Strange behavior of anonymous functions

Dear all,

I have two problems related to anonymous functions:

1-) I create an anonymous function taking several inputs inside a m-file function. Would like the anonymous function to take one input and use the other pre-created inputs from the workspace. for instance,

function f=myfunc(...) 

a=rand(10,1); b=rand(10,1); c=rand(10,1); ss=rand(10,1);
f=@(x)a(6,:)*x^2+b(4,:)*x+c(3,:)-ss(5)

I get an error when calling the function I get an error saying that "ss" is undefined for inputs of type double and that I many need some toolbox that I do not have. But then I change the second line in the function to

f=@(x)a(6,:)*x^2+b(4,:)*x+c(3,:)-ss(5,:)

and everything works. Has anyone encountered such a problem? Is there any explanation for that behavior?

2-) After creating the anonymous function, I put it in a structure that I would like to carry around. The structure contains the anonymous function and various other elements. mystruct=struct(...,'anonym',myfunc(...)); I now would like to update the anonymous function inside the structure. In particular, I would like to change the order of input "x" before passing it to the anonymous function. I thought I could just create another anonymous function as follows neworder=10:-1:1; mystruct.anonym=@(x)mystruct.anonym(x(neworder));

This works without any problem. The issue is that when I now inspect the workspace of "mystruct.anonym" by issuing w=functions(mystruct.anonym), w.workspace, it contains an almost interminable loop of mystruct.anonym woven inside the sub-workspaces. This, most certainly, uses some memory. Is there any way to get rid of this behavior?

Thanks,

[Adam]

Your first example works fine for me, although why would you use the (3,:) syntax for a, b and c, but not for s (I wouldn't use it for any of them in that example).

[Patrick Mboma]

this is just a piece of code to illustrate the issue. the actual code is a lot more complicated.

[Brendan Hamm]

Also runs perfectly fine for me. I would check the original code and see if maybe you have a typo. Something like:

s = rand(10,1);

f=@(x)a(6,:)*x^2+b(4,:)*x+c(3,:)-ss(5)

Answer 1

[per isakson]

1. If you have the Control or Signal Toolbox, ss as the name for variables used to cause trouble. See ss,Create state-space model, convert to state-space model. I once spent time to learn that:-(
2. TMW is working on it. Not much we can do. Try to keep the workspace small when creating the anonymous function.

[Patrick Mboma]

Hi Per,

I change the variable name and it still does not work. It very much looks to me like a bug in the version of matlab I am using (2014b) but I am not 100 percent sure. I will try and install a different version and see whether something changes.

To summarize the problem, ss (or any other name I choose) is created before the anonymous function. But it just does not appear in the workspace of the anonymous function unless I call it as ss(2,:).

[Patrick Mboma]

I have tried a different version of matlab and I get the same results. Although the variable exists in the workspace, somehow matlab must interpret it as a function, I guess...