I was looping through each subsitution which is inefficient!

Not necessarily. Alternation can disable some optimizations that the regex engine uses. This becomes more noticable on larger strings.

use strict;
use warnings;
use Benchmark qw(cmpthese timethese);

my $orig = "Test: FileNames can't have / \\ and other special characte
+rs";
$orig .= 'a' x 1000;


my %substs = qw(
    :       COLON
    /       FSLASH
    \\\\    BSLASH
);
sub loop {
    my $string = $orig;
    $string =~ s/$_/$substs{$_}/g
        for keys %substs;
    return;
}


my @substs = (
    qr/:/  => 'COLON',
    qr/\// => 'FSLASH',
    qr/\\/ => 'BSLASH',
);
sub loop2 {
    my $string = $orig;
    $string =~ s/$substs[$_]/$substs[$_ + 1]/g
        for map $_ * 2, 0 .. $#substs / 2;
    return;
}


my %s = qw(
    :   COLON
    /   FSLASH
    \\  BSLASH
);
my $t = join "|", map quotemeta, keys %s;
sub alter {
    my $string = $orig;
    $string =~ s/($t)/$s{$1}/g;
    return;
}


my %tests = (
    loop => \&loop,
    loop2 => \&loop2,
    alter => \&alter,
);

cmpthese(timethese(-2, \%tests));

__END__
Benchmark: running alter, loop, loop2 for at least 2 CPU seconds...
     alter:  3 wallclock secs ( 2.42 usr +  0.00 sys =  2.42 CPU) @ 16
+55.39/s (n=4011)
      loop:  2 wallclock secs ( 2.13 usr +  0.00 sys =  2.13 CPU) @ 64
+89.92/s (n=13843)
     loop2:  2 wallclock secs ( 2.02 usr +  0.00 sys =  2.02 CPU) @ 11
+480.97/s (n=23226)
         Rate alter  loop loop2
alter  1698/s    --  -72%  -86%
loop   6125/s  261%    --  -48%
loop2 11811/s  596%   93%    --
[download]

Note that the code isn't completely equivalent though. Using &loop or &loop2 you can replace a replacement. For instance, if a replacement has a colon in it &alter will leave it there while &loop will change it to "COLON".

ihb