1 package DBIx::Class::ResultSet;
5 use base qw/DBIx::Class/;
7 use DBIx::Class::ResultSetColumn;
8 use Scalar::Util qw/blessed weaken reftype/;
9 use DBIx::Class::_Util 'fail_on_internal_wantarray';
11 use Data::Compare (); # no imports!!! guard against insane architecture
13 # not importing first() as it will clash with our own method
17 # De-duplication in _merge_attr() is disabled, but left in for reference
18 # (the merger is used for other things that ought not to be de-duped)
19 *__HM_DEDUP = sub () { 0 };
29 # this is real - CDBICompat overrides it with insanity
30 # yes, prototype won't matter, but that's for now ;)
33 __PACKAGE__->mk_group_accessors('simple' => qw/_result_class result_source/);
37 DBIx::Class::ResultSet - Represents a query used for fetching a set of results.
41 my $users_rs = $schema->resultset('User');
42 while( $user = $users_rs->next) {
43 print $user->username;
46 my $registered_users_rs = $schema->resultset('User')->search({ registered => 1 });
47 my @cds_in_2005 = $schema->resultset('CD')->search({ year => 2005 })->all();
51 A ResultSet is an object which stores a set of conditions representing
52 a query. It is the backbone of DBIx::Class (i.e. the really
53 important/useful bit).
55 No SQL is executed on the database when a ResultSet is created, it
56 just stores all the conditions needed to create the query.
58 A basic ResultSet representing the data of an entire table is returned
59 by calling C<resultset> on a L<DBIx::Class::Schema> and passing in a
60 L<Source|DBIx::Class::Manual::Glossary/Source> name.
62 my $users_rs = $schema->resultset('User');
64 A new ResultSet is returned from calling L</search> on an existing
65 ResultSet. The new one will contain all the conditions of the
66 original, plus any new conditions added in the C<search> call.
68 A ResultSet also incorporates an implicit iterator. L</next> and L</reset>
69 can be used to walk through all the L<DBIx::Class::Row>s the ResultSet
72 The query that the ResultSet represents is B<only> executed against
73 the database when these methods are called:
74 L</find>, L</next>, L</all>, L</first>, L</single>, L</count>.
76 If a resultset is used in a numeric context it returns the L</count>.
77 However, if it is used in a boolean context it is B<always> true. So if
78 you want to check if a resultset has any results, you must use C<if $rs
81 =head1 CUSTOM ResultSet CLASSES THAT USE Moose
83 If you want to make your custom ResultSet classes with L<Moose>, use a template
86 package MyApp::Schema::ResultSet::User;
89 use namespace::autoclean;
91 extends 'DBIx::Class::ResultSet';
93 sub BUILDARGS { $_[2] }
97 __PACKAGE__->meta->make_immutable;
101 The L<MooseX::NonMoose> is necessary so that the L<Moose> constructor does not
102 clash with the regular ResultSet constructor. Alternatively, you can use:
104 __PACKAGE__->meta->make_immutable(inline_constructor => 0);
106 The L<BUILDARGS|Moose::Manual::Construction/BUILDARGS> is necessary because the
107 signature of the ResultSet C<new> is C<< ->new($source, \%args) >>.
111 =head2 Chaining resultsets
113 Let's say you've got a query that needs to be run to return some data
114 to the user. But, you have an authorization system in place that
115 prevents certain users from seeing certain information. So, you want
116 to construct the basic query in one method, but add constraints to it in
121 my $request = $self->get_request; # Get a request object somehow.
122 my $schema = $self->result_source->schema;
124 my $cd_rs = $schema->resultset('CD')->search({
125 title => $request->param('title'),
126 year => $request->param('year'),
129 $cd_rs = $self->apply_security_policy( $cd_rs );
131 return $cd_rs->all();
134 sub apply_security_policy {
143 =head3 Resolving conditions and attributes
145 When a resultset is chained from another resultset (e.g.:
146 C<< my $new_rs = $old_rs->search(\%extra_cond, \%attrs) >>), conditions
147 and attributes with the same keys need resolving.
149 If any of L</columns>, L</select>, L</as> are present, they reset the
150 original selection, and start the selection "clean".
152 The L</join>, L</prefetch>, L</+columns>, L</+select>, L</+as> attributes
153 are merged into the existing ones from the original resultset.
155 The L</where> and L</having> attributes, and any search conditions, are
156 merged with an SQL C<AND> to the existing condition from the original
159 All other attributes are overridden by any new ones supplied in the
162 =head2 Multiple queries
164 Since a resultset just defines a query, you can do all sorts of
165 things with it with the same object.
167 # Don't hit the DB yet.
168 my $cd_rs = $schema->resultset('CD')->search({
169 title => 'something',
173 # Each of these hits the DB individually.
174 my $count = $cd_rs->count;
175 my $most_recent = $cd_rs->get_column('date_released')->max();
176 my @records = $cd_rs->all;
178 And it's not just limited to SELECT statements.
184 $cd_rs->create({ artist => 'Fred' });
186 Which is the same as:
188 $schema->resultset('CD')->create({
189 title => 'something',
194 See: L</search>, L</count>, L</get_column>, L</all>, L</create>.
202 =item Arguments: L<$source|DBIx::Class::ResultSource>, L<\%attrs?|/ATTRIBUTES>
204 =item Return Value: L<$resultset|/search>
208 The resultset constructor. Takes a source object (usually a
209 L<DBIx::Class::ResultSourceProxy::Table>) and an attribute hash (see
210 L</ATTRIBUTES> below). Does not perform any queries -- these are
211 executed as needed by the other methods.
213 Generally you never construct a resultset manually. Instead you get one
215 C<< $schema->L<resultset|DBIx::Class::Schema/resultset>('$source_name') >>
216 or C<< $another_resultset->L<search|/search>(...) >> (the later called in
219 my $rs = $schema->resultset('CD')->search({ title => '100th Window' });
225 If called on an object, proxies to L</new_result> instead, so
227 my $cd = $schema->resultset('CD')->new({ title => 'Spoon' });
229 will return a CD object, not a ResultSet, and is equivalent to:
231 my $cd = $schema->resultset('CD')->new_result({ title => 'Spoon' });
233 Please also keep in mind that many internals call L</new_result> directly,
234 so overloading this method with the idea of intercepting new result object
235 creation B<will not work>. See also warning pertaining to L</create>.
243 return $class->new_result(@_) if ref $class;
245 my ($source, $attrs) = @_;
246 $source = $source->resolve
247 if $source->isa('DBIx::Class::ResultSourceHandle');
249 $attrs = { %{$attrs||{}} };
250 delete @{$attrs}{qw(_last_sqlmaker_alias_map _related_results_construction)};
252 if ($attrs->{page}) {
253 $attrs->{rows} ||= 10;
256 $attrs->{alias} ||= 'me';
259 result_source => $source,
260 cond => $attrs->{where},
265 # if there is a dark selector, this means we are already in a
266 # chain and the cleanup/sanification was taken care of by
268 $self->_normalize_selection($attrs)
269 unless $attrs->{_dark_selector};
272 $attrs->{result_class} || $source->result_class
282 =item Arguments: L<$cond|DBIx::Class::SQLMaker> | undef, L<\%attrs?|/ATTRIBUTES>
284 =item Return Value: $resultset (scalar context) | L<@result_objs|DBIx::Class::Manual::ResultClass> (list context)
288 my @cds = $cd_rs->search({ year => 2001 }); # "... WHERE year = 2001"
289 my $new_rs = $cd_rs->search({ year => 2005 });
291 my $new_rs = $cd_rs->search([ { year => 2005 }, { year => 2004 } ]);
292 # year = 2005 OR year = 2004
294 In list context, C<< ->all() >> is called implicitly on the resultset, thus
295 returning a list of L<result|DBIx::Class::Manual::ResultClass> objects instead.
296 To avoid that, use L</search_rs>.
298 If you need to pass in additional attributes but no additional condition,
299 call it as C<search(undef, \%attrs)>.
301 # "SELECT name, artistid FROM $artist_table"
302 my @all_artists = $schema->resultset('Artist')->search(undef, {
303 columns => [qw/name artistid/],
306 For a list of attributes that can be passed to C<search>, see
307 L</ATTRIBUTES>. For more examples of using this function, see
308 L<Searching|DBIx::Class::Manual::Cookbook/SEARCHING>. For a complete
309 documentation for the first argument, see L<SQL::Abstract/"WHERE CLAUSES">
310 and its extension L<DBIx::Class::SQLMaker>.
312 For more help on using joins with search, see L<DBIx::Class::Manual::Joining>.
316 Note that L</search> does not process/deflate any of the values passed in the
317 L<SQL::Abstract>-compatible search condition structure. This is unlike other
318 condition-bound methods L</new_result>, L</create> and L</find>. The user must ensure
319 manually that any value passed to this method will stringify to something the
320 RDBMS knows how to deal with. A notable example is the handling of L<DateTime>
321 objects, for more info see:
322 L<DBIx::Class::Manual::Cookbook/Formatting DateTime objects in queries>.
328 my $rs = $self->search_rs( @_ );
331 DBIx::Class::_ENV_::ASSERT_NO_INTERNAL_WANTARRAY and my $sog = fail_on_internal_wantarray($rs);
334 elsif (defined wantarray) {
338 # we can be called by a relationship helper, which in
339 # turn may be called in void context due to some braindead
340 # overload or whatever else the user decided to be clever
341 # at this particular day. Thus limit the exception to
342 # external code calls only
343 $self->throw_exception ('->search is *not* a mutator, calling it in void context makes no sense')
344 if (caller)[0] !~ /^\QDBIx::Class::/;
354 =item Arguments: L<$cond|DBIx::Class::SQLMaker>, L<\%attrs?|/ATTRIBUTES>
356 =item Return Value: L<$resultset|/search>
360 This method does the same exact thing as search() except it will
361 always return a resultset, even in list context.
