This is the way I've usually seen it done. I was curious, so I ran a benchmark against the two.
use Benchmark;
my @list;
for ( 0..9999 )
{
push @list, sprintf "%d", 100 * rand ;
}
timethese(
1000,
{
'keys_map' => sub { my @uniq = keys %{{ map {$_ => 1} @list }}
+; },
'grep_seen' => sub { my %seen; my @uniq = grep ! $seen{$_}+
++, @list; },
}
);
Yields the following output.
Benchmark: timing 1000 iterations of grep_seen, keys_map...
grep_seen: 13 wallclock secs (11.17 usr + 0.00 sys = 11.17 CPU) @ 89
+.52/s (n=1000)
keys_map: 30 wallclock secs (29.28 usr + 0.00 sys = 29.28 CPU) @ 34
+.15/s (n=1000)
