NAME

DBIx::Class::SQLMaker::Role::SQLA2Passthrough - A test of future possibilities

REQUIRES

This module ships inside the SQL-Abstract distribution but requires DBIx::Class at runtime. DBIx::Class is intentionally not a prerequisite of SQL-Abstract - DBIx::Class depends on SQL::Abstract, so declaring the reverse would create a dependency cycle - therefore you must have DBIx::Class installed separately to use this module. It is an experimental proof of concept (see the NAME above).

SYNOPSIS

• select and group_by options are processed using the richer SQLA2 code

• expand_join_condition is provided to more easily express rich joins

See examples/sqla2passthrough.pl for a small amount of running code.

SETUP

(on_connect_call => sub {
   my ($storage) = @_;
   $storage->sql_maker
           ->with::roles('DBIx::Class::SQLMaker::Role::SQLA2Passthrough');
})

expand_join_condition

__PACKAGE__->has_many(minions => 'Blah::Person' => sub {
  my ($args) = @_;
  $args->{self_resultsource}
       ->schema->storage->sql_maker
       ->expand_join_condition(
           $args
         );
});

on

__PACKAGE__->has_many(minions => 'Blah::Person' => on {
  { 'self.group_id' => 'foreign.group_id',
    'self.rank' => { '>', 'foreign.rank' } }
});

Or with ParameterizedJoinHack,

__PACKAGE__->parameterized_has_many(
    priority_tasks => 'MySchema::Result::Task',
    [['min_priority'] => sub {
        my $args = shift;
        return +{
            "$args->{foreign_alias}.owner_id" => {
                -ident => "$args->{self_alias}.id",
            },
            "$args->{foreign_alias}.priority" => {
                '>=' => $_{min_priority},
            },
        };
    }],
);

becomes

__PACKAGE__->parameterized_has_many(
    priority_tasks => 'MySchema::Result::Task',
    [['min_priority'] => on {
      { 'foreign.owner_id' => 'self.id',
        'foreign.priority' => { '>=', { -value => $_{min_priority} } } }
    }]
);

Note that foreign/self can appear in such a condition on either side, BUT if you want DBIx::Class to be able to use a join-less version you must ensure that the LHS is all foreign columns, i.e.

on {
  +{
    'foreign.x' => 'self.x',
    'self.y' => { -between => [ 'foreign.y1', 'foreign.y2' ] }
  }
}

is completely valid but DBIC will insist on doing a JOIN even if you have a fully populated row object to call search_related on - to avoid the spurious JOIN, you must specify it with explicit LHS foreign cols as:

on {
  +{
    'foreign.x' => 'self.x',
    'foreign.y1' => { '<=', 'self.y' },
    'foreign.y2' => { '>=', 'self.y' },
  }
}