Single struct as alternative to name-value pairs in user-defined function

Hello, is it possible when defining a function with name-value argument pairs to allow a user to pass the struct(s) that define these name-value pairs directly to the function?

For example, I can define

function [s1, s2] = TestStructArgs(s1, s2)
arguments
    s1 (1,1) struct
    s2 (1,1) struct
end
end

then call TestStructArgs(struct('a', 1, 'b', 2), struct('x', 3, 'y', 4))

Or, I can define

function [s1, s2] = TestStructArgs2(s1, s2)
arguments
    s1.a (1,1) double
    s1.b (1,1) double
    s2.x (1,1) double
    s2.y (1,1) double
end

end

and call TestStructArgs2('a', 1, 'b', 2, 'x', 3, 'y', 4)

But what if I would also like to call the second example by passing whole structs, TestStructArgs2(struct('a', 1, 'b', 2), struct('x', 3, 'y', 4)) ? There doesn't appear to be any ambiguity in allowing both approaches, if they're not mixed-and-matched.

Further, some MATLAB built-in functions permit this e.g. the following runs just fine

>> opts = struct('Color', 'r', 'LineWidth', 2);
>> x = linspace(0,1,200); y = cos(x);
>> plot(x, y, opts)

There is an answer on the messageboard from 2022 here which suggests adding the attribute StructExpand after arguments but this causes an error if I add it to the above example

Many thanks for any help

Answer 1

[Matt J]

You would have to use the older inputParser system to do that kind of thing, with the StructExpand property set to true (the default), e.g.,

TestStructArgs2(struct('a', 1,'b',2), struct('x', 3, 'y', 4))

s1 = struct with fields:
    a: 1
    b: 2
    c: 30

s2 = struct with fields:
    x: 3
    y: 4
    z: 60

function TestStructArgs2(varargin)
 p=inputParser;
 addParameter(p,'a',10,@isscalar);
 addParameter(p,'b',20,@isscalar);
 addParameter(p,'c',30,@isscalar);
 addParameter(p,'x',40,@isscalar);
 addParameter(p,'y',50,@isscalar);
 addParameter(p,'z',60,@isscalar);
 parse(p,varargin{:});
 p=p.Results;

 s1=struct('a',p.a,'b',p.b,'c',p.c)
 s2=struct('x',p.x,'y',p.y,'z',p.z)

end

[Jake]

Thanks Matt, that's really helpful. I feared I might have to fall back to inputParser and was intimidated by the amount of code I thought might be required, but yours is a nice, concise solution. It's a shame there's no way to achieve the same effect with argument blocks, however

[Matt J]

Yes, I agree. It's been suggested to TMW before, but I guess there are problems with the idea that we're not seeing.