Find the value which causes an error from a vector

Hello,

I do have the following code:

function [test] = func(T)
T_grenz=112;
if T>=T_grenz
    svol = 50*T;
elseif T<T_grenz
    svol = 100*T;
end
if svol==0
    svol = 0.001;
elseif isemtpy(svol)
    error("svol is empty at T="); %I want to add this error message correctly
end
test=1/svol;

This matlab script is used by a simulation software. The software calls the matlab function with like a 1x10000 vector and expects a 1x10000 vector as return value.

While running this for some reason matlab gets the error that svol is not defined. However I think in every case svol should have a value. Now I want to reproduce the error. How can I get exact values of my vector T to reproduce the bug. I do not want the whole vector. Just the one value that cause the error. Can I do this with vectors? Or do I have to write the for loop manually to iterate through the vector?

Answer 1

[Stephan]

This works with logical indexing, therefore avoids the if / else logic and should ensure that you do not need any error messages:

function test = func(T)
    T_grenz=112;
    [r,c] = size(T);
    svol = zeros(r,c);
    svol(T>=T_grenz) = 50*T(T>T_grenz);
    svol(T<T_grenz) = 100*T(T<T_grenz);
    svol(svol==0) = 0.001;
    test=1./svol;
end

[Christian Berwanger]

I tried this solution. My testparameters work with this again. However. The software which uses my matlab script gets the error: The logical indices contain a true value outside of the array bounds. which seems to be a matlab error, which is just parsed to the software I use.

So I would still like to see which values or vectors are causing the issue, which is not possible right now. Any suggestion on how I can get the error causing values (or in this case I guess the whole vectors are causing the error)?

[Stephan]

The result of this function is an array of the same size like the input - i think the problem is not inside this function or its output. What kind of software do you use?

• The way you use isempty can not work. isempty gives a logic 1 only on empty arrays.
• You want to select values depending on the value of T, but then you make operations on the whole svol vector - this will not work.

The best way is to use the debugger to learn what goes wrong:

[Star Strider]

Typographical error:

svol(T>=T_grenz) = 50*T(T>T_grenz);
                         ↑ ← HERE

Should be:

svol(T>=T_grenz) = 50*T(T>=T_grenz);

.

[Christian Berwanger]

Ok. One of my problems is that I wanted to deliver a minimal working example of the error and I might have oversimplified it.

svol is actually dependend on a lot of parameters which are defined before. So my function looks roughly more like:

function test = func(T,dT,p)
    q=abs(dT);
    if q==0:
        q=0.1;
    end
    A = 100+20.*log(q)-0.01*p
    Bm = 0.58-0.0003*p+0.00003*p*p
    Bs = 0.58-0.0001*p+0.000019*p*p
    C = 100.*exp(-0.0023*p)
    T_grenz=100+0.1.*log(q)-0.02.*p;
    [r,c] = size(T);
    svol = zeros(r,c);
    svol(T>=T_grenz) = A(T>=T_grenz)+Bm(T>=T_grenz).*T(T>=T_grenz)+0;
    svol(T<T_grenz) = A(T<T_grenz)+Bs(T<T_grenz).*T(T<T_grenz)+C(T<T_grenz).*p(T<T_grenz);
    svol(svol==0) = 0.001;
    test=1./svol;
end

I hope that I did that right.

I use Comsol Multiphysics for my simulation. In Comsol the matlab function is called with roughly 10.000x1 vectors as input parameters and expects a 10.000x1 output vector.

Now. When I create a small vector for myself to test the matlab function I get no errors and I get a result. However. When using Comsol I get an error. Thats why I want to get the error causing vector (or input parameter) which matlab gets from comsol.

[Christian Berwanger]

And I use the debugger. But nothing goes wrong in my test. Comsol uses the matlab server, where breakpoints doesn't work. Thats why I want to get the error causing data from comsol, so I can run them with the debugger