How to redirect function output arguments to a cell array

I have a third-party function (one I can't edit) that takes an arbitrary-size array of input values, but then returns one argument for each input value (rather than returning an array of output values, which would seem to make a lot more sense). So I need to compile this set of variables of unknown length into a single cell array. Is there a clean way to do this? All I've come up with is to build a wrapper that looks at the input array size and then produces a string like '[Y{1},Y{2},Y{3}]=sillyFunction(X)' and then passes that to eval(), but generally when I have to resort to using eval() I'm doing something wrong.

[Stephen23]

"...generally when I have to resort to using eval() I'm doing something wrong."

Yes.

https://www.mathworks.com/matlabcentral/discussions/tips/847976-tutorial-comma-separated-lists-and-how-to-use-them

Answer 1

[Steven Lord]

Use a comma-separated list. On that page there is an item in the "How to Use the Comma-Separated Lists" section that shows an example using them for collecting output arguments from a function.

numOut = 2;
C = cell(1, numOut);
x = randi([-10 10], 1, 10)

x = 1×10

2 5 2 -5 8 3 1 9 -8 -4

[C{:}] = max(x)

C = 1×2 cell array
    {[9]}    {[8]}

[value, location] = max(x)

value = 9

location = 8

The first cell in C filled in by the first max call contains the same information as the value output in the second call. Similarly the second cell of C matches the location output in the second call.

The only tricky thing is determining numOut, but from your description it sounds like you can compute that if you know the size or numel of the input that will be passed to the function you have to treat as a black box.

You can also use comma-separated lists for input.

Q = [C{:}] % concatenation

Q = 1×2

9 8

S = plus(C{:})

S = 17

[Lane]

This is what I needed--thanks.

Answer 2

[Sulaymon Eshkabilov]

One of the easy and quick solution to your exercise is alike the following:

OUT = sillyFunction(X); % The output is a cell array

function Y=sillyFunction(X)

Y{1}= ...;
Y{2}=...;
Y{3}=...;
...

end

[Lane]

It seems like you are proposing a modification of sillyFunction to produce a cell array. Like I said in my original post, though, I can't modify sillyFunction. I have to work with it in its original form, where it produces one output argument for every input element. Am I misunderstanding you?

Answer 3

[Sulaymon Eshkabilov]

If this is the case, then you'd need to convert the outputs into cell array after acquiring the outputs from the function - sillyFunction(X), e.g.:

for ii=1:numel(X)
OUT{ii}=sillyFunction(X(ii));
end

Answer 4

[Mauricio]

Based on previous answers, I created a simple generic function that captures the n-th output from any function func.

So I'm just putting it here since it could be useful for someone else:

function o = get_n_output(n,func,varargin)
% n -> position of the output to capture
% func -> function to call
% varargin -> ordered arguments for function func
%
% returns:
%    n-th output of func(varargin{:})
    output      = cell(1,n);
    [output{:}] = func(varargin{:});
    o           = output{n};
end

For example, if you have a function

function [a,b,c] = do_something(arg1,arg2,arg3)
% do something, e.g.
    a = arg1;
    b = arg2;
    c = arg3;
end

And you want to get only the 2nd output, you can call

b = get_n_output(2,@do_something,1,2,3); % returns b=2