368 my $rsrc = $self->result_source;
369 my ($call_cond, $call_attrs);
371 # Special-case handling for (undef, undef) or (undef)
372 # Note that (foo => undef) is valid deprecated syntax
373 @_ = () if not scalar grep { defined $_ } @_;
379 # fish out attrs in the ($condref, $attr) case
380 elsif (@_ == 2 and ( ! defined $_[0] or (ref $_[0]) ne '') ) {
381 ($call_cond, $call_attrs) = @_;
384 $self->throw_exception('Odd number of arguments to search')
388 carp_unique 'search( %condition ) is deprecated, use search( \%condition ) instead'
389 unless $rsrc->result_class->isa('DBIx::Class::CDBICompat');
391 for my $i (0 .. $#_) {
393 $self->throw_exception ('All keys in condition key/value pairs must be plain scalars')
394 if (! defined $_[$i] or ref $_[$i] ne '');
400 # see if we can keep the cache (no $rs changes)
402 my %safe = (alias => 1, cache => 1);
403 if ( ! List::Util::first { !$safe{$_} } keys %$call_attrs and (
406 ref $call_cond eq 'HASH' && ! keys %$call_cond
408 ref $call_cond eq 'ARRAY' && ! @$call_cond
410 $cache = $self->get_cache;
413 my $old_attrs = { %{$self->{attrs}} };
414 my ($old_having, $old_where) = delete @{$old_attrs}{qw(having where)};
416 my $new_attrs = { %$old_attrs };
418 # take care of call attrs (only if anything is changing)
419 if ($call_attrs and keys %$call_attrs) {
421 # copy for _normalize_selection
422 $call_attrs = { %$call_attrs };
424 my @selector_attrs = qw/select as columns cols +select +as +columns include_columns/;
426 # reset the current selector list if new selectors are supplied
427 if (List::Util::first { exists $call_attrs->{$_} } qw/columns cols select as/) {
428 delete @{$old_attrs}{(@selector_attrs, '_dark_selector')};
431 # Normalize the new selector list (operates on the passed-in attr structure)
432 # Need to do it on every chain instead of only once on _resolved_attrs, in
433 # order to allow detection of empty vs partial 'as'
434 $call_attrs->{_dark_selector} = $old_attrs->{_dark_selector}
435 if $old_attrs->{_dark_selector};
436 $self->_normalize_selection ($call_attrs);
438 # start with blind overwriting merge, exclude selector attrs
439 $new_attrs = { %{$old_attrs}, %{$call_attrs} };
440 delete @{$new_attrs}{@selector_attrs};
442 for (@selector_attrs) {
443 $new_attrs->{$_} = $self->_merge_attr($old_attrs->{$_}, $call_attrs->{$_})
444 if ( exists $old_attrs->{$_} or exists $call_attrs->{$_} );
447 # older deprecated name, use only if {columns} is not there
448 if (my $c = delete $new_attrs->{cols}) {
449 carp_unique( "Resultset attribute 'cols' is deprecated, use 'columns' instead" );
450 if ($new_attrs->{columns}) {
451 carp "Resultset specifies both the 'columns' and the legacy 'cols' attributes - ignoring 'cols'";
454 $new_attrs->{columns} = $c;
459 # join/prefetch use their own crazy merging heuristics
460 foreach my $key (qw/join prefetch/) {
461 $new_attrs->{$key} = $self->_merge_joinpref_attr($old_attrs->{$key}, $call_attrs->{$key})
462 if exists $call_attrs->{$key};
465 # stack binds together
466 $new_attrs->{bind} = [ @{ $old_attrs->{bind} || [] }, @{ $call_attrs->{bind} || [] } ];
470 for ($old_where, $call_cond) {
472 $new_attrs->{where} = $self->_stack_cond (
473 $_, $new_attrs->{where}
478 if (defined $old_having) {
479 $new_attrs->{having} = $self->_stack_cond (
480 $old_having, $new_attrs->{having}
484 my $rs = (ref $self)->new($rsrc, $new_attrs);
486 $rs->set_cache($cache) if ($cache);
492 sub _normalize_selection {
493 my ($self, $attrs) = @_;
496 if ( exists $attrs->{include_columns} ) {
497 carp_unique( "Resultset attribute 'include_columns' is deprecated, use '+columns' instead" );
498 $attrs->{'+columns'} = $self->_merge_attr(
499 $attrs->{'+columns'}, delete $attrs->{include_columns}
503 # columns are always placed first, however
505 # Keep the X vs +X separation until _resolved_attrs time - this allows to
506 # delay the decision on whether to use a default select list ($rsrc->columns)
507 # allowing stuff like the remove_columns helper to work
509 # select/as +select/+as pairs need special handling - the amount of select/as
510 # elements in each pair does *not* have to be equal (think multicolumn
511 # selectors like distinct(foo, bar) ). If the selector is bare (no 'as'
512 # supplied at all) - try to infer the alias, either from the -as parameter
513 # of the selector spec, or use the parameter whole if it looks like a column
514 # name (ugly legacy heuristic). If all fails - leave the selector bare (which
515 # is ok as well), but make sure no more additions to the 'as' chain take place
516 for my $pref ('', '+') {
518 my ($sel, $as) = map {
519 my $key = "${pref}${_}";
521 my $val = [ ref $attrs->{$key} eq 'ARRAY'
523 : $attrs->{$key} || ()
525 delete $attrs->{$key};
529 if (! @$as and ! @$sel ) {
532 elsif (@$as and ! @$sel) {
533 $self->throw_exception(
534 "Unable to handle ${pref}as specification (@$as) without a corresponding ${pref}select"
538 # no as part supplied at all - try to deduce (unless explicit end of named selection is declared)
539 # if any @$as has been supplied we assume the user knows what (s)he is doing
540 # and blindly keep stacking up pieces
541 unless ($attrs->{_dark_selector}) {
544 if ( ref $_ eq 'HASH' and exists $_->{-as} ) {
545 push @$as, $_->{-as};
547 # assume any plain no-space, no-parenthesis string to be a column spec
548 # FIXME - this is retarded but is necessary to support shit like 'count(foo)'
549 elsif ( ! ref $_ and $_ =~ /^ [^\s\(\)]+ $/x) {
552 # if all else fails - raise a flag that no more aliasing will be allowed
554 $attrs->{_dark_selector} = {
556 string => ($dark_sel_dumper ||= do {
557 require Data::Dumper::Concise;
558 Data::Dumper::Concise::DumperObject()->Indent(0);
559 })->Values([$_])->Dump
567 elsif (@$as < @$sel) {
568 $self->throw_exception(
569 "Unable to handle an ${pref}as specification (@$as) with less elements than the corresponding ${pref}select"
572 elsif ($pref and $attrs->{_dark_selector}) {
573 $self->throw_exception(
574 "Unable to process named '+select', resultset contains an unnamed selector $attrs->{_dark_selector}{string}"
580 $attrs->{"${pref}select"} = $self->_merge_attr($attrs->{"${pref}select"}, $sel);
581 $attrs->{"${pref}as"} = $self->_merge_attr($attrs->{"${pref}as"}, $as);
586 my ($self, $left, $right) = @_;
589 (ref $_ eq 'ARRAY' and !@$_)
591 (ref $_ eq 'HASH' and ! keys %$_)
592 ) and $_ = undef for ($left, $right);
594 # either on of the two undef or both undef
595 if ( ( (defined $left) xor (defined $right) ) or ! defined $left ) {
596 return defined $left ? $left : $right;
599 my $cond = $self->result_source->schema->storage->_collapse_cond({ -and => [$left, $right] });
601 for my $c (grep { ref $cond->{$_} eq 'ARRAY' and ($cond->{$_}[0]||'') eq '-and' } keys %$cond) {
603 my @vals = sort @{$cond->{$c}}[ 1..$#{$cond->{$c}} ];
604 my @fin = shift @vals;
607 push @fin, $v unless Data::Compare::Compare( $fin[-1], $v );
610 $cond->{$c} = (@fin == 1) ? $fin[0] : [-and => @fin ];
616 =head2 search_literal
618 B<CAVEAT>: C<search_literal> is provided for Class::DBI compatibility and
619 should only be used in that context. C<search_literal> is a convenience
620 method. It is equivalent to calling C<< $schema->search(\[]) >>, but if you
621 want to ensure columns are bound correctly, use L</search>.
623 See L<DBIx::Class::Manual::Cookbook/SEARCHING> and
624 L<DBIx::Class::Manual::FAQ/Searching> for searching techniques that do not
625 require C<search_literal>.
629 =item Arguments: $sql_fragment, @standalone_bind_values
631 =item Return Value: L<$resultset|/search> (scalar context) | L<@result_objs|DBIx::Class::Manual::ResultClass> (list context)
635 my @cds = $cd_rs->search_literal('year = ? AND title = ?', qw/2001 Reload/);
636 my $newrs = $artist_rs->search_literal('name = ?', 'Metallica');
638 Pass a literal chunk of SQL to be added to the conditional part of the
641 Example of how to use C<search> instead of C<search_literal>
643 my @cds = $cd_rs->search_literal('cdid = ? AND (artist = ? OR artist = ?)', (2, 1, 2));
644 my @cds = $cd_rs->search(\[ 'cdid = ? AND (artist = ? OR artist = ?)', [ 'cdid', 2 ], [ 'artist', 1 ], [ 'artist', 2 ] ]);
649 my ($self, $sql, @bind) = @_;
651 if ( @bind && ref($bind[-1]) eq 'HASH' ) {
654 return $self->search(\[ $sql, map [ {} => $_ ], @bind ], ($attr || () ));
661 =item Arguments: \%columns_values | @pk_values, { key => $unique_constraint, L<%attrs|/ATTRIBUTES> }?
663 =item Return Value: L<$result|DBIx::Class::Manual::ResultClass> | undef
667 Finds and returns a single row based on supplied criteria. Takes either a
668 hashref with the same format as L</create> (including inference of foreign
669 keys from related objects), or a list of primary key values in the same
670 order as the L<primary columns|DBIx::Class::ResultSource/primary_columns>
671 declaration on the L</result_source>.
673 In either case an attempt is made to combine conditions already existing on
674 the resultset with the condition passed to this method.
676 To aid with preparing the correct query for the storage you may supply the
677 C<key> attribute, which is the name of a
678 L<unique constraint|DBIx::Class::ResultSource/add_unique_constraint> (the
679 unique constraint corresponding to the
680 L<primary columns|DBIx::Class::ResultSource/primary_columns> is always named
681 C<primary>). If the C<key> attribute has been supplied, and DBIC is unable
682 to construct a query that satisfies the named unique constraint fully (
683 non-NULL values for each column member of the constraint) an exception is
686 If no C<key> is specified, the search is carried over all unique constraints
687 which are fully defined by the available condition.
689 If no such constraint is found, C<find> currently defaults to a simple
690 C<< search->(\%column_values) >> which may or may not do what you expect.
691 Note that this fallback behavior may be deprecated in further versions. If
692 you need to search with arbitrary conditions - use L</search>. If the query
693 resulting from this fallback produces more than one row, a warning to the
694 effect is issued, though only the first row is constructed and returned as
697 In addition to C<key>, L</find> recognizes and applies standard
698 L<resultset attributes|/ATTRIBUTES> in the same way as L</search> does.
700 Note that if you have extra concerns about the correctness of the resulting
701 query you need to specify the C<key> attribute and supply the entire condition
702 as an argument to find (since it is not always possible to perform the
703 combination of the resultset condition with the supplied one, especially if
704 the resultset condition contains literal sql).
706 For example, to find a row by its primary key:
708 my $cd = $schema->resultset('CD')->find(5);
710 You can also find a row by a specific unique constraint:
712 my $cd = $schema->resultset('CD')->find(
714 artist => 'Massive Attack',
715 title => 'Mezzanine',
717 { key => 'cd_artist_title' }
720 See also L</find_or_create> and L</update_or_create>.
726 my $attrs = (@_ > 1 && ref $_[$#_] eq 'HASH' ? pop(@_) : {});
728 my $rsrc = $self->result_source;
731 if (exists $attrs->{key}) {
732 $constraint_name = defined $attrs->{key}
734 : $self->throw_exception("An undefined 'key' resultset attribute makes no sense")
738 # Parse out the condition from input
741 if (ref $_[0] eq 'HASH') {
742 $call_cond = { %{$_[0]} };
745 # if only values are supplied we need to default to 'primary'
746 $constraint_name = 'primary' unless defined $constraint_name;
748 my @c_cols = $rsrc->unique_constraint_columns($constraint_name);
750 $self->throw_exception(
751 "No constraint columns, maybe a malformed '$constraint_name' constraint?"
754 $self->throw_exception (
755 'find() expects either a column/value hashref, or a list of values '
756 . "corresponding to the columns of the specified unique constraint '$constraint_name'"
757 ) unless @c_cols == @_;
760 @{$call_cond}{@c_cols} = @_;
764 for my $key (keys %$call_cond) {
766 my $keyref = ref($call_cond->{$key})
768 my $relinfo = $rsrc->relationship_info($key)
770 my $val = delete $call_cond->{$key};
772 next if $keyref eq 'ARRAY'; # has_many for multi_create
774 my $rel_q = $rsrc->_resolve_condition(
775 $relinfo->{cond}, $val, $key, $key
777 die "Can't handle complex relationship conditions in find" if ref($rel_q) ne 'HASH';
778 @related{keys %$rel_q} = values %$rel_q;
782 # relationship conditions take precedence (?)
