1 package DBIx::Class::Relationship::Base;
6 use base qw/DBIx::Class/;
8 use Scalar::Util qw/weaken blessed/;
14 DBIx::Class::Relationship::Base - Inter-table relationships
18 __PACKAGE__->add_relationship(
19 spiders => 'My::DB::Result::Creatures',
23 "$args->{foreign_alias}.id" => { -ident => "$args->{self_alias}.id" },
24 "$args->{foreign_alias}.type" => 'arachnid'
31 This class provides methods to describe the relationships between the
32 tables in your database model. These are the "bare bones" relationships
33 methods, for predefined ones, look in L<DBIx::Class::Relationship>.
37 =head2 add_relationship
41 =item Arguments: 'relname', 'Foreign::Class', $condition, $attrs
45 __PACKAGE__->add_relationship('relname',
49 Create a custom relationship between one result source and another
50 source, indicated by its class name.
54 The condition argument describes the C<ON> clause of the C<JOIN>
55 expression used to connect the two sources when creating SQL queries.
57 To create simple equality joins, supply a hashref containing the
58 remote table column name as the key(s), and the local table column
59 name as the value(s), for example given:
61 My::Schema::Author->has_many(
62 books => 'My::Schema::Book',
63 { 'foreign.author_id' => 'self.id' }
68 $author_rs->search_related('books')->next
70 will result in the following C<JOIN> clause:
72 ... FROM author me LEFT JOIN book books ON books.author_id = me.id ...
74 This describes a relationship between the C<Author> table and the
75 C<Book> table where the C<Book> table has a column C<author_id>
76 containing the ID value of the C<Author>.
78 C<foreign> and C<self> are pseudo aliases and must be entered
79 literally. They will be replaced with the actual correct table alias
80 when the SQL is produced.
84 My::Schema::Book->has_many(
85 editions => 'My::Schema::Edition',
87 'foreign.publisher_id' => 'self.publisher_id',
88 'foreign.type_id' => 'self.type_id',
94 $book_rs->search_related('editions')->next
96 will result in the C<JOIN> clause:
99 LEFT JOIN edition editions ON
100 editions.publisher_id = me.publisher_id
101 AND editions.type_id = me.type_id ...
103 This describes the relationship from C<Book> to C<Edition>, where the
104 C<Edition> table refers to a publisher and a type (e.g. "paperback"):
106 As is the default in L<SQL::Abstract>, the key-value pairs will be
107 C<AND>ed in the result. C<OR> can be achieved with an arrayref, for
108 example a condition like:
110 My::Schema::Item->has_many(
111 related_item_links => My::Schema::Item::Links,
113 { 'foreign.left_itemid' => 'self.id' },
114 { 'foreign.right_itemid' => 'self.id' },
118 will translate to the following C<JOIN> clause:
120 ... FROM item me JOIN item_relations related_item_links ON
121 related_item_links.left_itemid = me.id
122 OR related_item_links.right_itemid = me.id ...
124 This describes the relationship from C<Item> to C<Item::Links>, where
125 C<Item::Links> is a many-to-many linking table, linking items back to
126 themselves in a peer fashion (without a "parent-child" designation)
128 To specify joins which describe more than a simple equality of column
129 values, the custom join condition coderef syntax can be used. For
132 My::Schema::Artist->has_many(
133 cds_80s => 'My::Schema::CD',
138 "$args->{foreign_alias}.artist" => { -ident => "$args->{self_alias}.artistid" },
139 "$args->{foreign_alias}.year" => { '>', "1979", '<', "1990" },
146 $artist_rs->search_related('cds_80s')->next;
148 will result in the C<JOIN> clause:
150 ... FROM artist me LEFT JOIN cd cds_80s ON
151 cds_80s.artist = me.artistid
155 with the bind values:
159 C<< $args->{foreign_alias} >> and C<< $args->{self_alias} >> are supplied the
160 same values that would be otherwise substituted for C<foreign> and C<self>
161 in the simple hashref syntax case.
