Is there any other way than try & catch to handle error in matlab?

I have a code that is executed to give a result. But it is possible that it would not execute due to some alteration done to the code before execution. How do I control the error that is expected if the code is not running. For example:

    result = compute(x, y)

The value of result can be a real number or the compute function does not complete the execution. How do I assign a particular value (say -1) to result if the function fails to execute.

Answer 1

[Walter Roberson]

Do not have your function call

result = compute(x, y)

instead have it call

clear compute
result = test_function(@compute, substitute_value, x, y);

where

function output = test_function(fh, substitute_value, varargin)
   try
     output = feval(fh, varargin{:});
   catch
     output = substitute_value;
   end

Now, as long as the function does not exit MATLAB or crash MATLAB, any error will be caught.

As long as only the function compute changes, and not the function test_function, then you should be fine.

Note: the above code does assume that compute() is not a function handle.

Answer 2

[Jan]

You cannot predict reliably, if the function fails or not without running it. Such a test is impossible even in theory (search for "Turing halting problem").

You can check if the inputs cause problems inside the function also:

function r = compute(x, y)
if y == 0
  r = NaN;
else
  r = x / y;
end

But a try catch block is the perfect solution for catching errors, if you cannot modify the function accordingly. Why do you want to avoid it?

[Hayatullahi Adeyemo]

Thanks for your contribution. Actually, I am working on mutation testing whereby a function is altered and a test is generated to see if the result of the altered function would be different from the result of the original function. If the results are different, then the alteration is significant else otherwise. One of the alterations is statement deletion. So even if I include try-catch block, one of the components of the try-catch block could be the statement to be deleted as a result of random deletion that is applied; which would result in the error we are trying to avoid. This is why I need a way to detect if the code gives error or it is executed successfully without changing anything in the original/altered code. I believe the intention is clear now. If it is not clear, pls let me know. Thanks!

[Jan]

The posted link is dead.

Do you modify M-code dynamically? And dare to run it? Bold.

quit = 17;
plot(quit, 3);

Now remove the first line and run the code again. Even try/catch in the caller will not help. Calling the code in a sandbox, e.g. as an additional Matlab engine, cannot solve problems like:

s = '*.*';
s = 'C:\Temp\file.mat';   % <- removed randomly
delete(s);

My impression is, that the problems are much more serious than a removed try.

[Walter Roberson]

It appears the link is https://en.wikipedia.org/wiki/Mutation_testing

[Hayatullahi Adeyemo]

Yes, you are right. I have modified it.