783 @{$call_cond}{keys %related} = values %related;
785 my $alias = exists $attrs->{alias} ? $attrs->{alias} : $self->{attrs}{alias};
787 if (defined $constraint_name) {
788 $final_cond = $self->_qualify_cond_columns (
790 $self->_build_unique_cond (
798 elsif ($self->{attrs}{accessor} and $self->{attrs}{accessor} eq 'single') {
799 # This means that we got here after a merger of relationship conditions
800 # in ::Relationship::Base::search_related (the row method), and furthermore
801 # the relationship is of the 'single' type. This means that the condition
802 # provided by the relationship (already attached to $self) is sufficient,
803 # as there can be only one row in the database that would satisfy the
807 # no key was specified - fall down to heuristics mode:
808 # run through all unique queries registered on the resultset, and
809 # 'OR' all qualifying queries together
810 my (@unique_queries, %seen_column_combinations);
811 for my $c_name ($rsrc->unique_constraint_names) {
812 next if $seen_column_combinations{
813 join "\x00", sort $rsrc->unique_constraint_columns($c_name)
816 push @unique_queries, try {
817 $self->_build_unique_cond ($c_name, $call_cond, 'croak_on_nulls')
821 $final_cond = @unique_queries
822 ? [ map { $self->_qualify_cond_columns($_, $alias) } @unique_queries ]
823 : $self->_non_unique_find_fallback ($call_cond, $attrs)
827 # Run the query, passing the result_class since it should propagate for find
828 my $rs = $self->search ($final_cond, {result_class => $self->result_class, %$attrs});
829 if ($rs->_resolved_attrs->{collapse}) {
831 carp "Query returned more than one row" if $rs->next;
839 # This is a stop-gap method as agreed during the discussion on find() cleanup:
840 # http://lists.scsys.co.uk/pipermail/dbix-class/2010-October/009535.html
842 # It is invoked when find() is called in legacy-mode with insufficiently-unique
843 # condition. It is provided for overrides until a saner way forward is devised
845 # *NOTE* This is not a public method, and it's *GUARANTEED* to disappear down
846 # the road. Please adjust your tests accordingly to catch this situation early
847 # DBIx::Class::ResultSet->can('_non_unique_find_fallback') is reasonable
849 # The method will not be removed without an adequately complete replacement
850 # for strict-mode enforcement
851 sub _non_unique_find_fallback {
852 my ($self, $cond, $attrs) = @_;
854 return $self->_qualify_cond_columns(
856 exists $attrs->{alias}
858 : $self->{attrs}{alias}
863 sub _qualify_cond_columns {
864 my ($self, $cond, $alias) = @_;
866 my %aliased = %$cond;
867 for (keys %aliased) {
868 $aliased{"$alias.$_"} = delete $aliased{$_}
875 sub _build_unique_cond {
876 my ($self, $constraint_name, $extra_cond, $croak_on_null) = @_;
878 my @c_cols = $self->result_source->unique_constraint_columns($constraint_name);
880 # combination may fail if $self->{cond} is non-trivial
881 my ($final_cond) = try {
882 $self->_merge_with_rscond ($extra_cond)
887 # trim out everything not in $columns
888 $final_cond = { map {
889 exists $final_cond->{$_}
890 ? ( $_ => $final_cond->{$_} )
894 if (my @missing = grep
895 { ! ($croak_on_null ? defined $final_cond->{$_} : exists $final_cond->{$_}) }
898 $self->throw_exception( sprintf ( "Unable to satisfy requested constraint '%s', no values for column(s): %s",
900 join (', ', map { "'$_'" } @missing),
907 !$ENV{DBIC_NULLABLE_KEY_NOWARN}
909 my @undefs = sort grep { ! defined $final_cond->{$_} } (keys %$final_cond)
911 carp_unique ( sprintf (
912 "NULL/undef values supplied for requested unique constraint '%s' (NULL "
913 . 'values in column(s): %s). This is almost certainly not what you wanted, '
914 . 'though you can set DBIC_NULLABLE_KEY_NOWARN to disable this warning.',
916 join (', ', map { "'$_'" } @undefs),
923 =head2 search_related
927 =item Arguments: $rel_name, $cond?, L<\%attrs?|/ATTRIBUTES>
929 =item Return Value: L<$resultset|/search> (scalar context) | L<@result_objs|DBIx::Class::Manual::ResultClass> (list context)
933 $new_rs = $cd_rs->search_related('artist', {
937 Searches the specified relationship, optionally specifying a condition and
938 attributes for matching records. See L</ATTRIBUTES> for more information.
940 In list context, C<< ->all() >> is called implicitly on the resultset, thus
941 returning a list of result objects instead. To avoid that, use L</search_related_rs>.
943 See also L</search_related_rs>.
948 return shift->related_resultset(shift)->search(@_);
951 =head2 search_related_rs
953 This method works exactly the same as search_related, except that
954 it guarantees a resultset, even in list context.
958 sub search_related_rs {
959 return shift->related_resultset(shift)->search_rs(@_);
966 =item Arguments: none
968 =item Return Value: L<$cursor|DBIx::Class::Cursor>
972 Returns a storage-driven cursor to the given resultset. See
973 L<DBIx::Class::Cursor> for more information.
980 return $self->{cursor} ||= do {
981 my $attrs = $self->_resolved_attrs;
982 $self->result_source->storage->select(
983 $attrs->{from}, $attrs->{select}, $attrs->{where}, $attrs
992 =item Arguments: L<$cond?|DBIx::Class::SQLMaker>
994 =item Return Value: L<$result|DBIx::Class::Manual::ResultClass> | undef
998 my $cd = $schema->resultset('CD')->single({ year => 2001 });
1000 Inflates the first result without creating a cursor if the resultset has
1001 any records in it; if not returns C<undef>. Used by L</find> as a lean version
1004 While this method can take an optional search condition (just like L</search>)
1005 being a fast-code-path it does not recognize search attributes. If you need to
1006 add extra joins or similar, call L</search> and then chain-call L</single> on the
1007 L<DBIx::Class::ResultSet> returned.
1013 As of 0.08100, this method enforces the assumption that the preceding
1014 query returns only one row. If more than one row is returned, you will receive
1017 Query returned more than one row
1019 In this case, you should be using L</next> or L</find> instead, or if you really
1020 know what you are doing, use the L</rows> attribute to explicitly limit the size
1023 This method will also throw an exception if it is called on a resultset prefetching
1024 has_many, as such a prefetch implies fetching multiple rows from the database in
1025 order to assemble the resulting object.
1032 my ($self, $where) = @_;
1034 $self->throw_exception('single() only takes search conditions, no attributes. You want ->search( $cond, $attrs )->single()');
1037 my $attrs = { %{$self->_resolved_attrs} };
1039 $self->throw_exception(
1040 'single() can not be used on resultsets collapsing a has_many. Use find( \%cond ) or next() instead'
1041 ) if $attrs->{collapse};
1044 if (defined $attrs->{where}) {
1047 [ map { ref $_ eq 'ARRAY' ? [ -or => $_ ] : $_ }
1048 $where, delete $attrs->{where} ]
1051 $attrs->{where} = $where;
1055 my $data = [ $self->result_source->storage->select_single(
1056 $attrs->{from}, $attrs->{select},
1057 $attrs->{where}, $attrs
1060 return undef unless @$data;
1061 $self->{_stashed_rows} = [ $data ];
1062 $self->_construct_results->[0];
1069 =item Arguments: L<$cond?|DBIx::Class::SQLMaker>
1071 =item Return Value: L<$resultsetcolumn|DBIx::Class::ResultSetColumn>
1075 my $max_length = $rs->get_column('length')->max;
1077 Returns a L<DBIx::Class::ResultSetColumn> instance for a column of the ResultSet.
1082 my ($self, $column) = @_;
1083 my $new = DBIx::Class::ResultSetColumn->new($self, $column);
1091 =item Arguments: L<$cond|DBIx::Class::SQLMaker>, L<\%attrs?|/ATTRIBUTES>
1093 =item Return Value: L<$resultset|/search> (scalar context) | L<@result_objs|DBIx::Class::Manual::ResultClass> (list context)
1097 # WHERE title LIKE '%blue%'
1098 $cd_rs = $rs->search_like({ title => '%blue%'});
1100 Performs a search, but uses C<LIKE> instead of C<=> as the condition. Note
1101 that this is simply a convenience method retained for ex Class::DBI users.
1102 You most likely want to use L</search> with specific operators.
1104 For more information, see L<DBIx::Class::Manual::Cookbook>.
1106 This method is deprecated and will be removed in 0.09. Use L</search()>
1107 instead. An example conversion is:
1109 ->search_like({ foo => 'bar' });
1113 ->search({ foo => { like => 'bar' } });
1120 'search_like() is deprecated and will be removed in DBIC version 0.09.'
1121 .' Instead use ->search({ x => { -like => "y%" } })'
1122 .' (note the outer pair of {}s - they are important!)'
1124 my $attrs = (@_ > 1 && ref $_[$#_] eq 'HASH' ? pop(@_) : {});
1125 my $query = ref $_[0] eq 'HASH' ? { %{shift()} }: {@_};
1126 $query->{$_} = { 'like' => $query->{$_} } for keys %$query;
1127 return $class->search($query, { %$attrs });
1134 =item Arguments: $first, $last
1136 =item Return Value: L<$resultset|/search> (scalar context) | L<@result_objs|DBIx::Class::Manual::ResultClass> (list context)
1140 Returns a resultset or object list representing a subset of elements from the
1141 resultset slice is called on. Indexes are from 0, i.e., to get the first
1142 three records, call:
1144 my ($one, $two, $three) = $rs->slice(0, 2);
1149 my ($self, $min, $max) = @_;
1150 my $attrs = {}; # = { %{ $self->{attrs} || {} } };
1151 $attrs->{offset} = $self->{attrs}{offset} || 0;
1152 $attrs->{offset} += $min;
1153 $attrs->{rows} = ($max ? ($max - $min + 1) : 1);
1154 return $self->search(undef, $attrs);
1161 =item Arguments: none
1163 =item Return Value: L<$result|DBIx::Class::Manual::ResultClass> | undef
1167 Returns the next element in the resultset (C<undef> is there is none).
1169 Can be used to efficiently iterate over records in the resultset:
1171 my $rs = $schema->resultset('CD')->search;
1172 while (my $cd = $rs->next) {
1176 Note that you need to store the resultset object, and call C<next> on it.
1177 Calling C<< resultset('Table')->next >> repeatedly will always return the
1178 first record from the resultset.
1185 if (my $cache = $self->get_cache) {
1186 $self->{all_cache_position} ||= 0;
1187 return $cache->[$self->{all_cache_position}++];
1190 if ($self->{attrs}{cache}) {
1191 delete $self->{pager};
1192 $self->{all_cache_position} = 1;
1193 return ($self->all)[0];
1196 return shift(@{$self->{_stashed_results}}) if @{ $self->{_stashed_results}||[] };
1198 $self->{_stashed_results} = $self->_construct_results
1201 return shift @{$self->{_stashed_results}};
1204 # Constructs as many results as it can in one pass while respecting
1205 # cursor laziness. Several modes of operation:
1207 # * Always builds everything present in @{$self->{_stashed_rows}}
1208 # * If called with $fetch_all true - pulls everything off the cursor and
1209 # builds all result structures (or objects) in one pass
1210 # * If $self->_resolved_attrs->{collapse} is true, checks the order_by
1211 # and if the resultset is ordered properly by the left side:
1212 # * Fetches stuff off the cursor until the "master object" changes,
1213 # and saves the last extra row (if any) in @{$self->{_stashed_rows}}
1215 # * Just fetches, and collapses/constructs everything as if $fetch_all
1216 # was requested (there is no other way to collapse except for an
1218 # * If no collapse is requested - just get the next row, construct and
1220 sub _construct_results {
1221 my ($self, $fetch_all) = @_;
1223 my $rsrc = $self->result_source;
1224 my $attrs = $self->_resolved_attrs;
1229 ! $attrs->{order_by}
1233 my @pcols = $rsrc->primary_columns
1235 # default order for collapsing unless the user asked for something
1236 $attrs->{order_by} = [ map { join '.', $attrs->{alias}, $_} @pcols ];
1237 $attrs->{_ordered_for_collapse} = 1;
1238 $attrs->{_order_is_artificial} = 1;
1241 # this will be used as both initial raw-row collector AND as a RV of
1242 # _construct_results. Not regrowing the array twice matters a lot...
