Do you even need prototypes?

use 5.010;

sub size { @_ + 0 }

say size 'a' .. 'z';

my @foo = (1, 2, 3);
say size @foo;
[download]

hmm good point...

you are forcing everything into an array,

maybe

sub size { 
  scalar @_ 
}
[download]

is easier to read.

it's like explicitly wrapping lists and arrays into scalar @{[ ... ]}

But w/o prototypes this would mean to copy huge arrays around...

And something like size %hash couldn't be extended to find scalar keys %hash.

Anyway your example fits as a proof of concept! =)

(But an efficient implementation should be done in XS. Edit: see benchmark)

Cheers Rolf

Inline::C allows a pretty fast implementation...

use 5.010;
use Benchmark qw(cmpthese);
use Inline C => <<'INLINE';
int count1 (SV* name1, ...)
{
   Inline_Stack_Vars;
   return Inline_Stack_Items;
}
INLINE

sub count2 { scalar @_ }
sub count3 { 0 + @_ }
sub count4 { my $_ = @_ }
sub count5 (\@) { scalar @{$_[0]} }

my @array = 'a'..'z';

my %implementations = (
    count1 => sub { count1(@array) },
    count2 => sub { count2(@array) },
    count3 => sub { count3(@array) },
    count4 => sub { count4(@array) },
    count5 => sub { count5(@array) },
);

for my $i (sort keys %implementations) {
    say $i, ": ", $implementations{$i}->();
}

cmpthese(250_000, \%implementations);
[download]

XS may well allow faster still.

perl -E'sub Monkey::do{say$_,for@_,do{($monkey=[caller(0)]->[3])=~s{::}{ }and$monkey}}"Monkey say"->Monkey::do'

LanX:

With the scalar operator and arrays evaluating to their length in scalar contect, I don't see a need for it. But if someone decided to change perl to do something like this, I'd rather see them overload length to accept a hash, array or reference to a hash or array.

...roboticus

When your only tool is a hammer, all problems look like your thumb.

> I don't see a need for it.

Honestly, did you immediately spot the problem in BUK's code?

Re: Idiom to return 0 or random number of array elements

UPDATE:

Do you really prefer explaining goatse "op"  =()= and turtle "op" @{[ ]} to newbes who simply need to count the elements a function returns?

UPDATE:

> I'd rather see them overload length to accept a hash, array or reference to a hash or array.

Then it's quite impossible to tell what length ($str) is supposed to return. 1 for the list-size or the char-size of $str?

Cheers Rolf

LanX:

No, I didn't notice the error. And I've never used goatse, turtle or kite. Perhaps if I had, I'd see the need for it.

As far as the practicality of what length $str is supposed to return, I didn't give it much thought. I was just expressing my color preference for new new shed out back. ;^)

...roboticus

When your only tool is a hammer, all problems look like your thumb.

size is a terrible name, because it doesn't imply any unit. It could stand for string length in characters, string length in bytes, used memory, number of allocated elements in a container, number of actual elements in a container or any other size you can think of.

It makes sense to you now because it's not too bad in the mental context you are in, but from the outside it's not good.

Anyway, if you don't fiddle with prototypes at all, the argument list of a subroutine is automatically in list context, so you can just write

sub elems { scalar @_ }
[download]

and don't need a new keyword at all.

I care less about the name than the functionality, could be also "count" or whatever suits.

I suppose "elems" is already known from ruby or perl6? So why not.¹

But² tobyinks and your suggestion w/o prototypes is considerably slower on large arrays.

  DB<106> use Time::HiRes qw(time)
 
  DB<107> sub elems { scalar @_ }
 
  DB<108> @a=(1..1e6);0
 
  DB<109> $t=time; $count = scalar @a; print time-$t
7.10487365722656e-05
  DB<110> $t=time; $count = elems @a; print time-$t
0.0554749965667725
[download]

Perl 5.10!

Cheers Rolf

¹) Well the plural in "elems" somehow indicates a list to be returned.

²) as already said

It is most probably the copying of the argument list (the full array) in the function call that takes this time difference.

Passing a reference to the array would probably solve the issue, but the syntax would become less obvious to use.