That seems like it should be possible. Probably not much fun to whip up an example without even a starting point though. :) Can you post the code you've tried? Using SQLite as a the DB is the easiest way to make it portable for others to test/hack.

There are ways to have your class files autogenerated too--DBIx::Class::Schema::Loader--or just have the classes in memory. So maybe you should list your requirements before picking the code path to follow.

package DB::User;
use base qw(DBIx::Class);
__PACKAGE__->load_components(qw(PK::Auto Core));
__PACKAGE__->table('user');
__PACKAGE__->add_columns(qw(id login password));
__PACKAGE__->set_primary_key('id');

package DB;
use base qw(DBIx::Class::Schema);
__PACKAGE__->load_classes(qw(User));

package main;
use strict;
use Data::Dumper;

my $schema = DB->connect( 
    'DBI:mysql:database:localhost',
    'username', 'password', {RaiseError => 1}
);

my @users = $schema->resultset('User')->all;
print Dumper($users[1]{_column_data});
[download]

Resurrecting this thread, because I have also been struggling with this of late, and I thought it would be a good idea to post what I have learnt.

I found that to make things work, I needed to put the table classes before the main schema class, in the source file, and that I can't have any extra levels in the class hierarchy between the Schema class and the table classes. (Not sure why)

Here is an example script that I got working

#!/usr/bin/perl

# Test/Demo of a single file perl script using DBIx::Class.

# Notes on how to make this work:
#   The Table class(es) need to appear before the Schema class or
#       the rest of the program.
#   The package of the table classes can't have any extra levels
#       (eg <schema>::Result::<table> won't work.
#   Remember the deploy command if the initial DB is empty.

# Put the DBIx::Class packages in the same file
package LocalTest::Schema::TestTable;
use base 'DBIx::Class::Core';
use DateTime;
__PACKAGE__->table('tblTest');
__PACKAGE__->load_components(qw/InflateColumn::DateTime/);
__PACKAGE__->add_columns(
    'id'   => { data_type=>'int',         is_auto_increment=>1 },
    'date' => { data_type=>'datetime'                          },
    'text' => { data_type=>'varchar', size=>20, is_nullable=>1 },
);
__PACKAGE__->set_primary_key('id');

sub new
{
    my( $class, $arg ) = @_;
    $arg->{date}  ||= DateTime->now;
    return $class->next::method( $arg );
}

package LocalTest::Schema;
use base 'DBIx::Class::Schema';
__PACKAGE__->load_namespaces;
__PACKAGE__->load_classes('TestTable');

# Back to the main program
package main;

use strict;
use warnings;
use diagnostics;
use English;
use DateTime;

# use LocalTest::Schema;    # This import does not work, & not needed.

my $schema = LocalTest::Schema->connect('dbi:SQLite:dbname=:memory:');
$schema->deploy();

my $row1 = $schema->resultset('TestTable')->create({'text'=>'foo_bar'}
+);
my $row2 = $schema->resultset('TestTable')->create({});
my $row3 = $schema->resultset('TestTable')->create({'text'=>'frodo'});

my $row_rs = $schema->resultset('TestTable')->search({'id'=>[1,2]});
my @rows = $row_rs->all;

printf "Found %d rows\n", $row_rs->count;
[download]

A trick you can use to make the human read code flow more like you expect-

# package stuff I expect/want to see first.

BEGIN {
# package stuff the compiler needs to see first.
}

package main;

# my stuff that does stuff and junk.

exit 0;

# subs, if any

__END__

=pod

=cut

leftover notes...
[download]