1243 # a surprising amount actually
1244 my $rows = delete $self->{_stashed_rows};
1246 my $cursor; # we may not need one at all
1248 my $did_fetch_all = $fetch_all;
1251 # FIXME SUBOPTIMAL - we can do better, cursor->next/all (well diff. methods) should return a ref
1252 $rows = [ ($rows ? @$rows : ()), $self->cursor->all ];
1254 elsif( $attrs->{collapse} ) {
1256 # a cursor will need to be closed over in case of collapse
1257 $cursor = $self->cursor;
1259 $attrs->{_ordered_for_collapse} = (
1265 ->_main_source_order_by_portion_is_stable($rsrc, $attrs->{order_by}, $attrs->{where})
1267 ) unless defined $attrs->{_ordered_for_collapse};
1269 if (! $attrs->{_ordered_for_collapse}) {
1272 # instead of looping over ->next, use ->all in stealth mode
1273 # *without* calling a ->reset afterwards
1274 # FIXME ENCAPSULATION - encapsulation breach, cursor method additions pending
1275 if (! $cursor->{_done}) {
1276 $rows = [ ($rows ? @$rows : ()), $cursor->all ];
1277 $cursor->{_done} = 1;
1282 if (! $did_fetch_all and ! @{$rows||[]} ) {
1283 # FIXME SUBOPTIMAL - we can do better, cursor->next/all (well diff. methods) should return a ref
1284 $cursor ||= $self->cursor;
1285 if (scalar (my @r = $cursor->next) ) {
1290 return undef unless @{$rows||[]};
1292 # sanity check - people are too clever for their own good
1293 if ($attrs->{collapse} and my $aliastypes = $attrs->{_last_sqlmaker_alias_map} ) {
1295 my $multiplied_selectors;
1296 for my $sel_alias ( grep { $_ ne $attrs->{alias} } keys %{ $aliastypes->{selecting} } ) {
1298 $aliastypes->{multiplying}{$sel_alias}
1300 $aliastypes->{premultiplied}{$sel_alias}
1302 $multiplied_selectors->{$_} = 1 for values %{$aliastypes->{selecting}{$sel_alias}{-seen_columns}}
1306 for my $i (0 .. $#{$attrs->{as}} ) {
1307 my $sel = $attrs->{select}[$i];
1309 if (ref $sel eq 'SCALAR') {
1312 elsif( ref $sel eq 'REF' and ref $$sel eq 'ARRAY' ) {
1316 $self->throw_exception(
1317 'Result collapse not possible - selection from a has_many source redirected to the main object'
1318 ) if ($multiplied_selectors->{$sel} and $attrs->{as}[$i] !~ /\./);
1322 # hotspot - skip the setter
1323 my $res_class = $self->_result_class;
1325 my $inflator_cref = $self->{_result_inflator}{cref} ||= do {
1326 $res_class->can ('inflate_result')
1327 or $self->throw_exception("Inflator $res_class does not provide an inflate_result() method");
1330 my $infmap = $attrs->{as};
1332 $self->{_result_inflator}{is_core_row} = ( (
1335 ( \&DBIx::Class::Row::inflate_result || die "No ::Row::inflate_result() - can't happen" )
1336 ) ? 1 : 0 ) unless defined $self->{_result_inflator}{is_core_row};
1338 $self->{_result_inflator}{is_hri} = ( (
1339 ! $self->{_result_inflator}{is_core_row}
1342 require DBIx::Class::ResultClass::HashRefInflator
1344 DBIx::Class::ResultClass::HashRefInflator->can('inflate_result')
1346 ) ? 1 : 0 ) unless defined $self->{_result_inflator}{is_hri};
1349 if (! $attrs->{_related_results_construction}) {
1350 # construct a much simpler array->hash folder for the one-table cases right here
1351 if ($self->{_result_inflator}{is_hri}) {
1352 for my $r (@$rows) {
1353 $r = { map { $infmap->[$_] => $r->[$_] } 0..$#$infmap };
1356 # FIXME SUBOPTIMAL this is a very very very hot spot
1357 # while rather optimal we can *still* do much better, by
1358 # building a smarter Row::inflate_result(), and
1359 # switch to feeding it data via a much leaner interface
1361 # crude unscientific benchmarking indicated the shortcut eval is not worth it for
1362 # this particular resultset size
1363 elsif (@$rows < 60) {
1364 for my $r (@$rows) {
1365 $r = $inflator_cref->($res_class, $rsrc, { map { $infmap->[$_] => $r->[$_] } (0..$#$infmap) } );
1370 '$_ = $inflator_cref->($res_class, $rsrc, { %s }) for @$rows',
1371 join (', ', map { "\$infmap->[$_] => \$_->[$_]" } 0..$#$infmap )
1377 $self->{_result_inflator}{is_hri} ? 'hri'
1378 : $self->{_result_inflator}{is_core_row} ? 'classic_pruning'
1379 : 'classic_nonpruning'
1382 # $args and $attrs to _mk_row_parser are separated to delineate what is
1383 # core collapser stuff and what is dbic $rs specific
1384 @{$self->{_row_parser}{$parser_type}}{qw(cref nullcheck)} = $rsrc->_mk_row_parser({
1386 inflate_map => $infmap,
1387 collapse => $attrs->{collapse},
1388 premultiplied => $attrs->{_main_source_premultiplied},
1389 hri_style => $self->{_result_inflator}{is_hri},
1390 prune_null_branches => $self->{_result_inflator}{is_hri} || $self->{_result_inflator}{is_core_row},
1391 }, $attrs) unless $self->{_row_parser}{$parser_type}{cref};
1393 # column_info metadata historically hasn't been too reliable.
1394 # We need to start fixing this somehow (the collapse resolver
1395 # can't work without it). Add an explicit check for the *main*
1396 # result, hopefully this will gradually weed out such errors
1398 # FIXME - this is a temporary kludge that reduces performance
1399 # It is however necessary for the time being
1400 my ($unrolled_non_null_cols_to_check, $err);
1402 if (my $check_non_null_cols = $self->{_row_parser}{$parser_type}{nullcheck} ) {
1405 'Collapse aborted due to invalid ResultSource metadata - the following '
1406 . 'selections are declared non-nullable but NULLs were retrieved: '
1410 COL: for my $i (@$check_non_null_cols) {
1411 ! defined $_->[$i] and push @violating_idx, $i and next COL for @$rows;
1414 $self->throw_exception( $err . join (', ', map { "'$infmap->[$_]'" } @violating_idx ) )
1417 $unrolled_non_null_cols_to_check = join (',', @$check_non_null_cols);
1421 ($did_fetch_all or ! $attrs->{collapse}) ? undef
1422 : defined $unrolled_non_null_cols_to_check ? eval sprintf <<'EOS', $unrolled_non_null_cols_to_check
1424 # FIXME SUBOPTIMAL - we can do better, cursor->next/all (well diff. methods) should return a ref
1425 my @r = $cursor->next or return;
1426 if (my @violating_idx = grep { ! defined $r[$_] } (%s) ) {
1427 $self->throw_exception( $err . join (', ', map { "'$infmap->[$_]'" } @violating_idx ) )
1433 # FIXME SUBOPTIMAL - we can do better, cursor->next/all (well diff. methods) should return a ref
1434 my @r = $cursor->next or return;
1439 $self->{_row_parser}{$parser_type}{cref}->(
1441 $next_cref ? ( $next_cref, $self->{_stashed_rows} = [] ) : (),
1444 # Special-case multi-object HRI - there is no $inflator_cref pass
1445 unless ($self->{_result_inflator}{is_hri}) {
1446 $_ = $inflator_cref->($res_class, $rsrc, @$_) for @$rows
1450 # The @$rows check seems odd at first - why wouldn't we want to warn
1451 # regardless? The issue is things like find() etc, where the user
1452 # *knows* only one result will come back. In these cases the ->all
1453 # is not a pessimization, but rather something we actually want
1455 'Unable to properly collapse has_many results in iterator mode due '
1456 . 'to order criteria - performed an eager cursor slurp underneath. '
1457 . 'Consider using ->all() instead'
1458 ) if ( ! $fetch_all and @$rows > 1 );
1463 =head2 result_source
1467 =item Arguments: L<$result_source?|DBIx::Class::ResultSource>
1469 =item Return Value: L<$result_source|DBIx::Class::ResultSource>
1473 An accessor for the primary ResultSource object from which this ResultSet
1480 =item Arguments: $result_class?
1482 =item Return Value: $result_class
1486 An accessor for the class to use when creating result objects. Defaults to
1487 C<< result_source->result_class >> - which in most cases is the name of the
1488 L<"table"|DBIx::Class::Manual::Glossary/"ResultSource"> class.
1490 Note that changing the result_class will also remove any components
1491 that were originally loaded in the source class via
1492 L<DBIx::Class::ResultSource/load_components>. Any overloaded methods
1493 in the original source class will not run.
1498 my ($self, $result_class) = @_;
1499 if ($result_class) {
1501 # don't fire this for an object
1502 $self->ensure_class_loaded($result_class)
1503 unless ref($result_class);
1505 if ($self->get_cache) {
1506 carp_unique('Changing the result_class of a ResultSet instance with cached results is a noop - the cache contents will not be altered');
1508 # FIXME ENCAPSULATION - encapsulation breach, cursor method additions pending
1509 elsif ($self->{cursor} && $self->{cursor}{_pos}) {
1510 $self->throw_exception('Changing the result_class of a ResultSet instance with an active cursor is not supported');
1513 $self->_result_class($result_class);
1515 delete $self->{_result_inflator};
1517 $self->_result_class;
1524 =item Arguments: L<$cond|DBIx::Class::SQLMaker>, L<\%attrs?|/ATTRIBUTES>
1526 =item Return Value: $count
1530 Performs an SQL C<COUNT> with the same query as the resultset was built
1531 with to find the number of elements. Passing arguments is equivalent to
1532 C<< $rs->search ($cond, \%attrs)->count >>
1538 return $self->search(@_)->count if @_ and defined $_[0];
1539 return scalar @{ $self->get_cache } if $self->get_cache;
1541 my $attrs = { %{ $self->_resolved_attrs } };
1543 # this is a little optimization - it is faster to do the limit
1544 # adjustments in software, instead of a subquery
1545 my ($rows, $offset) = delete @{$attrs}{qw/rows offset/};
1548 if ($self->_has_resolved_attr (qw/collapse group_by/)) {
1549 $crs = $self->_count_subq_rs ($attrs);
1552 $crs = $self->_count_rs ($attrs);
1554 my $count = $crs->next;
1556 $count -= $offset if $offset;
1557 $count = $rows if $rows and $rows < $count;
1558 $count = 0 if ($count < 0);
1567 =item Arguments: L<$cond|DBIx::Class::SQLMaker>, L<\%attrs?|/ATTRIBUTES>
1569 =item Return Value: L<$count_rs|DBIx::Class::ResultSetColumn>
1573 Same as L</count> but returns a L<DBIx::Class::ResultSetColumn> object.
1574 This can be very handy for subqueries:
1576 ->search( { amount => $some_rs->count_rs->as_query } )
1578 As with regular resultsets the SQL query will be executed only after
1579 the resultset is accessed via L</next> or L</all>. That would return
1580 the same single value obtainable via L</count>.
