You might want to look at Tie::DBI (by Lincoln Stein) or AnyData (by me). AnyData is a tied-hash interface to DBI (it underlies DBD::AnyData but has a separate user interface). Here's some sample AnyData code:

# create a table
#
$table = adTie( 'CSV','my_db.csv','o',
                 {col_names=>'name,country,sex'}
               );

# insert a row
#
$table->{Sue} = {country=>'de',sex=>'f'};

# delete a single row
#
delete $table->{Tom};

# select a single value
#
$str  = $table->{Sue}->{country};

# loop through table
#
while ( my $row = each %$table ) {
   print $row->{name} if $row->{sex} eq 'f';
}

# select multiple rows
#
$rows = $table->{{age=>'> 25'}}


# update multiple rows
#
$table->{{country=>'Nz'}}={country=>'nz'};
[download]

I really recommend that you write to dbi-dev@perl.org to get ideas from Tim and other DBI/DBD authors.