It isn't complicated at all, and you do not need any modules beyond DB_File itself (which is part of the standard distro).

Tieing to a DB_File BTREE is no harder than tie'ing to anything else --- to keep it in memory (instead of in a file) just use undef in place of the filename argument. You can easily supply custom sort routines (default is lexical sorting).

use DB_File;
$DB_BTREE->{compare} = sub { $_[1] cmp $_[0] };
tie my %h, 'DB_File', undef, O_RDWR|O_CREAT, 0666, $DB_BTREE
    or die "DB_File tie failed: $!";
%h = 'a' .. 'z';
print "$_ : $h{$_}\n" for keys %h;
[download]

danger,
This looks potentially really cool and I would like to incorporate this technique as one of the possible options of my module. The problem is I can't seem to get DB_BTREE->{compare} to do anything meaningful. I get the same result no matter what I make the sort routine.

#!/usr/bin/perl -w
use strict;
use DB_File;

tie my %h, 'DB_File', undef, O_RDWR|O_CREAT, 0666, $DB_BTREE
   or die "DB_File tie failed: $!";

#$DB_BTREE->{compare} = sub { $h{$_[1]} <=> $h{$_[0]} };
#$DB_BTREE->{compare} = sub { $h{$_[0]} <=> $h{$_[1]} };
#$DB_BTREE->{compare} = sub { $_[0] cmp $_[1] };
#$DB_BTREE->{compare} = sub { $_[1] cmp $_[0] };


%h = (
    'John'   => 33,
    'Jacob'  => 29,
    'Jingle' => 15,
    'Heimer' => 48,
    'Smitz'  => 12,
);
print "$_ : $h{$_}\n" for keys %h;
[download]

What am I missing? - L~R

Set $DB_BTREE->{compare} before tieing the hash. However, accessing the tied hash in the compare routine leads to deep recursion (FETCH method uses the compare routine to navigate the tree).