1586 return $self->search(@_)->count_rs if @_;
1588 # this may look like a lack of abstraction (count() does about the same)
1589 # but in fact an _rs *must* use a subquery for the limits, as the
1590 # software based limiting can not be ported if this $rs is to be used
1591 # in a subquery itself (i.e. ->as_query)
1592 if ($self->_has_resolved_attr (qw/collapse group_by offset rows/)) {
1593 return $self->_count_subq_rs($self->{_attrs});
1596 return $self->_count_rs($self->{_attrs});
1601 # returns a ResultSetColumn object tied to the count query
1604 my ($self, $attrs) = @_;
1606 my $rsrc = $self->result_source;
1608 my $tmp_attrs = { %$attrs };
1609 # take off any limits, record_filter is cdbi, and no point of ordering nor locking a count
1610 delete @{$tmp_attrs}{qw/rows offset order_by record_filter for/};
1612 # overwrite the selector (supplied by the storage)
1613 $rsrc->resultset_class->new($rsrc, {
1615 select => $rsrc->storage->_count_select ($rsrc, $attrs),
1617 })->get_column ('count');
1621 # same as above but uses a subquery
1623 sub _count_subq_rs {
1624 my ($self, $attrs) = @_;
1626 my $rsrc = $self->result_source;
1628 my $sub_attrs = { %$attrs };
1629 # extra selectors do not go in the subquery and there is no point of ordering it, nor locking it
1630 delete @{$sub_attrs}{qw/collapse columns as select order_by for/};
1632 # if we multi-prefetch we group_by something unique, as this is what we would
1633 # get out of the rs via ->next/->all. We *DO WANT* to clobber old group_by regardless
1634 if ( $attrs->{collapse} ) {
1635 $sub_attrs->{group_by} = [ map { "$attrs->{alias}.$_" } @{
1636 $rsrc->_identifying_column_set || $self->throw_exception(
1637 'Unable to construct a unique group_by criteria properly collapsing the '
1638 . 'has_many prefetch before count()'
1643 # Calculate subquery selector
1644 if (my $g = $sub_attrs->{group_by}) {
1646 my $sql_maker = $rsrc->storage->sql_maker;
1648 # necessary as the group_by may refer to aliased functions
1650 for my $sel (@{$attrs->{select}}) {
1651 $sel_index->{$sel->{-as}} = $sel
1652 if (ref $sel eq 'HASH' and $sel->{-as});
1655 # anything from the original select mentioned on the group-by needs to make it to the inner selector
1656 # also look for named aggregates referred in the having clause
1657 # having often contains scalarrefs - thus parse it out entirely
1659 if ($attrs->{having}) {
1660 local $sql_maker->{having_bind};
1661 local $sql_maker->{quote_char} = $sql_maker->{quote_char};
1662 local $sql_maker->{name_sep} = $sql_maker->{name_sep};
1663 unless (defined $sql_maker->{quote_char} and length $sql_maker->{quote_char}) {
1664 $sql_maker->{quote_char} = [ "\x00", "\xFF" ];
1665 # if we don't unset it we screw up retarded but unfortunately working
1666 # 'MAX(foo.bar)' => { '>', 3 }
1667 $sql_maker->{name_sep} = '';
1670 my ($lquote, $rquote, $sep) = map { quotemeta $_ } ($sql_maker->_quote_chars, $sql_maker->name_sep);
1672 my $having_sql = $sql_maker->_parse_rs_attrs ({ having => $attrs->{having} });
1675 # search for both a proper quoted qualified string, for a naive unquoted scalarref
1676 # and if all fails for an utterly naive quoted scalar-with-function
1677 while ($having_sql =~ /
1678 $rquote $sep $lquote (.+?) $rquote
1680 [\s,] \w+ \. (\w+) [\s,]
1682 [\s,] $lquote (.+?) $rquote [\s,]
1684 my $part = $1 || $2 || $3; # one of them matched if we got here
1685 unless ($seen_having{$part}++) {
1692 my $colpiece = $sel_index->{$_} || $_;
1694 # unqualify join-based group_by's. Arcane but possible query
1695 # also horrible horrible hack to alias a column (not a func.)
1696 # (probably need to introduce SQLA syntax)
1697 if ($colpiece =~ /\./ && $colpiece !~ /^$attrs->{alias}\./) {
1700 $colpiece = \ sprintf ('%s AS %s', map { $sql_maker->_quote ($_) } ($colpiece, $as) );
1702 push @{$sub_attrs->{select}}, $colpiece;
1706 my @pcols = map { "$attrs->{alias}.$_" } ($rsrc->primary_columns);
1707 $sub_attrs->{select} = @pcols ? \@pcols : [ 1 ];
1710 return $rsrc->resultset_class
1711 ->new ($rsrc, $sub_attrs)
1713 ->search ({}, { columns => { count => $rsrc->storage->_count_select ($rsrc, $attrs) } })
1714 ->get_column ('count');
1718 =head2 count_literal
1720 B<CAVEAT>: C<count_literal> is provided for Class::DBI compatibility and
1721 should only be used in that context. See L</search_literal> for further info.
1725 =item Arguments: $sql_fragment, @standalone_bind_values
1727 =item Return Value: $count
1731 Counts the results in a literal query. Equivalent to calling L</search_literal>
1732 with the passed arguments, then L</count>.
1736 sub count_literal { shift->search_literal(@_)->count; }
1742 =item Arguments: none
1744 =item Return Value: L<@result_objs|DBIx::Class::Manual::ResultClass>
1748 Returns all elements in the resultset.
1755 $self->throw_exception("all() doesn't take any arguments, you probably wanted ->search(...)->all()");
1758 delete @{$self}{qw/_stashed_rows _stashed_results/};
1760 if (my $c = $self->get_cache) {
1764 $self->cursor->reset;
1766 my $objs = $self->_construct_results('fetch_all') || [];
1768 $self->set_cache($objs) if $self->{attrs}{cache};
1777 =item Arguments: none
1779 =item Return Value: $self
1783 Resets the resultset's cursor, so you can iterate through the elements again.
1784 Implicitly resets the storage cursor, so a subsequent L</next> will trigger
1792 delete @{$self}{qw/_stashed_rows _stashed_results/};
1793 $self->{all_cache_position} = 0;
1794 $self->cursor->reset;
1802 =item Arguments: none
1804 =item Return Value: L<$result|DBIx::Class::Manual::ResultClass> | undef
1808 L<Resets|/reset> the resultset (causing a fresh query to storage) and returns
1809 an object for the first result (or C<undef> if the resultset is empty).
1814 return $_[0]->reset->next;
1820 # Determines whether and what type of subquery is required for the $rs operation.
1821 # If grouping is necessary either supplies its own, or verifies the current one
1822 # After all is done delegates to the proper storage method.
1824 sub _rs_update_delete {
1825 my ($self, $op, $values) = @_;
1827 my $rsrc = $self->result_source;
1828 my $storage = $rsrc->schema->storage;
1830 my $attrs = { %{$self->_resolved_attrs} };
1832 my $join_classifications;
1833 my ($existing_group_by) = delete @{$attrs}{qw(group_by _grouped_by_distinct)};
1835 # do we need a subquery for any reason?
1837 defined $existing_group_by
1839 # if {from} is unparseable wrap a subq
1840 ref($attrs->{from}) ne 'ARRAY'
1842 # limits call for a subq
1843 $self->_has_resolved_attr(qw/rows offset/)
1846 # simplify the joinmap, so we can further decide if a subq is necessary
1847 if (!$needs_subq and @{$attrs->{from}} > 1) {
1849 ($attrs->{from}, $join_classifications) =
1850 $storage->_prune_unused_joins ($attrs);
1852 # any non-pruneable non-local restricting joins imply subq
1853 $needs_subq = defined List::Util::first { $_ ne $attrs->{alias} } keys %{ $join_classifications->{restricting} || {} };
1856 # check if the head is composite (by now all joins are thrown out unless $needs_subq)
1858 (ref $attrs->{from}[0]) ne 'HASH'
1860 ref $attrs->{from}[0]{ $attrs->{from}[0]{-alias} }
1864 # do we need anything like a subquery?
1865 if (! $needs_subq) {
1866 # Most databases do not allow aliasing of tables in UPDATE/DELETE. Thus
1867 # a condition containing 'me' or other table prefixes will not work
1868 # at all. Tell SQLMaker to dequalify idents via a gross hack.
1870 my $sqla = $rsrc->storage->sql_maker;
1871 local $sqla->{_dequalify_idents} = 1;
1872 \[ $sqla->_recurse_where($self->{cond}) ];
1876 # we got this far - means it is time to wrap a subquery
1877 my $idcols = $rsrc->_identifying_column_set || $self->throw_exception(
1879 "Unable to perform complex resultset %s() without an identifying set of columns on source '%s'",
1885 # make a new $rs selecting only the PKs (that's all we really need for the subq)
1886 delete $attrs->{$_} for qw/select as collapse/;
1887 $attrs->{columns} = [ map { "$attrs->{alias}.$_" } @$idcols ];
1889 # this will be consumed by the pruner waaaaay down the stack
1890 $attrs->{_force_prune_multiplying_joins} = 1;
1892 my $subrs = (ref $self)->new($rsrc, $attrs);
1894 if (@$idcols == 1) {
1895 $cond = { $idcols->[0] => { -in => $subrs->as_query } };
1897 elsif ($storage->_use_multicolumn_in) {
1898 # no syntax for calling this properly yet
1899 # !!! EXPERIMENTAL API !!! WILL CHANGE !!!
1900 $cond = $storage->sql_maker->_where_op_multicolumn_in (
1901 $idcols, # how do I convey a list of idents...? can binds reside on lhs?
1906 # if all else fails - get all primary keys and operate over a ORed set
1907 # wrap in a transaction for consistency
1908 # this is where the group_by/multiplication starts to matter
1912 # we do not need to check pre-multipliers, since if the premulti is there, its
1913 # parent (who is multi) will be there too
1914 keys %{ $join_classifications->{multiplying} || {} }
1916 # make sure if there is a supplied group_by it matches the columns compiled above
1917 # perfectly. Anything else can not be sanely executed on most databases so croak
1918 # right then and there
1919 if ($existing_group_by) {
1920 my @current_group_by = map
1921 { $_ =~ /\./ ? $_ : "$attrs->{alias}.$_" }
1926 join ("\x00", sort @current_group_by)
1928 join ("\x00", sort @{$attrs->{columns}} )
1930 $self->throw_exception (
1931 "You have just attempted a $op operation on a resultset which does group_by"
1932 . ' on columns other than the primary keys, while DBIC internally needs to retrieve'
1933 . ' the primary keys in a subselect. All sane RDBMS engines do not support this'
1934 . ' kind of queries. Please retry the operation with a modified group_by or'
1935 . ' without using one at all.'
1940 $subrs = $subrs->search({}, { group_by => $attrs->{columns} });
1943 $guard = $storage->txn_scope_guard;
1946 for my $row ($subrs->cursor->all) {
1948 { $idcols->[$_] => $row->[$_] }
1955 my $res = $storage->$op (
1957 $op eq 'update' ? $values : (),
1961 $guard->commit if $guard;
1970 =item Arguments: \%values
1972 =item Return Value: $underlying_storage_rv
1976 Sets the specified columns in the resultset to the supplied values in a
1977 single query. Note that this will not run any accessor/set_column/update
1978 triggers, nor will it update any result object instances derived from this
1979 resultset (this includes the contents of the L<resultset cache|/set_cache>
1980 if any). See L</update_all> if you need to execute any on-update
1981 triggers or cascades defined either by you or a
1982 L<result component|DBIx::Class::Manual::Component/WHAT IS A COMPONENT>.
1984 The return value is a pass through of what the underlying
1985 storage backend returned, and may vary. See L<DBI/execute> for the most
1990 Note that L</update> does not process/deflate any of the values passed in.
1991 This is unlike the corresponding L<DBIx::Class::Row/update>. The user must
1992 ensure manually that any value passed to this method will stringify to
1993 something the RDBMS knows how to deal with. A notable example is the
1994 handling of L<DateTime> objects, for more info see:
1995 L<DBIx::Class::Manual::Cookbook/Formatting DateTime objects in queries>.
2000 my ($self, $values) = @_;
2001 $self->throw_exception('Values for update must be a hash')
2002 unless ref $values eq 'HASH';
2004 return $self->_rs_update_delete ('update', $values);
2011 =item Arguments: \%values
2013 =item Return Value: 1
2017 Fetches all objects and updates them one at a time via
2018 L<DBIx::Class::Row/update>. Note that C<update_all> will run DBIC defined
2019 triggers, while L</update> will not.
