Whene there are so many WTDI, a Benchmark may be appropriate to chose:

my $olddate = '20020131';

MyBenchmark::benchmark(
        $ARGV[0],[
        ['regexp \d{}' => sub {
                my $new;
                $olddate =~ m/^(\d{4})(\d{2})(\d{2})$/ and
                $new = "$1-$2-$3";
                $new
        }],
        ['regexp ..' => sub {
                my $new;
                my @p = $olddate =~ /^(....)(..)(..)$/;
                $new = join '-',@p;
                $new
        }],
        ['s/// \d{}' => sub {
                my $new = $olddate;
                $new =~ s/^(\d{4})(\d{2})(\d{2})$/$1-$2-$3/;
                $new
        }],
        ['substr'  => sub {
                my $new = $olddate;
                substr (((substr ($new,4,0) = '-'),$new),7,0)='-';
                $new
        }],
        ['join unpack' => sub {
                my $new;
                $new = join '-', unpack 'a4 a2 a2', $olddate;
                $new
        }],
        ['sprintf unpack' => sub {
                my $new;
                $new = sprintf '%s-%s-%s', unpack 'a4 a2 a2', $olddate
+;
                $new
        }],
        ], '2002-01-31' #expected output
);
[download]

Note that MyBenchmark::benchmark is just a function i've written that executes all subs beforehand and compares their results with an expected value. The benchmark itself is done via Benchmark::cmpthese.

join unpack is fastest for me, sprintf unpack is next.

--
http://fruiture.de