DBIx::Class join pruning

DBIx::Class has a join optimiser to prune out any unnecessary joins from the queries it generates.  Occasionally this can cause code that looks good to fail unexpectedly.  The most obvious time this normally happens when using literal SQL.

The most common example of this is when doing an aggregate operation with a group by.  Here I’m trying to sum and count some values on a joined table.  This looks like pretty standard DBIC code, but it fails.  I don’t use the linked table in the group by, just in the columns where it’s my own raw SQL.

sub total
    my $self = shift;
    my @columns = $self->result_source->columns;
    my $me = $self->current_source_alias;
    return $self->search(undef, {
        join => ['items'],
        '+columns' => [
            { total => \'sum(amount) as total' },
            { lines => \'count(amount) as lines' },
        group_by => [map {"$me.$_"} @columns],

        total => { '>' => 10 }
        '+columns' => ['total', 'lines']


DBI Exception: DBD::Pg::st execute failed: ERROR:  column "amount" does not exist
LINE 1: ... (SELECT "me"."id", "me"."name", "me"."type", sum(amount) as...
                                                             ^ [for Statement "SELECT "me"."id", "me"."name", "me"."type", "me"."total", "me"."lines" FROM (SELECT "me"."id", "me"."name", "me"."type", sum(amount) as total, count(amount) as lines FROM "orders" "me" GROUP BY "me"."id", "me"."name", "me"."type") "me" WHERE ( "total" > ? )" with ParamValues: 1='10']

Looking closely at the SQL generated the items alias isn’t joined in.  The optimiser has pruned that join because it couldn’t see any reference to the column.  It turns out that DBIC does really try to do the right thing, it’s just we didn’t give it enough info to give it a hint.

Adding the ‘items.’ qualifier to the column names in the SQL allows DBIC to realise that that join is necessary.

   return $self->search(undef, {
        join => ['items'],
        '+columns' => [
            # NOTE: fully qualified name
            # even though they are not ambiguous.
            { total => \'sum(items.amount) as total' },
            { lines => \'count(items.amount) as lines' },
        group_by => [map {"$me.$_"} @columns],

With that table alias, not strictly necessary for the SQL since those column names are unambiguous, DBIC realises that it is needs the join.

SELECT "me"."id", "me"."name", "me"."type", "me"."total", "me"."lines"
  FROM (
    SELECT "me"."id", "me"."name", "me"."type", sum( items.amount ) AS total, count( items.amount ) AS lines
      FROM "orders" "me"
      LEFT JOIN "order_items" "items"
        ON "items"."order_id" = "me"."id"
    GROUP BY "me"."id", "me"."name", "me"."type"
   ) "me"
WHERE "total" > '10'

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s