2024 my ($self, $values) = @_;
2025 $self->throw_exception('Values for update_all must be a hash')
2026 unless ref $values eq 'HASH';
2028 my $guard = $self->result_source->schema->txn_scope_guard;
2029 $_->update({%$values}) for $self->all; # shallow copy - update will mangle it
2038 =item Arguments: none
2040 =item Return Value: $underlying_storage_rv
2044 Deletes the rows matching this resultset in a single query. Note that this
2045 will not run any delete triggers, nor will it alter the
2046 L<in_storage|DBIx::Class::Row/in_storage> status of any result object instances
2047 derived from this resultset (this includes the contents of the
2048 L<resultset cache|/set_cache> if any). See L</delete_all> if you need to
2049 execute any on-delete triggers or cascades defined either by you or a
2050 L<result component|DBIx::Class::Manual::Component/WHAT IS A COMPONENT>.
2052 The return value is a pass through of what the underlying storage backend
2053 returned, and may vary. See L<DBI/execute> for the most common case.
2059 $self->throw_exception('delete does not accept any arguments')
2062 return $self->_rs_update_delete ('delete');
2069 =item Arguments: none
2071 =item Return Value: 1
2075 Fetches all objects and deletes them one at a time via
2076 L<DBIx::Class::Row/delete>. Note that C<delete_all> will run DBIC defined
2077 triggers, while L</delete> will not.
2083 $self->throw_exception('delete_all does not accept any arguments')
2086 my $guard = $self->result_source->schema->txn_scope_guard;
2087 $_->delete for $self->all;
2096 =item Arguments: [ \@column_list, \@row_values+ ] | [ \%col_data+ ]
2098 =item Return Value: L<\@result_objects|DBIx::Class::Manual::ResultClass> (scalar context) | L<@result_objects|DBIx::Class::Manual::ResultClass> (list context)
2102 Accepts either an arrayref of hashrefs or alternatively an arrayref of
2109 The context of this method call has an important effect on what is
2110 submitted to storage. In void context data is fed directly to fastpath
2111 insertion routines provided by the underlying storage (most often
2112 L<DBI/execute_for_fetch>), bypassing the L<new|DBIx::Class::Row/new> and
2113 L<insert|DBIx::Class::Row/insert> calls on the
2114 L<Result|DBIx::Class::Manual::ResultClass> class, including any
2115 augmentation of these methods provided by components. For example if you
2116 are using something like L<DBIx::Class::UUIDColumns> to create primary
2117 keys for you, you will find that your PKs are empty. In this case you
2118 will have to explicitly force scalar or list context in order to create
2123 In non-void (scalar or list) context, this method is simply a wrapper
2124 for L</create>. Depending on list or scalar context either a list of
2125 L<Result|DBIx::Class::Manual::ResultClass> objects or an arrayref
2126 containing these objects is returned.
2128 When supplying data in "arrayref of arrayrefs" invocation style, the
2129 first element should be a list of column names and each subsequent
2130 element should be a data value in the earlier specified column order.
2133 $schema->resultset("Artist")->populate([
2134 [ qw( artistid name ) ],
2135 [ 100, 'A Formally Unknown Singer' ],
2136 [ 101, 'A singer that jumped the shark two albums ago' ],
2137 [ 102, 'An actually cool singer' ],
2140 For the arrayref of hashrefs style each hashref should be a structure
2141 suitable for passing to L</create>. Multi-create is also permitted with
2144 $schema->resultset("Artist")->populate([
2145 { artistid => 4, name => 'Manufactured Crap', cds => [
2146 { title => 'My First CD', year => 2006 },
2147 { title => 'Yet More Tweeny-Pop crap', year => 2007 },
2150 { artistid => 5, name => 'Angsty-Whiny Girl', cds => [
2151 { title => 'My parents sold me to a record company', year => 2005 },
2152 { title => 'Why Am I So Ugly?', year => 2006 },
2153 { title => 'I Got Surgery and am now Popular', year => 2007 }
2158 If you attempt a void-context multi-create as in the example above (each
2159 Artist also has the related list of CDs), and B<do not> supply the
2160 necessary autoinc foreign key information, this method will proxy to the
2161 less efficient L</create>, and then throw the Result objects away. In this
2162 case there are obviously no benefits to using this method over L</create>.
2169 # cruft placed in standalone method
2170 my $data = $self->_normalize_populate_args(@_);
2172 return unless @$data;
2174 if(defined wantarray) {
2175 my @created = map { $self->create($_) } @$data;
2176 return wantarray ? @created : \@created;
2179 my $first = $data->[0];
2181 # if a column is a registered relationship, and is a non-blessed hash/array, consider
2182 # it relationship data
2183 my (@rels, @columns);
2184 my $rsrc = $self->result_source;
2185 my $rels = { map { $_ => $rsrc->relationship_info($_) } $rsrc->relationships };
2186 for (keys %$first) {
2187 my $ref = ref $first->{$_};
2188 $rels->{$_} && ($ref eq 'ARRAY' or $ref eq 'HASH')
2194 my @pks = $rsrc->primary_columns;
2196 ## do the belongs_to relationships
2197 foreach my $index (0..$#$data) {
2199 # delegate to create() for any dataset without primary keys with specified relationships
2200 if (grep { !defined $data->[$index]->{$_} } @pks ) {
2202 if (grep { ref $data->[$index]{$r} eq $_ } qw/HASH ARRAY/) { # a related set must be a HASH or AoH
2203 my @ret = $self->populate($data);
2209 foreach my $rel (@rels) {
2210 next unless ref $data->[$index]->{$rel} eq "HASH";
2211 my $result = $self->related_resultset($rel)->create($data->[$index]->{$rel});
2212 my ($reverse_relname, $reverse_relinfo) = %{$rsrc->reverse_relationship_info($rel)};
2213 my $related = $result->result_source->_resolve_condition(
2214 $reverse_relinfo->{cond},
2220 delete $data->[$index]->{$rel};
2221 $data->[$index] = {%{$data->[$index]}, %$related};
2223 push @columns, keys %$related if $index == 0;
2227 ## inherit the data locked in the conditions of the resultset
2228 my ($rs_data) = $self->_merge_with_rscond({});
2229 delete @{$rs_data}{@columns};
2231 ## do bulk insert on current row
2232 $rsrc->storage->insert_bulk(
2234 [@columns, keys %$rs_data],
2235 [ map { [ @$_{@columns}, values %$rs_data ] } @$data ],
2238 ## do the has_many relationships
2239 foreach my $item (@$data) {
2243 foreach my $rel (@rels) {
2244 next unless ref $item->{$rel} eq "ARRAY" && @{ $item->{$rel} };
2246 $main_row ||= $self->new_result({map { $_ => $item->{$_} } @pks});
2248 my $child = $main_row->$rel;
2250 my $related = $child->result_source->_resolve_condition(
2251 $rels->{$rel}{cond},
2257 my @rows_to_add = ref $item->{$rel} eq 'ARRAY' ? @{$item->{$rel}} : ($item->{$rel});
2258 my @populate = map { {%$_, %$related} } @rows_to_add;
2260 $child->populate( \@populate );
2267 # populate() arguments went over several incarnations
2268 # What we ultimately support is AoH
2269 sub _normalize_populate_args {
2270 my ($self, $arg) = @_;
2272 if (ref $arg eq 'ARRAY') {
2276 elsif (ref $arg->[0] eq 'HASH') {
2279 elsif (ref $arg->[0] eq 'ARRAY') {
2281 my @colnames = @{$arg->[0]};
2282 foreach my $values (@{$arg}[1 .. $#$arg]) {
2283 push @ret, { map { $colnames[$_] => $values->[$_] } (0 .. $#colnames) };
2289 $self->throw_exception('Populate expects an arrayref of hashrefs or arrayref of arrayrefs');
2296 =item Arguments: none
2298 =item Return Value: L<$pager|Data::Page>
2302 Returns a L<Data::Page> object for the current resultset. Only makes
2303 sense for queries with a C<page> attribute.
2305 To get the full count of entries for a paged resultset, call
2306 C<total_entries> on the L<Data::Page> object.
2313 return $self->{pager} if $self->{pager};
2315 my $attrs = $self->{attrs};
2316 if (!defined $attrs->{page}) {
2317 $self->throw_exception("Can't create pager for non-paged rs");
2319 elsif ($attrs->{page} <= 0) {
2320 $self->throw_exception('Invalid page number (page-numbers are 1-based)');
2322 $attrs->{rows} ||= 10;
2324 # throw away the paging flags and re-run the count (possibly
2325 # with a subselect) to get the real total count
2326 my $count_attrs = { %$attrs };
2327 delete @{$count_attrs}{qw/rows offset page pager/};
2329 my $total_rs = (ref $self)->new($self->result_source, $count_attrs);
2331 require DBIx::Class::ResultSet::Pager;
2332 return $self->{pager} = DBIx::Class::ResultSet::Pager->new(
2333 sub { $total_rs->count }, #lazy-get the total
2335 $self->{attrs}{page},
2343 =item Arguments: $page_number
2345 =item Return Value: L<$resultset|/search>
2349 Returns a resultset for the $page_number page of the resultset on which page
2350 is called, where each page contains a number of rows equal to the 'rows'
2351 attribute set on the resultset (10 by default).
2356 my ($self, $page) = @_;
2357 return (ref $self)->new($self->result_source, { %{$self->{attrs}}, page => $page });
2364 =item Arguments: \%col_data
2366 =item Return Value: L<$result|DBIx::Class::Manual::ResultClass>
2370 Creates a new result object in the resultset's result class and returns
2371 it. The row is not inserted into the database at this point, call
2372 L<DBIx::Class::Row/insert> to do that. Calling L<DBIx::Class::Row/in_storage>
2373 will tell you whether the result object has been inserted or not.
2375 Passes the hashref of input on to L<DBIx::Class::Row/new>.
2380 my ($self, $values) = @_;
2382 $self->throw_exception( "new_result takes only one argument - a hashref of values" )
2385 $self->throw_exception( "new_result expects a hashref" )
2386 unless (ref $values eq 'HASH');
2388 my ($merged_cond, $cols_from_relations) = $self->_merge_with_rscond($values);
2390 my $new = $self->result_class->new({
2392 ( @$cols_from_relations
2393 ? (-cols_from_relations => $cols_from_relations)
2396 -result_source => $self->result_source, # DO NOT REMOVE THIS, REQUIRED
2400 reftype($new) eq 'HASH'
2406 carp_unique (sprintf (
2407 "%s->new returned a blessed empty hashref - a strong indicator something is wrong with its inheritance chain",
2408 $self->result_class,
2415 # _merge_with_rscond
2417 # Takes a simple hash of K/V data and returns its copy merged with the
2418 # condition already present on the resultset. Additionally returns an
2419 # arrayref of value/condition names, which were inferred from related
2420 # objects (this is needed for in-memory related objects)
2421 sub _merge_with_rscond {
2422 my ($self, $data) = @_;
2424 my (%new_data, @cols_from_relations);
2426 my $alias = $self->{attrs}{alias};
2428 if (! defined $self->{cond}) {
2429 # just massage $data below
2431 elsif ($self->{cond} eq $DBIx::Class::ResultSource::UNRESOLVABLE_CONDITION) {
2432 %new_data = %{ $self->{attrs}{related_objects} || {} }; # nothing might have been inserted yet
2433 @cols_from_relations = keys %new_data;
2435 elsif (ref $self->{cond} ne 'HASH') {
2436 $self->throw_exception(
2437 "Can't abstract implicit construct, resultset condition not a hash"
2441 if ($self->{cond}) {
2442 my $implied = $self->_remove_alias(
2443 $self->result_source->schema->storage->_collapse_cond($self->{cond}),
2447 for my $c (keys %$implied) {
2448 my $v = $implied->{$c};
2452 overload::Method($v, '""')
2457 ref $v eq 'HASH' and keys %$v == 1 and exists $v->{'='} and (
2458 ref $v->{'='} eq 'SCALAR'
2460 ( ref $v->{'='} eq 'REF' and ref ${$v->{'='}} eq 'ARRAY' )
2463 $new_data{$c} = $v->{'='};
2469 # precedence must be given to passed values over values inherited from
2470 # the cond, so the order here is important.