163 The coderef is expected to return a valid L<SQL::Abstract> query-structure, just
164 like what one would supply as the first argument to
165 L<DBIx::Class::ResultSet/search>. The return value will be passed directly to
166 L<SQL::Abstract> and the resulting SQL will be used verbatim as the C<ON>
167 clause of the C<JOIN> statement associated with this relationship.
169 While every coderef-based condition must return a valid C<ON> clause, it may
170 elect to additionally return a simplified join-free condition hashref when
171 invoked as C<< $row_object->relationship >>, as opposed to
172 C<< $rs->related_resultset('relationship') >>. In this case C<$row_object> is
173 passed to the coderef as C<< $args->{self_rowobj} >>, so a user can do the
181 "$args->{foreign_alias}.artist" => { -ident => "$args->{self_alias}.artistid" },
182 "$args->{foreign_alias}.year" => { '>', "1979", '<', "1990" },
184 $args->{self_rowobj} && {
185 "$args->{foreign_alias}.artist" => $args->{self_rowobj}->artistid,
186 "$args->{foreign_alias}.year" => { '>', "1979", '<', "1990" },
193 my $artist = $schema->resultset("Artist")->find({ id => 4 });
194 $artist->cds_80s->all;
196 Can skip a C<JOIN> altogether and instead produce:
198 SELECT cds_80s.cdid, cds_80s.artist, cds_80s.title, cds_80s.year, cds_80s.genreid, cds_80s.single_track
200 WHERE cds_80s.artist = ?
204 With the bind values:
208 Note that in order to be able to use
209 L<< $row->create_related|DBIx::Class::Relationship::Base/create_related >>,
210 the coderef must not only return as its second such a "simple" condition
211 hashref which does not depend on joins being available, but the hashref must
212 contain only plain values/deflatable objects, such that the result can be
213 passed directly to L<DBIx::Class::Relationship::Base/set_from_related>. For
214 instance the C<year> constraint in the above example prevents the relationship
215 from being used to to create related objects (an exception will be thrown).
217 In order to allow the user to go truly crazy when generating a custom C<ON>
218 clause, the C<$args> hashref passed to the subroutine contains some extra
219 metadata. Currently the supplied coderef is executed as:
221 $relationship_info->{cond}->({
222 self_alias => The alias of the invoking resultset ('me' in case of a row object),
223 foreign_alias => The alias of the to-be-joined resultset (often matches relname),
224 self_resultsource => The invocant's resultsource,
225 foreign_relname => The relationship name (does *not* always match foreign_alias),
226 self_rowobj => The invocant itself in case of $row_obj->relationship
231 The L<standard ResultSet attributes|DBIx::Class::ResultSet/ATTRIBUTES> may
232 be used as relationship attributes. In particular, the 'where' attribute is
233 useful for filtering relationships:
235 __PACKAGE__->has_many( 'valid_users', 'MyApp::Schema::User',
236 { 'foreign.user_id' => 'self.user_id' },
237 { where => { valid => 1 } }
240 The following attributes are also valid:
246 Explicitly specifies the type of join to use in the relationship. Any SQL
247 join type is valid, e.g. C<LEFT> or C<RIGHT>. It will be placed in the SQL
248 command immediately before C<JOIN>.
250 =item proxy =E<gt> $column | \@columns | \%column
256 An arrayref containing a list of accessors in the foreign class to create in
257 the main class. If, for example, you do the following:
259 MyApp::Schema::CD->might_have(liner_notes => 'MyApp::Schema::LinerNotes',
261 proxy => [ qw/notes/ ],
264 Then, assuming MyApp::Schema::LinerNotes has an accessor named notes, you can do:
266 my $cd = MyApp::Schema::CD->find(1);
267 $cd->notes('Notes go here'); # set notes -- LinerNotes object is
268 # created if it doesn't exist
272 A hashref where each key is the accessor you want installed in the main class,
273 and its value is the name of the original in the fireign class.
