Problem

You want to process a large dataset you'd like to commit to a DBM file in a particular order.

Solution

Use the DB_File's B-tree bindings and supply a comparison function of your own devising:

use DB_File;

# specify the Perl sub to do key comparison using the
# exported $DB_BTREE hash reference
$DB_BTREE->{'compare'} = sub {
    my ($key1, $key2) = @_ ;
    "\L$key1" cmp "\L$key2" ;
};

tie(%hash, "DB_File", $filename, O_RDWR|O_CREAT, 0666, $DB_BTREE)
    or die "can't tie $filename: $!";

Description

An annoyance of hashes, whether in memory or as DBM files, is that they do not maintain proper ordering. The CPAN module Tie::IxHash can make a regular hash in memory maintain its insertion order, but that doesn't help you for DBM databases or arbitrary sorting criteria.

The DB_File module supports a nice solution to this using a B-tree implementation. One advantage of a B-tree over a regular DBM hash is its ordering. When the user defines a comparison function, all calls to keys, values, and each are automatically ordered. For example, Example 14.4 is a program that maintains a hash whose keys will always be sorted case-insensitively.

#!/usr/bin/perl
# sortdemo - show auto dbm sorting
use strict;
use DB_File;

$DB_BTREE->{'compare'} = sub {
    my ($key1, $key2) = @_ ;
    "\L$key1" cmp "\L$key2" ;
};

my %hash;
my $filename = '/tmp/sorthash.db';
tie(%hash, "DB_File", $filename, O_RDWR|O_CREAT, 0666, $DB_BTREE)
    or die "can't tie $filename: $!";

my $i = 0;
for my $word (qw(Can't you go camp down by Gibraltar)) {
    $hash{$word} = ++$i;
}

while (my($word, $number) = each %hash) {
    printf "%-12s %d\n", $word, $number;
}

By default, the entries in a B-tree DB_File database are stored alphabetically. Here, though, we provide a case-insensitive comparison function, so using each to fetch all the keys would show:

by           6
camp         4
Can't        1
down         5
Gibraltar    7
go           3
you          2

This sorting property on hashes is so convenient that it's worth using even without a permanent database. If you pass undef where the filename is expected on the tie, DB_File will create a file in /tmp and then immediately unlink it, giving an anonymous database:

tie(%hash, "DB_File", undef, O_RDWR|O_CREAT, 0666, $DB_BTREE)
        or die "can't tie: $!";

Remember these two things if you supply a comparison for your BTREE database. One, the new compare function must be specified when you create the database. Two, you cannot change the ordering once the database has been created; you must use the same compare function every time you access the database.

Using BTREE databases under DB_File also permits duplicate or partial keys. See its documentation for examples.

See Also

Recipe 5.6