multi-level subindexing

I'm tryig to achieve a proper 1-line ndex/sub-indexing operation, but have thus far failed to get the syntax right

ArrayA = sparse(100,100000);
ArrayA(randi(100*100000,[7000 1])) = (82-36).*rand(7000,1)+36;
[ArrayAFull(:,1),ArrayAFull(:,2),ArrayAFull(:,3)] = find(ArrayA);
ArrayB = sparse(100,100000);
ArrayB(randi(100*100000,[7000 1])) = (90-30).*rand(7000,1)+30;
[ArrayBFull(:,1),ArrayBFull(:,2),ArrayBFull(:,3)] = find(ArrayB);

[~,IdxA,IdxB] = intersect(ArrayAFull(:,[1,2]),ArrayBFull(:,[1,2]),'stable','rows');
t1 = ArrayBFull(ArrayBFull(IdxB,1)==1,:); %This is where things go sour
%The next two lines get me the actually desired matrix
t2 = ArrayBFull(IdxB,:);
t2 = t2(t2(:,1)==1,:); 
isequal(t1,t2); %false

Is it possible to compress the t2 lines into a single line, or no?

Answer 1

[Walter Roberson]

Yes, it is, but it is fairly ugly to do so. You can call upon subsref() manually, possibly having used substruct() to help build the appropriate structure.

It is usually a lot easier to use a helper function. For example,

SelectRowWhere = @(expression, where) expression(expression(:,1)==where,:);

t2 = SelectRowWhere(ArrayBFul(IdxB,:), 1)

[Gabriel Stanley]

yugh. You're right about the ugly part. I'll try this later today and see if it's faster or slower than the two-line implementation, then go with the faster of the two. Regardless, thank you for the education.

[Walter Roberson]

The two line version would typically be faster, as calling an anonymous function is notably slower than calling a non-anonymous function.

[Gabriel Stanley]

Understood. Ty for the heads-up.