views:

187

answers:

3

I have a Perl file like this:

use strict;
f1();

sub f3()
{ f2(); }

sub f1()
{}
sub f2()
{}

In short, f1 is called before it is defined. So, Perl throws a warning: "f1 called too early to check prototype". But same is the case with f2, the only diff being that it is called from inside another subroutine. It doesn't throw a warning for f2. Why?

What is the best way to resolve this issue?

  1. declare the subroutine before it is called
  2. call the sub like this: &f1();
+8  A: 

You can completely avoid this issue by not using prototypes in the first place:

use strict;

f1();

sub f3 { f2() }

sub f1 {}
sub f2 {}

Don't use prototypes unless you know why you are using them:

This is all very powerful, of course, and should be used only in moderation to make the world a better place.

Sinan Ünür
I don't like the "don't use prototypes" bit. They're quite nice at times, for some things.
Chris Lutz
Not using prototypes is one of the reasons people complain about readability and maintainability in Perl.
jheddings
@jheddings: You're missing the point: Perl's prototypes are, 99%, *not a useful feature*. They make the code *worse*. Don't believe me? Read http://groups.google.com/group/comp.lang.perl.modules/msg/84484de5eb01085b?dmode=source@)`.
hobbs
What is the difference between declaring the function and using
That said, *real* function signatures are a useful feature. But they don't come with Perl at all, yet. :)
hobbs
Nathan Fellman
@Chris Lutz: ah, but experienced and knowledgeable users can disregard "don't use X" cautions; they are for the unexperienced and/or unknowledgeable.
ysth
@Nathan, no, the solution is not to use the `()` prototype at all, because it's most likely there for no reason at all.
hobbs
Actually, not using prototypes also solves the problem -- without having to declare the sub before using it.
Leonardo Herrera
I agree, don't try to pretend prototypes are function signatures. If you do you (or the person maintaining your code later) will cry. They're there to force context. If you don't understand what that means DON'T USE THEM! The () prototype is relatively innocent but even that causes issues.
Schwern
+2  A: 

The lack of a warning on the call to f2() from f3() appears to be a bug.

use strict;
use warnings;

f1();

sub f1 {
    my @a = qw(a b c);
    f2(@a);
}

sub f2(\@) { print $_[0] }

This prints "a". If you either predeclare f2() or swap the order of the subroutine definitions, it prints "ARRAY(0x182c054)".

As for resolving the situation, it depends. My preferences (in order) would be:

  1. Remove the prototypes from the subroutine definitions. Perl's prototypes don't do what most people expect them to. They're really only useful for declaring subs that act like builtins. Unless you're trying to extend Perl's syntax, don't use them.
  2. Predeclare the subroutines before using them. This lets Perl know about the prototype before the encountering any calls.
  3. Reorder the code so that the subroutine definitions appear before any calls.
  4. Call the subroutines using the &foo() notation to bypass prototype checking.
Michael Carman
+3  A: 

If you are going to call it with the parenthesis, why are you even using prototypes?

sub f1(){ ... }

f1();

The only time I would use the empty prototype is for a subroutine that I want to work like a constant.

sub PI(){ 3.14159 }

print 'I like ', PI, ", don't you?\n";

I would actually recommend against using Perl 5 prototypes, unless you want your subroutine to behave differently than it would otherwise.

sub rad2deg($){ ... }

say '6.2831 radians is equal to ', rad2deg 6.2831, " degrees, of course.\n";

In this example, you would have to use parenthesis, if it didn't have a prototype. Otherwise it would have gotten an extra argument, and the last string would never get printed.

Brad Gilbert
You've got a spurious "\n" at the end of the say statement there
Mark Fowler
I meant to change it to `print` so that I didn't have to `use feature qw'say';` or `use 5.10.1;`
Brad Gilbert