I'm curious which comparison routines get optimized.
No block,
$a <=> $b,
$b <=> $a,
$a cmp $b and
$b cmp $a.
I don't see why you'd be surprised that creating a single hash would be faster than creating a ton of tiny anonymous arrays.
No lookup would even be faster. With a GRT you don't use block with sort, and hence, no array or hash lookup.
use Benchmark "cmpthese";
our @files = glob "/bin/*";
our (@o, @s, @g);
cmpthese -1 => {
ordinary => '@o = sort {-s $a <=> -s $b} @files',
st => '@s = map $_ -> [0],
sort {$a -> [1] <=> $b -> [1]}
map [$_ => -s], @files',
grt => '@g = map substr ($_, 8),
sort
map sprintf ("%08d%s", -s, $_), @files',
};
die unless "@o" eq "@s" && "@o" eq "@g";
__END__
Rate ordinary st grt
ordinary 684/s -- -54% -63%
st 1492/s 118% -- -20%
grt 1864/s 172% 25% --