275 MyApp::Schema::Track->belongs_to( cd => 'DBICTest::Schema::CD', 'cd', {
276 proxy => { cd_title => 'title' },
279 This will create an accessor named C<cd_title> on the C<$track> row object.
283 NOTE: you can pass a nested struct too, for example:
285 MyApp::Schema::Track->belongs_to( cd => 'DBICTest::Schema::CD', 'cd', {
286 proxy => [ 'year', { cd_title => 'title' } ],
291 Specifies the type of accessor that should be created for the relationship.
292 Valid values are C<single> (for when there is only a single related object),
293 C<multi> (when there can be many), and C<filter> (for when there is a single
294 related object, but you also want the relationship accessor to double as
295 a column accessor). For C<multi> accessors, an add_to_* method is also
296 created, which calls C<create_related> for the relationship.
298 =item is_foreign_key_constraint
300 If you are using L<SQL::Translator> to create SQL for you and you find that it
301 is creating constraints where it shouldn't, or not creating them where it
302 should, set this attribute to a true or false value to override the detection
303 of when to create constraints.
307 If C<cascade_copy> is true on a C<has_many> relationship for an
308 object, then when you copy the object all the related objects will
309 be copied too. To turn this behaviour off, pass C<< cascade_copy => 0 >>
310 in the C<$attr> hashref.
312 The behaviour defaults to C<< cascade_copy => 1 >> for C<has_many>
317 By default, DBIx::Class cascades deletes across C<has_many>,
318 C<has_one> and C<might_have> relationships. You can disable this
319 behaviour on a per-relationship basis by supplying
320 C<< cascade_delete => 0 >> in the relationship attributes.
322 The cascaded operations are performed after the requested delete,
323 so if your database has a constraint on the relationship, it will
324 have deleted/updated the related records or raised an exception
325 before DBIx::Class gets to perform the cascaded operation.
329 By default, DBIx::Class cascades updates across C<has_one> and
330 C<might_have> relationships. You can disable this behaviour on a
331 per-relationship basis by supplying C<< cascade_update => 0 >> in
332 the relationship attributes.
334 This is not a RDMS style cascade update - it purely means that when
335 an object has update called on it, all the related objects also
336 have update called. It will not change foreign keys automatically -
337 you must arrange to do this yourself.
339 =item on_delete / on_update
341 If you are using L<SQL::Translator> to create SQL for you, you can use these
342 attributes to explicitly set the desired C<ON DELETE> or C<ON UPDATE> constraint
343 type. If not supplied the SQLT parser will attempt to infer the constraint type by
344 interrogating the attributes of the B<opposite> relationship. For any 'multi'
345 relationship with C<< cascade_delete => 1 >>, the corresponding belongs_to
346 relationship will be created with an C<ON DELETE CASCADE> constraint. For any
347 relationship bearing C<< cascade_copy => 1 >> the resulting belongs_to constraint
348 will be C<ON UPDATE CASCADE>. If you wish to disable this autodetection, and just
349 use the RDBMS' default constraint type, pass C<< on_delete => undef >> or
350 C<< on_delete => '' >>, and the same for C<on_update> respectively.
354 Tells L<SQL::Translator> that the foreign key constraint it creates should be
355 deferrable. In other words, the user may request that the constraint be ignored
356 until the end of the transaction. Currently, only the PostgreSQL producer
357 actually supports this.
