Something like this should do

{
  package A;
  local *x = sub { die "aaa" };
 
  #run tests
  ...
}
# &x reset after leaving block
[download]

if you want to explicitly reset &x try to store the old coderef before.

$old_coderef = \&x;
 
# run tests
... # do what you want 
... # (but think about the omen ;-)
 
*x = $old_coderef;
[download]

It works very well. This is real perl alchemy :-)

It would probably be quicker to post the class methods x and y on here and ask why they don't work. :-)

Why doesn't function y simply log whatever it is doing through use of debug statements?

I have to disagree with you. This is a perfectly reasonable request. In fact, this is considered a good practice for testing complex software.

There are lots of reasons for doing things like this:

1. You may not have control over function y
2. It might do something that you don't want to do during testing (such as submit a credit card order...)
3. It might take a long time to run.
4. It might be hard to reproduce certain kinds of output.
5. It might not be written yet, or it might have bugs of its own (simulating part of the software allows you to develop/test multiple interdependent pieces of code simultaneously).


When's the last time you used duct tape on a duct? --Larry Wall

Thanks for the reasons. I normally suppress things I don't want to do during a test run, but I can see your points :-)
A Monk aims to give answers to those who have none, and to learn from those who know more.

Actually they work till i bump on a defect.
I test the class with a .t file, so i want to check the behavior when there is a failure.
Currently there are no known defects so i can not simulate it if i don't change the function.

EDIT: Please ignore this message, a good answer has been given above. You can alter the symbol table with *globs

The description does not really fit the title. In order to change a perl function in runtime, you would have to alter the symbol table. It can be done but it requires a lot of perl experience I think. Unless of course your function was defined as a reference to a function, in which case you can alter it as you would with any variable, which is trivial. Anyway I don't really see what you want to do.

It seems to me that what you really want is to just use 'eval':

package A;
sub Y { die "aargh" }
sub X {
   eval { Y };
   if ($@) { print "humm... Y failed..." }
   {
       print "ok, no error here";
   }
}
[download]

The title is fine; you have misunderstood the question. Take a look at Rolf's solution.



When's the last time you used duct tape on a duct? --Larry Wall

• invent a time machine
• travel back a week and a bit
• go to my talk on unit testing at the London Perl Workshop