A hash would not appear to be the ideal data structure to contain your transition tables. A hash contains a lot of magic behind the scenes to insure that each key maps to exactly one container in the internal structure. A precaution which your application expressly does not need. A hash also has to work over a wide population of keys and insure that the population of keys is distributed evenly over the internal storage. Your, admittedly simple, example has a fixed and small key range of a..z which could be stored sequentially.

It would seem from your example code that needs would be served by a simple array with a symbolic index. Perl supplies something called a pseudo-hash though I don't know how efficient it is. The module Hash::Util allows a similar functionality with "locked" hashes.

#!/usr/bin/perl
# this code build a table of array-refs keyed to alphabetic 
# characters.  The arrays could be used to hold state transistions.

require 5.8.8;
use strict;
use fields;
use Data::Dumper;

my $inputs = fields::phash([('a'..'z')], [ map { [1,1] } ('a'..'z')]);

$inputs->{c} = [ 0, 1 ];
$inputs->{f} = [ 2, 1 ];
$inputs->{g} = [ 0, 1 ];

print Dumper map { $inputs->{$_} } ( 'a'..'g' );
__END__
Troll [127] as maint ~/ %perl src/symarray.pl                         
+ [6:40pm]
$VAR1 = [
          1,
          1
        ];
$VAR2 = [
          1,
          1
        ];
$VAR3 = [
          0,
          1
        ];
$VAR4 = [
          1,
          1
        ];
$VAR5 = [
          1,
          1
        ];
$VAR6 = [
          2,
          1
        ];
$VAR7 = [
          0,
          1
        ];
[download]

I'm not sure if this is as efficient as I would like. Question for perlmonks; Does Perl actually create an array here or otherwise optimize or do I still suffer from the overhead of hash lookups?