361 Tells L<SQL::Translator> to add an index for this constraint. Can also be
362 specified globally in the args to L<DBIx::Class::Schema/deploy> or
363 L<DBIx::Class::Schema/create_ddl_dir>. Default is on, set to 0 to disable.
367 =head2 register_relationship
371 =item Arguments: $relname, $rel_info
375 Registers a relationship on the class. This is called internally by
376 DBIx::Class::ResultSourceProxy to set up Accessors and Proxies.
380 sub register_relationship { }
382 =head2 related_resultset
386 =item Arguments: $relationship_name
388 =item Return Value: $related_resultset
392 $rs = $cd->related_resultset('artist');
394 Returns a L<DBIx::Class::ResultSet> for the relationship named
399 sub related_resultset {
401 $self->throw_exception("Can't call *_related as class methods")
404 my $rel_info = $self->relationship_info($rel);
405 $self->throw_exception( "No such relationship ${rel}" )
408 return $self->{related_resultsets}{$rel} ||= do {
409 my $attrs = (@_ > 1 && ref $_[$#_] eq 'HASH' ? pop(@_) : {});
410 $attrs = { %{$rel_info->{attrs} || {}}, %$attrs };
412 $self->throw_exception( "Invalid query: @_" )
413 if (@_ > 1 && (@_ % 2 == 1));
414 my $query = ((@_ > 1) ? {@_} : shift);
416 my $source = $self->result_source;
418 # condition resolution may fail if an incomplete master-object prefetch
419 # is encountered - that is ok during prefetch construction (not yet in_storage)
420 my ($cond, $is_crosstable) = try {
421 $source->_resolve_condition( $rel_info->{cond}, $rel, $self, $rel )
424 if ($self->in_storage) {
425 $self->throw_exception ($_);
428 $DBIx::Class::ResultSource::UNRESOLVABLE_CONDITION; # RV
431 # keep in mind that the following if() block is part of a do{} - no return()s!!!
432 if ($is_crosstable) {
433 $self->throw_exception (
434 "A cross-table relationship condition returned for statically declared '$rel'")
435 unless ref $rel_info->{cond} eq 'CODE';
437 # A WHOREIFFIC hack to reinvoke the entire condition resolution
438 # with the correct alias. Another way of doing this involves a
439 # lot of state passing around, and the @_ positions are already
440 # mapped out, making this crap a less icky option.
442 # The point of this exercise is to retain the spirit of the original
443 # $obj->search_related($rel) where the resulting rset will have the
444 # root alias as 'me', instead of $rel (as opposed to invoking
445 # $rs->search_related)
447 local $source->{_relationships}{me} = $source->{_relationships}{$rel}; # make the fake 'me' rel
448 my $obj_table_alias = lc($source->source_name) . '__row';
449 $obj_table_alias =~ s/\W+/_/g;
451 $source->resultset->search(
452 $self->ident_condition($obj_table_alias),
453 { alias => $obj_table_alias },
454 )->search_related('me', $query, $attrs)
457 # FIXME - this conditional doesn't seem correct - got to figure out
458 # at some point what it does. Also the entire UNRESOLVABLE_CONDITION
459 # business seems shady - we could simply not query *at all*
460 if ($cond eq $DBIx::Class::ResultSource::UNRESOLVABLE_CONDITION) {
461 my $reverse = $source->reverse_relationship_info($rel);
462 foreach my $rev_rel (keys %$reverse) {
463 if ($reverse->{$rev_rel}{attrs}{accessor} && $reverse->{$rev_rel}{attrs}{accessor} eq 'multi') {
464 weaken($attrs->{related_objects}{$rev_rel}[0] = $self);
466 weaken($attrs->{related_objects}{$rev_rel} = $self);
470 elsif (ref $cond eq 'ARRAY') {
472 if (ref $_ eq 'HASH') {
474 foreach my $key (keys %$_) {
475 my $newkey = $key !~ /\./ ? "me.$key" : $key;
476 $hash->{$newkey} = $_->{$key};
484 elsif (ref $cond eq 'HASH') {
485 foreach my $key (grep { ! /\./ } keys %$cond) {
486 $cond->{"me.$key"} = delete $cond->{$key};
490 $query = ($query ? { '-and' => [ $cond, $query ] } : $cond);
491 $self->result_source->related_source($rel)->resultset->search(
498 =head2 search_related
500 @objects = $rs->search_related('relname', $cond, $attrs);
501 $objects_rs = $rs->search_related('relname', $cond, $attrs);
503 Run a search on a related resultset. The search will be restricted to the
504 item or items represented by the L<DBIx::Class::ResultSet> it was called
505 upon. This method can be called on a ResultSet, a Row or a ResultSource class.
510 return shift->related_resultset(shift)->search(@_);
513 =head2 search_related_rs
515 ( $objects_rs ) = $rs->search_related_rs('relname', $cond, $attrs);
517 This method works exactly the same as search_related, except that
518 it guarantees a resultset, even in list context.
