How to design this for-loop?

Hello,

I'm writing a code for comparing the position of dots to one another and i'm having trouble coding for-loops for efficiency. This is the code:

for iS=1:length(RealSPs1) 
  for iC=1:NClust 
    D1(iS,iC) = min(abs(RealSPs1(iS)-SPs{iC})); 
  end
  end

for iS=1:length(RealSPs2) 
  for iC=1:NClust 
    D2(iS,iC) = min(abs(RealSPs2(iS)-SPs{iC})); 
  end
  end

This code works perfectly fine and D gives a nice 10x3 matrix. Which is what I want. However, I also have the same loop for D3 and, possibly up to D5 or D6. Besides the fact that it takes up a lot of space, the number of D can vary depending on NClust and adding more or less of these blocks of code when NClust has a different value seems very time consuming. So, is there a way to make a for-loop that stacks all iterations of D in some sort of 10x3xNClust array?

[Rik]

You should restructure your data at the point it gets into those variables. Using numbered variables is red flag, because you will be forced to use eval to deal with them.

Stephen Cobeldick is very pasionate about this topic (and with good reason), so I have taken the liberty of adapting some of his thoughts on this topic: Introspective programing (eval and other functions like it) is slow, and the MATLAB documentation warns specifically against it. Using eval is slow, makes debugging difficult, and removes all code helper tools (code checking, syntax hints, code completion, variable highlighting, etc.). A longer discussion can be found here. If you are not an expert and you think you need eval, there are probably better solutions to your problem.

Using cell arrays is a much better solution. An added benefit is that you can use length to create an outer loop.

[Stephen23]

"So, is there a way to make a for-loop that stacks all iterations of D in some sort of 10x3xNClust array?"

Yes, better code design is the solution.

Magically accessing variable names is slow, buggy, complex, insecure, and makes debugging difficult. But you actually have no need to do this: simply use one array and some indexing. Indexing is simple, neat, and highly efficient (unlike what you are proposing).

You should simply put your data into one ND numeric array (if the matrix sizes are compatible), or one cell array (if the number of points can vary between matrices). As you already know the size of the final array, why not simply preallocate an ND numeric array of the correct size and index into that?

Read more here:

https://www.mathworks.com/matlabcentral/answers/304528-tutorial-why-variables-should-not-be-named-dynamically-eval

@Rik Wisselink: I find that "passionate" distracts somewhat from the fact that TMW has clear guidelines advising to avoid string evaluation and all expert users give the same advice, because it makes it seem like one "passionate" old guy is something that can be safely ignored...

[Rik]

Fair enough. I'll change the text. I thought it was a good way to give credit where credit is due, but you are right, it might take away some of the message we both want to convey: people shouldn't use it.