Re^2: on the fly reference to subroutine output

[thirdm]

This will make a copy of the array and give a reference to that, right? Is there a way to get a reference directly to the array passed back by sub1 with some assurance the copy won't happen? Or is this the sort of optimization that is up to the compiler (and may or may not happen?)?

This example makes me think of rvalue references and move semantics in the new C++, where you can now return a std::vector, say, and be assured that the copy will really only grab the innards of the temporary returned rather than allocate anything new and copy things over. I wonder if there's something roughly equivalent in Perl (5 or 6).

[Anonymous Monk]

The incorrect assumption here is that sub1 is returning an array.

In perl5, subs can return scalars or lists. Since it's not returning a (scalar) reference to an array, the only option is to store the list into a (possibly anonymous) array and pass a reference into sub2

tjd

[thirdm]

Ah, I'm always forgetting this.

But underneath, is there a chance at some kind of return value optimization? Say if you have...

sub1 {
...
   return @a;
}

$r = [sub1];

Even if it's officially turning the values of @a into a list en route to copying them into the array that the brackets create a reference to, could perl not notice @a is no longer needed otherwise and instead simply make [sub1] be \@a behind the scenes? It would be neat if you could write sub1 that way instead of making it return \@a even if you anticipated the copy to be expensive.