2473 %{ $self->_remove_alias($data, $alias) },
2476 return (\%new_data, \@cols_from_relations);
2479 # _has_resolved_attr
2481 # determines if the resultset defines at least one
2482 # of the attributes supplied
2484 # used to determine if a subquery is necessary
2486 # supports some virtual attributes:
2488 # This will scan for any joins being present on the resultset.
2489 # It is not a mere key-search but a deep inspection of {from}
2492 sub _has_resolved_attr {
2493 my ($self, @attr_names) = @_;
2495 my $attrs = $self->_resolved_attrs;
2499 for my $n (@attr_names) {
2500 if (grep { $n eq $_ } (qw/-join/) ) {
2501 $extra_checks{$n}++;
2505 my $attr = $attrs->{$n};
2507 next if not defined $attr;
2509 if (ref $attr eq 'HASH') {
2510 return 1 if keys %$attr;
2512 elsif (ref $attr eq 'ARRAY') {
2520 # a resolved join is expressed as a multi-level from
2522 $extra_checks{-join}
2524 ref $attrs->{from} eq 'ARRAY'
2526 @{$attrs->{from}} > 1
2534 # Remove the specified alias from the specified query hash. A copy is made so
2535 # the original query is not modified.
2538 my ($self, $query, $alias) = @_;
2540 my %orig = %{ $query || {} };
2543 foreach my $key (keys %orig) {
2545 $unaliased{$key} = $orig{$key};
2548 $unaliased{$1} = $orig{$key}
2549 if $key =~ m/^(?:\Q$alias\E\.)?([^.]+)$/;
2559 =item Arguments: none
2561 =item Return Value: \[ $sql, L<@bind_values|/DBIC BIND VALUES> ]
2565 Returns the SQL query and bind vars associated with the invocant.
2567 This is generally used as the RHS for a subquery.
2574 my $attrs = { %{ $self->_resolved_attrs } };
2576 my $aq = $self->result_source->storage->_select_args_to_query (
2577 $attrs->{from}, $attrs->{select}, $attrs->{where}, $attrs
2587 =item Arguments: \%col_data, { key => $unique_constraint, L<%attrs|/ATTRIBUTES> }?
2589 =item Return Value: L<$result|DBIx::Class::Manual::ResultClass>
2593 my $artist = $schema->resultset('Artist')->find_or_new(
2594 { artist => 'fred' }, { key => 'artists' });
2596 $cd->cd_to_producer->find_or_new({ producer => $producer },
2597 { key => 'primary' });
2599 Find an existing record from this resultset using L</find>. if none exists,
2600 instantiate a new result object and return it. The object will not be saved
2601 into your storage until you call L<DBIx::Class::Row/insert> on it.
2603 You most likely want this method when looking for existing rows using a unique
2604 constraint that is not the primary key, or looking for related rows.
2606 If you want objects to be saved immediately, use L</find_or_create> instead.
2608 B<Note>: Make sure to read the documentation of L</find> and understand the
2609 significance of the C<key> attribute, as its lack may skew your search, and
2610 subsequently result in spurious new objects.
2612 B<Note>: Take care when using C<find_or_new> with a table having
2613 columns with default values that you intend to be automatically
2614 supplied by the database (e.g. an auto_increment primary key column).
2615 In normal usage, the value of such columns should NOT be included at
2616 all in the call to C<find_or_new>, even when set to C<undef>.
2622 my $attrs = (@_ > 1 && ref $_[$#_] eq 'HASH' ? pop(@_) : {});
2623 my $hash = ref $_[0] eq 'HASH' ? shift : {@_};
2624 if (keys %$hash and my $row = $self->find($hash, $attrs) ) {
2627 return $self->new_result($hash);
2634 =item Arguments: \%col_data
2636 =item Return Value: L<$result|DBIx::Class::Manual::ResultClass>
2640 Attempt to create a single new row or a row with multiple related rows
2641 in the table represented by the resultset (and related tables). This
2642 will not check for duplicate rows before inserting, use
2643 L</find_or_create> to do that.
2645 To create one row for this resultset, pass a hashref of key/value
2646 pairs representing the columns of the table and the values you wish to
2647 store. If the appropriate relationships are set up, foreign key fields
2648 can also be passed an object representing the foreign row, and the
2649 value will be set to its primary key.
2651 To create related objects, pass a hashref of related-object column values
2652 B<keyed on the relationship name>. If the relationship is of type C<multi>
2653 (L<DBIx::Class::Relationship/has_many>) - pass an arrayref of hashrefs.
2654 The process will correctly identify columns holding foreign keys, and will
2655 transparently populate them from the keys of the corresponding relation.
2656 This can be applied recursively, and will work correctly for a structure
2657 with an arbitrary depth and width, as long as the relationships actually
2658 exists and the correct column data has been supplied.
2660 Instead of hashrefs of plain related data (key/value pairs), you may
2661 also pass new or inserted objects. New objects (not inserted yet, see
2662 L</new_result>), will be inserted into their appropriate tables.
2664 Effectively a shortcut for C<< ->new_result(\%col_data)->insert >>.
2666 Example of creating a new row.
2668 $person_rs->create({
2669 name=>"Some Person",
2670 email=>"somebody@someplace.com"
2673 Example of creating a new row and also creating rows in a related C<has_many>
2674 or C<has_one> resultset. Note Arrayref.
2677 { artistid => 4, name => 'Manufactured Crap', cds => [
2678 { title => 'My First CD', year => 2006 },
2679 { title => 'Yet More Tweeny-Pop crap', year => 2007 },
2684 Example of creating a new row and also creating a row in a related
2685 C<belongs_to> resultset. Note Hashref.
2688 title=>"Music for Silly Walks",
2691 name=>"Silly Musician",
2699 When subclassing ResultSet never attempt to override this method. Since
2700 it is a simple shortcut for C<< $self->new_result($attrs)->insert >>, a
2701 lot of the internals simply never call it, so your override will be
2702 bypassed more often than not. Override either L<DBIx::Class::Row/new>
2703 or L<DBIx::Class::Row/insert> depending on how early in the
2704 L</create> process you need to intervene. See also warning pertaining to
2712 my ($self, $col_data) = @_;
2713 $self->throw_exception( "create needs a hashref" )
2714 unless ref $col_data eq 'HASH';
2715 return $self->new_result($col_data)->insert;
2718 =head2 find_or_create
2722 =item Arguments: \%col_data, { key => $unique_constraint, L<%attrs|/ATTRIBUTES> }?
2724 =item Return Value: L<$result|DBIx::Class::Manual::ResultClass>
2728 $cd->cd_to_producer->find_or_create({ producer => $producer },
2729 { key => 'primary' });
2731 Tries to find a record based on its primary key or unique constraints; if none
2732 is found, creates one and returns that instead.
2734 my $cd = $schema->resultset('CD')->find_or_create({
2736 artist => 'Massive Attack',
2737 title => 'Mezzanine',
2741 Also takes an optional C<key> attribute, to search by a specific key or unique
2742 constraint. For example:
2744 my $cd = $schema->resultset('CD')->find_or_create(
2746 artist => 'Massive Attack',
2747 title => 'Mezzanine',
2749 { key => 'cd_artist_title' }
2752 B<Note>: Make sure to read the documentation of L</find> and understand the
2753 significance of the C<key> attribute, as its lack may skew your search, and
2754 subsequently result in spurious row creation.
2756 B<Note>: Because find_or_create() reads from the database and then
2757 possibly inserts based on the result, this method is subject to a race
2758 condition. Another process could create a record in the table after
2759 the find has completed and before the create has started. To avoid
2760 this problem, use find_or_create() inside a transaction.
2762 B<Note>: Take care when using C<find_or_create> with a table having
2763 columns with default values that you intend to be automatically
2764 supplied by the database (e.g. an auto_increment primary key column).
2765 In normal usage, the value of such columns should NOT be included at
2766 all in the call to C<find_or_create>, even when set to C<undef>.
2768 See also L</find> and L</update_or_create>. For information on how to declare
2769 unique constraints, see L<DBIx::Class::ResultSource/add_unique_constraint>.
2771 If you need to know if an existing row was found or a new one created use
2772 L</find_or_new> and L<DBIx::Class::Row/in_storage> instead. Don't forget
2773 to call L<DBIx::Class::Row/insert> to save the newly created row to the
2776 my $cd = $schema->resultset('CD')->find_or_new({
2778 artist => 'Massive Attack',
2779 title => 'Mezzanine',
2783 if( !$cd->in_storage ) {
2790 sub find_or_create {
2792 my $attrs = (@_ > 1 && ref $_[$#_] eq 'HASH' ? pop(@_) : {});
2793 my $hash = ref $_[0] eq 'HASH' ? shift : {@_};
2794 if (keys %$hash and my $row = $self->find($hash, $attrs) ) {
2797 return $self->create($hash);
2800 =head2 update_or_create
2804 =item Arguments: \%col_data, { key => $unique_constraint, L<%attrs|/ATTRIBUTES> }?
2806 =item Return Value: L<$result|DBIx::Class::Manual::ResultClass>
2810 $resultset->update_or_create({ col => $val, ... });
2812 Like L</find_or_create>, but if a row is found it is immediately updated via
2813 C<< $found_row->update (\%col_data) >>.
2816 Takes an optional C<key> attribute to search on a specific unique constraint.
2819 # In your application
2820 my $cd = $schema->resultset('CD')->update_or_create(
2822 artist => 'Massive Attack',
2823 title => 'Mezzanine',
2826 { key => 'cd_artist_title' }
2829 $cd->cd_to_producer->update_or_create({
2830 producer => $producer,
2836 B<Note>: Make sure to read the documentation of L</find> and understand the
2837 significance of the C<key> attribute, as its lack may skew your search, and
2838 subsequently result in spurious row creation.
2840 B<Note>: Take care when using C<update_or_create> with a table having
2841 columns with default values that you intend to be automatically
2842 supplied by the database (e.g. an auto_increment primary key column).
2843 In normal usage, the value of such columns should NOT be included at
2844 all in the call to C<update_or_create>, even when set to C<undef>.
2846 See also L</find> and L</find_or_create>. For information on how to declare
2847 unique constraints, see L<DBIx::Class::ResultSource/add_unique_constraint>.
2849 If you need to know if an existing row was updated or a new one created use
2850 L</update_or_new> and L<DBIx::Class::Row/in_storage> instead. Don't forget
2851 to call L<DBIx::Class::Row/insert> to save the newly created row to the
2856 sub update_or_create {
2858 my $attrs = (@_ > 1 && ref $_[$#_] eq 'HASH' ? pop(@_) : {});
2859 my $cond = ref $_[0] eq 'HASH' ? shift : {@_};
2861 my $row = $self->find($cond, $attrs);
2863 $row->update($cond);
2867 return $self->create($cond);
2870 =head2 update_or_new
2874 =item Arguments: \%col_data, { key => $unique_constraint, L<%attrs|/ATTRIBUTES> }?
