Using structures in functions within for loop

I've been cleaning up the workspace and editor a little by placing variables within structures. But I'm struggling with accessing and indexing structures and fields within structures in loops and functions within these loops.

Say I bundled up all key variables (1x1 doubles) into a structure, 'S' and pass them into a function, 'myfunction' which unpacks these variables, does some simple calculations (returning some 1x1doubles) then returns a structure of results, 'R'. Say I then pick one of these variables, 'V' to iterate over to produce an array of results from the function. How do I enter the structures into the function and return them within a for loop?

For example:

S.A = 10

S.B = 100

S.V = 1:1:10;

for i = 1:length(S.V)

[R(i)] = myfuction(S(i));

end

function [R] = myfunction(S)

R.C = S.A * S.V;

R.D = S.B * S.V;

end

Obviously the example above is super simple but I'm just trying to get my head around using structures within functions within loops (or another way to bundle up variables). The for loop would work for variables but how do I make it work for structures (or another form of bundling many variables)?

Kind regards,

Oliver

Answer 1

[Dave B]

In your example, you have a bunch of values for V (i.e. a 1 x 10 vector) and single values (i.e. scalars) for A and B. Do you want the function to use the the scalars for A and B and each of the values for V?

Several options here. First of all, in this limited case you don't need a loop. You're simply returning a vector C and D with corresponding length to the vector V

S.A = 10;
S.B = 100;
S.V = 1:1:10;
R = myfunction(S);

function R = myfunction(S)
R.C = S.A * S.V;
R.D = S.B * S.V;
end

A second strategy would be to make an array S, each one containing the values of A, B, and V:

for i = 1:10
   S(i).A = 10;
   S(i).B = 100;
   S(i).V = i;
end

for i = 1:numel(s)
    R(i) = myfunction(S(i));
end
% same myfunction as above

A third strategy would be to pass in which value of V you wanted:

S.A = 10
S.B = 100
S.V = 1:1:10;
for i = 1:length(S.V)
    R(i) = myfuction(S,i);
end
function [R] = myfunction(S)
R.C = S.A * S.V(i);
R.D = S.B * S.V(i);
end

Of course, there are infinitely more options depending on what the harder problems are.

A good way to think about this space maybe is to consider: a struct can be a scalar or an array; each of a struct's fields can each be a scalar or an array.

[Oliver Higbee]

Thanks Dave, this certainly gives me a few options, thanks.

To give a little background, 'S' for me contains ~30 weather and electrical parameters. I use this code to sweep these parameters depending on what parameter I'd need to study. So sometimes these varaibles will be scalars and others they will be arrays. And sometimes I need to use nested loops to sweep over two of these variables.

Option 1 - I think this is me oversimplying the example. Some parameters are raised to a power in the function etc.

Option 2 - Gives me flexibility to do this with minimal changes to the code but is resource heavy (these arrays can be be 1000+ long).

Option 3 - Too many changes needed to the function each time I change parameters to be swept. (The function contains many equations)

That is a good way to think of a structure so thanks, unfortunately with the constant changes to which variables I study, it would very time consuming to seperate the structures into scalars and arrays.

Kind regards,

Oliver

[Dave B]

Oliver -

I think the key here is that you have some knowledge about how you want to sweep across parameters but you need to pass that information to MATLAB somewhere. The case you describe of having to sweep over two of the variables in particular makes me think there's some information that you have that you need to communicate to MATLAB? (i.e. when it should it do a(i)+b(i)+c vs. a(i)+b(j)+c if that makes sense)

WRT 1: I'm sure you've considered this, but it's really important to note that lots of things in MATLAB vectorizable (including for example raising to a power), and even your nested cases may be manageable if you think about them as a matrix. Of course there's lots of logic that can be conceived in an array sense, but I mention this because vectorized operations are core to writing idiomatic MATLAB code (and often how you make things speedy).

WRT 2: worth trying to see if it matters?

WRT 3: If you wanted to treat all vectors in the same way, in theory you could write this code to cycle over vectors and index them with i and take scalars as is. I don't recommend that, it sounds messy!