[Dbix-class] is_foreign_key_constraint => false is ineffective

[Ash Berlin]

Jon Schutz wrote:
> I've added is_foreign_key_constraint => 0 as an attribute in a
> relationship, in order to query across the join without imposing
> database level constraints - but the constraint is still appearing in
> the resulting schema.
> 
> The doco in Relationship::Base says "set this attribute to a true or
> false value to override the detection of when to create constraints."
> 
> Can anyone (particular Jess & Matt whose

Except I wrote that bit of code ;)

 names can be found in the
> source) explain the logic in the following bit of code to me?  From
> SQL::Translator::Parser::DBIx::Class.pm :
> 
>                # Make sure we dont create the same foreign key constraint twice
>                 # Decide if this is a foreign key based on whether the self
>                 # items are our primary columns.
>                 # If the sets are different, then we assume it's a foreign key from
>                 # us to another table.
>                 # OR: If is_foreign_key_constraint attr is explicity set (or set to false) on the relation
>                 if ( ! exists $created_FK_rels{$rel_table}->{$key_test} &&
>                      ( exists $rel_info->{attrs}{is_foreign_key_constraint} && 
>                        $rel_info->{attrs}{is_foreign_key_constraint} ||
>                        !$source->compare_relationship_keys(\@keys, \@primary)
>                      )
>                    )
>                 {
>                     $created_FK_rels{$rel_table}->{$key_test} = 1;
>                     $table->add_constraint(
>                                ...
>                     );
>                 }
> 
> I read it as - if the foreign key has already been created, don't try to
> recreate the same one.  If is_foreign_key_constraint has been specified
> and is set, or if the keys are not primary keys, then create the foreign
> key relation.  So it seems we can force foreign keys with
> is_foreign_key_constraint = true, but can't forcibly disable them as the
> compare_relationship_keys test wins in the end.
> 
> Shouldn't that logic be : if is_foreign_key_constraint has been
> specified then do whatever it says, otherwise test using
> compare_relationship_keys ? i.e.
> 
>                if ( ! exists $created_FK_rels{$rel_table}->{$key_test} &&
> 		     ( exists $rel_info->{attrs}{is_foreign_key_constraint} ? 
> 			$rel_info->{attrs}{is_foreign_key_constraint} :
> 			$source->compare_relationship_keys(\@keys, \@primary) 
> 		     ) 
> 		   ) ...
>  
> ?
> 

Yes good point. Of course I only ever wanted it to force creating of
keys, not stop it. Fancy putting together some test cases for this? (By
creating a patch against t/86sqlt.t and other files?

-ash