That's what I expect from common sense, yes.
A quick benchmark seems to confirm this; asking an empty hash is much faster than asking a filled one, but the number of keys in the full hash seems to have no significant influence on the speed:
use strict;
use warnings;
use Benchmark qw(cmpthese);
for my $len (2..6) {
our (%full, %empty);
my $a = 'A';
$full{$a++} = $a while length($a) < $len;
print "\nLength $len with ", scalar(%full), "items\n";
cmpthese -1, {
full => sub {die unless %full},
empty => sub {die if %empty},
}
}
__END__
Length 2 with 22/32items
Rate full empty
full 1286220/s -- -77%
empty 5481192/s 326% --
Length 3 with 510/1024items
Rate full empty
full 1052183/s -- -82%
empty 5748771/s 446% --
Length 4 with 13961/32768items
Rate full empty
full 998734/s -- -84%
empty 6056132/s 506% --
Length 5 with 310721/524288items
Rate full empty
full 963764/s -- -83%
empty 5681139/s 489% --
Length 6 with 8655839/16777216items
Rate full empty
full 849541/s -- -86%
empty 6124859/s 621% --
Note that the number of hash items grows exponentially, while the number of iterations decreases linearly at best.