522 sub search_related_rs {
523 return shift->related_resultset(shift)->search_rs(@_);
528 $obj->count_related('relname', $cond, $attrs);
530 Returns the count of all the items in the related resultset, restricted by the
531 current item or where conditions. Can be called on a
532 L<DBIx::Class::Manual::Glossary/"ResultSet"> or a
533 L<DBIx::Class::Manual::Glossary/"Row"> object.
539 return $self->search_related(@_)->count;
544 my $new_obj = $obj->new_related('relname', \%col_data);
546 Create a new item of the related foreign class. If called on a
547 L<Row|DBIx::Class::Manual::Glossary/"Row"> object, it will magically
548 set any foreign key columns of the new object to the related primary
549 key columns of the source object for you. The newly created item will
550 not be saved into your storage until you call L<DBIx::Class::Row/insert>
556 my ($self, $rel, $values, $attrs) = @_;
558 # FIXME - this is a bad position for this (also an identical copy in
559 # set_from_related), but I have no saner way to hook, and I absolutely
560 # want this to throw at least for coderefs, instead of the "insert a NULL
561 # when it gets hard" insanity --ribasushi
563 # sanity check - currently throw when a complex coderef rel is encountered
564 # FIXME - should THROW MOAR!
566 if (ref $self) { # cdbi calls this as a class method, /me vomits
568 my $rsrc = $self->result_source;
569 my (undef, $crosstable, $relcols) = $rsrc->_resolve_condition (
570 $rsrc->relationship_info($rel)->{cond}, $rel, $self, $rel
573 $self->throw_exception("Custom relationship '$rel' does not resolve to a join-free condition fragment")
576 if (@{$relcols || []} and @$relcols = grep { ! exists $values->{$_} } @$relcols) {
577 $self->throw_exception(sprintf (
578 "Custom relationship '%s' not definitive - returns conditions instead of values for column(s): %s",
580 map { "'$_'" } @$relcols
585 my $row = $self->search_related($rel)->new($values, $attrs);
589 =head2 create_related
591 my $new_obj = $obj->create_related('relname', \%col_data);
593 Creates a new item, similarly to new_related, and also inserts the item's data
594 into your storage medium. See the distinction between C<create> and C<new>
595 in L<DBIx::Class::ResultSet> for details.
602 my $obj = $self->new_related($rel, @_)->insert;
603 delete $self->{related_resultsets}->{$rel};
609 my $found_item = $obj->find_related('relname', @pri_vals | \%pri_vals);
611 Attempt to find a related object using its primary key or unique constraints.
612 See L<DBIx::Class::ResultSet/find> for details.
619 return $self->search_related($rel)->find(@_);
622 =head2 find_or_new_related
624 my $new_obj = $obj->find_or_new_related('relname', \%col_data);
626 Find an item of a related class. If none exists, instantiate a new item of the
627 related class. The object will not be saved into your storage until you call
628 L<DBIx::Class::Row/insert> on it.
632 sub find_or_new_related {
634 my $obj = $self->find_related(@_);
635 return defined $obj ? $obj : $self->new_related(@_);
638 =head2 find_or_create_related
640 my $new_obj = $obj->find_or_create_related('relname', \%col_data);
642 Find or create an item of a related class. See
643 L<DBIx::Class::ResultSet/find_or_create> for details.
647 sub find_or_create_related {
649 my $obj = $self->find_related(@_);
650 return (defined($obj) ? $obj : $self->create_related(@_));
653 =head2 update_or_create_related
655 my $updated_item = $obj->update_or_create_related('relname', \%col_data, \%attrs?);
657 Update or create an item of a related class. See
658 L<DBIx::Class::ResultSet/update_or_create> for details.
662 sub update_or_create_related {
665 return $self->related_resultset($rel)->update_or_create(@_);
668 =head2 set_from_related
670 $book->set_from_related('author', $author_obj);
671 $book->author($author_obj); ## same thing
673 Set column values on the current object, using related values from the given
674 related object. This is used to associate previously separate objects, for
675 example, to set the correct author for a book, find the Author object, then
676 call set_from_related on the book.
678 This is called internally when you pass existing objects as values to
679 L<DBIx::Class::ResultSet/create>, or pass an object to a belongs_to accessor.
681 The columns are only set in the local copy of the object, call L</update> to
682 set them in the storage.
686 sub set_from_related {
687 my ($self, $rel, $f_obj) = @_;
689 my $rsrc = $self->result_source;
690 my $rel_info = $rsrc->relationship_info($rel)
691 or $self->throw_exception( "No such relationship ${rel}" );
693 if (defined $f_obj) {
694 my $f_class = $rel_info->{class};
695 $self->throw_exception( "Object $f_obj isn't a ".$f_class )
696 unless blessed $f_obj and $f_obj->isa($f_class);
700 # FIXME - this is a bad position for this (also an identical copy in
701 # new_related), but I have no saner way to hook, and I absolutely
702 # want this to throw at least for coderefs, instead of the "insert a NULL
703 # when it gets hard" insanity --ribasushi
705 # sanity check - currently throw when a complex coderef rel is encountered
706 # FIXME - should THROW MOAR!
707 my ($cond, $crosstable, $relcols) = $rsrc->_resolve_condition (
708 $rel_info->{cond}, $f_obj, $rel, $rel
710 $self->throw_exception("Custom relationship '$rel' does not resolve to a join-free condition fragment")
712 $self->throw_exception(sprintf (
713 "Custom relationship '%s' not definitive - returns conditions instead of values for column(s): %s",
715 map { "'$_'" } @$relcols
716 )) if @{$relcols || []};
718 $self->set_columns($cond);
723 =head2 update_from_related
725 $book->update_from_related('author', $author_obj);
727 The same as L</"set_from_related">, but the changes are immediately updated
732 sub update_from_related {
734 $self->set_from_related(@_);
738 =head2 delete_related
740 $obj->delete_related('relname', $cond, $attrs);
742 Delete any related item subject to the given conditions.
748 my $obj = $self->search_related(@_)->delete;
749 delete $self->{related_resultsets}->{$_[0]};
755 B<Currently only available for C<has_many>, C<many-to-many> and 'multi' type
760 =item Arguments: ($foreign_vals | $obj), $link_vals?
764 my $role = $schema->resultset('Role')->find(1);
765 $actor->add_to_roles($role);
766 # creates a My::DBIC::Schema::ActorRoles linking table row object
768 $actor->add_to_roles({ name => 'lead' }, { salary => 15_000_000 });
769 # creates a new My::DBIC::Schema::Role row object and the linking table
770 # object with an extra column in the link
772 Adds a linking table object for C<$obj> or C<$foreign_vals>. If the first
773 argument is a hash reference, the related object is created first with the
774 column values in the hash. If an object reference is given, just the linking
775 table object is created. In either case, any additional column values for the
776 linking table object can be specified in C<$link_vals>.
780 B<Currently only available for C<many-to-many> relationships.>
784 =item Arguments: (\@hashrefs | \@objs), $link_vals?
788 my $actor = $schema->resultset('Actor')->find(1);
789 my @roles = $schema->resultset('Role')->search({ role =>
790 { '-in' => ['Fred', 'Barney'] } } );
792 $actor->set_roles(\@roles);
793 # Replaces all of $actor's previous roles with the two named
795 $actor->set_roles(\@roles, { salary => 15_000_000 });
796 # Sets a column in the link table for all roles
799 Replace all the related objects with the given reference to a list of
800 objects. This does a C<delete> B<on the link table resultset> to remove the
801 association between the current object and all related objects, then calls
802 C<add_to_$rel> repeatedly to link all the new objects.
804 Note that this means that this method will B<not> delete any objects in the
805 table on the right side of the relation, merely that it will delete the link
808 Due to a mistake in the original implementation of this method, it will also
809 accept a list of objects or hash references. This is B<deprecated> and will be
810 removed in a future version.
812 =head2 remove_from_$rel
814 B<Currently only available for C<many-to-many> relationships.>
818 =item Arguments: $obj
822 my $role = $schema->resultset('Role')->find(1);
823 $actor->remove_from_roles($role);
824 # removes $role's My::DBIC::Schema::ActorRoles linking table row object
826 Removes the link between the current object and the related object. Note that
827 the related object itself won't be deleted unless you call ->delete() on
828 it. This method just removes the link between the two objects.
832 Matt S. Trout <mst@shadowcatsystems.co.uk>
836 You may distribute this code under the same terms as Perl itself.