2876 =item Return Value: L<$result|DBIx::Class::Manual::ResultClass>
2880 $resultset->update_or_new({ col => $val, ... });
2882 Like L</find_or_new> but if a row is found it is immediately updated via
2883 C<< $found_row->update (\%col_data) >>.
2887 # In your application
2888 my $cd = $schema->resultset('CD')->update_or_new(
2890 artist => 'Massive Attack',
2891 title => 'Mezzanine',
2894 { key => 'cd_artist_title' }
2897 if ($cd->in_storage) {
2898 # the cd was updated
2901 # the cd is not yet in the database, let's insert it
2905 B<Note>: Make sure to read the documentation of L</find> and understand the
2906 significance of the C<key> attribute, as its lack may skew your search, and
2907 subsequently result in spurious new objects.
2909 B<Note>: Take care when using C<update_or_new> with a table having
2910 columns with default values that you intend to be automatically
2911 supplied by the database (e.g. an auto_increment primary key column).
2912 In normal usage, the value of such columns should NOT be included at
2913 all in the call to C<update_or_new>, even when set to C<undef>.
2915 See also L</find>, L</find_or_create> and L</find_or_new>.
2921 my $attrs = ( @_ > 1 && ref $_[$#_] eq 'HASH' ? pop(@_) : {} );
2922 my $cond = ref $_[0] eq 'HASH' ? shift : {@_};
2924 my $row = $self->find( $cond, $attrs );
2925 if ( defined $row ) {
2926 $row->update($cond);
2930 return $self->new_result($cond);
2937 =item Arguments: none
2939 =item Return Value: L<\@result_objs|DBIx::Class::Manual::ResultClass> | undef
2943 Gets the contents of the cache for the resultset, if the cache is set.
2945 The cache is populated either by using the L</prefetch> attribute to
2946 L</search> or by calling L</set_cache>.
2958 =item Arguments: L<\@result_objs|DBIx::Class::Manual::ResultClass>
2960 =item Return Value: L<\@result_objs|DBIx::Class::Manual::ResultClass>
2964 Sets the contents of the cache for the resultset. Expects an arrayref
2965 of objects of the same class as those produced by the resultset. Note that
2966 if the cache is set, the resultset will return the cached objects rather
2967 than re-querying the database even if the cache attr is not set.
2969 The contents of the cache can also be populated by using the
2970 L</prefetch> attribute to L</search>.
2975 my ( $self, $data ) = @_;
2976 $self->throw_exception("set_cache requires an arrayref")
2977 if defined($data) && (ref $data ne 'ARRAY');
2978 $self->{all_cache} = $data;
2985 =item Arguments: none
2987 =item Return Value: undef
2991 Clears the cache for the resultset.
2996 shift->set_cache(undef);
3003 =item Arguments: none
3005 =item Return Value: true, if the resultset has been paginated
3013 return !!$self->{attrs}{page};
3020 =item Arguments: none
3022 =item Return Value: true, if the resultset has been ordered with C<order_by>.
3030 return scalar $self->result_source->storage->_extract_order_criteria($self->{attrs}{order_by});
3033 =head2 related_resultset
3037 =item Arguments: $rel_name
3039 =item Return Value: L<$resultset|/search>
3043 Returns a related resultset for the supplied relationship name.
3045 $artist_rs = $schema->resultset('CD')->related_resultset('Artist');
3049 sub related_resultset {
3050 my ($self, $rel) = @_;
3052 return $self->{related_resultsets}{$rel}
3053 if defined $self->{related_resultsets}{$rel};
3055 return $self->{related_resultsets}{$rel} = do {
3056 my $rsrc = $self->result_source;
3057 my $rel_info = $rsrc->relationship_info($rel);
3059 $self->throw_exception(
3060 "search_related: result source '" . $rsrc->source_name .
3061 "' has no such relationship $rel")
3064 my $attrs = $self->_chain_relationship($rel);
3066 my $join_count = $attrs->{seen_join}{$rel};
3068 my $alias = $self->result_source->storage
3069 ->relname_to_table_alias($rel, $join_count);
3071 # since this is search_related, and we already slid the select window inwards
3072 # (the select/as attrs were deleted in the beginning), we need to flip all
3073 # left joins to inner, so we get the expected results
3074 # read the comment on top of the actual function to see what this does
3075 $attrs->{from} = $rsrc->schema->storage->_inner_join_to_node ($attrs->{from}, $alias);
3078 #XXX - temp fix for result_class bug. There likely is a more elegant fix -groditi
3079 delete @{$attrs}{qw(result_class alias)};
3081 my $rel_source = $rsrc->related_source($rel);
3085 # The reason we do this now instead of passing the alias to the
3086 # search_rs below is that if you wrap/overload resultset on the
3087 # source you need to know what alias it's -going- to have for things
3088 # to work sanely (e.g. RestrictWithObject wants to be able to add
3089 # extra query restrictions, and these may need to be $alias.)
3091 my $rel_attrs = $rel_source->resultset_attributes;
3092 local $rel_attrs->{alias} = $alias;
3094 $rel_source->resultset
3098 where => $attrs->{where},
3102 if (my $cache = $self->get_cache) {
3103 my @related_cache = map
3104 { @{$_->related_resultset($rel)->get_cache||[]} }
3108 $new->set_cache(\@related_cache) if @related_cache;
3115 =head2 current_source_alias
3119 =item Arguments: none
3121 =item Return Value: $source_alias
3125 Returns the current table alias for the result source this resultset is built
3126 on, that will be used in the SQL query. Usually it is C<me>.
3128 Currently the source alias that refers to the result set returned by a
3129 L</search>/L</find> family method depends on how you got to the resultset: it's
3130 C<me> by default, but eg. L</search_related> aliases it to the related result
3131 source name (and keeps C<me> referring to the original result set). The long
3132 term goal is to make L<DBIx::Class> always alias the current resultset as C<me>
3133 (and make this method unnecessary).
3135 Thus it's currently necessary to use this method in predefined queries (see
3136 L<DBIx::Class::Manual::Cookbook/Predefined searches>) when referring to the
3137 source alias of the current result set:
3139 # in a result set class
3141 my ($self, $user) = @_;
3143 my $me = $self->current_source_alias;
3145 return $self->search({
3146 "$me.modified" => $user->id,
3152 sub current_source_alias {
3153 return (shift->{attrs} || {})->{alias} || 'me';
3156 =head2 as_subselect_rs
3160 =item Arguments: none
3162 =item Return Value: L<$resultset|/search>
3166 Act as a barrier to SQL symbols. The resultset provided will be made into a
3167 "virtual view" by including it as a subquery within the from clause. From this
3168 point on, any joined tables are inaccessible to ->search on the resultset (as if
3169 it were simply where-filtered without joins). For example:
3171 my $rs = $schema->resultset('Bar')->search({'x.name' => 'abc'},{ join => 'x' });
3173 # 'x' now pollutes the query namespace
3175 # So the following works as expected
3176 my $ok_rs = $rs->search({'x.other' => 1});
3178 # But this doesn't: instead of finding a 'Bar' related to two x rows (abc and
3179 # def) we look for one row with contradictory terms and join in another table
3180 # (aliased 'x_2') which we never use
3181 my $broken_rs = $rs->search({'x.name' => 'def'});
3183 my $rs2 = $rs->as_subselect_rs;
3185 # doesn't work - 'x' is no longer accessible in $rs2, having been sealed away
3186 my $not_joined_rs = $rs2->search({'x.other' => 1});
3188 # works as expected: finds a 'table' row related to two x rows (abc and def)
3189 my $correctly_joined_rs = $rs2->search({'x.name' => 'def'});
3191 Another example of when one might use this would be to select a subset of
3192 columns in a group by clause:
3194 my $rs = $schema->resultset('Bar')->search(undef, {
3195 group_by => [qw{ id foo_id baz_id }],
3196 })->as_subselect_rs->search(undef, {
3197 columns => [qw{ id foo_id }]
3200 In the above example normally columns would have to be equal to the group by,
3201 but because we isolated the group by into a subselect the above works.
3205 sub as_subselect_rs {
3208 my $attrs = $self->_resolved_attrs;
3210 my $fresh_rs = (ref $self)->new (
3211 $self->result_source
3214 # these pieces will be locked in the subquery
3215 delete $fresh_rs->{cond};
3216 delete @{$fresh_rs->{attrs}}{qw/where bind/};
3218 return $fresh_rs->search( {}, {
3220 $attrs->{alias} => $self->as_query,
3221 -alias => $attrs->{alias},
3222 -rsrc => $self->result_source,
3224 alias => $attrs->{alias},
3228 # This code is called by search_related, and makes sure there
3229 # is clear separation between the joins before, during, and
3230 # after the relationship. This information is needed later
3231 # in order to properly resolve prefetch aliases (any alias
3232 # with a relation_chain_depth less than the depth of the
3233 # current prefetch is not considered)
3235 # The increments happen twice per join. An even number means a
3236 # relationship specified via a search_related, whereas an odd
3237 # number indicates a join/prefetch added via attributes
3239 # Also this code will wrap the current resultset (the one we
3240 # chain to) in a subselect IFF it contains limiting attributes
3241 sub _chain_relationship {
3242 my ($self, $rel) = @_;
3243 my $source = $self->result_source;
3244 my $attrs = { %{$self->{attrs}||{}} };
3246 # we need to take the prefetch the attrs into account before we
3247 # ->_resolve_join as otherwise they get lost - captainL
3248 my $join = $self->_merge_joinpref_attr( $attrs->{join}, $attrs->{prefetch} );
3250 delete @{$attrs}{qw/join prefetch collapse group_by distinct _grouped_by_distinct select as columns +select +as +columns/};
3252 my $seen = { %{ (delete $attrs->{seen_join}) || {} } };
3255 my @force_subq_attrs = qw/offset rows group_by having/;
3258 ($attrs->{from} && ref $attrs->{from} ne 'ARRAY')
3260 $self->_has_resolved_attr (@force_subq_attrs)
3262 # Nuke the prefetch (if any) before the new $rs attrs
3263 # are resolved (prefetch is useless - we are wrapping
3264 # a subquery anyway).
3265 my $rs_copy = $self->search;
3266 $rs_copy->{attrs}{join} = $self->_merge_joinpref_attr (
3267 $rs_copy->{attrs}{join},
3268 delete $rs_copy->{attrs}{prefetch},
3273 -alias => $attrs->{alias},
3274 $attrs->{alias} => $rs_copy->as_query,
3276 delete @{$attrs}{@force_subq_attrs, qw/where bind/};
3277 $seen->{-relation_chain_depth} = 0;
3279 elsif ($attrs->{from}) { #shallow copy suffices
3280 $from = [ @{$attrs->{from}} ];
3285 -alias => $attrs->{alias},
3286 $attrs->{alias} => $source->from,
3290 my $jpath = ($seen->{-relation_chain_depth})
3291 ? $from->[-1][0]{-join_path}
3294 my @requested_joins = $source->_resolve_join(
3301 push @$from, @requested_joins;
3303 $seen->{-relation_chain_depth}++;
3305 # if $self already had a join/prefetch specified on it, the requested
3306 # $rel might very well be already included. What we do in this case
3307 # is effectively a no-op (except that we bump up the chain_depth on
3308 # the join in question so we could tell it *is* the search_related)
3311 # we consider the last one thus reverse
3312 for my $j (reverse @requested_joins) {
3313 my ($last_j) = keys %{$j->[0]{-join_path}[-1]};
3314 if ($rel eq $last_j) {
3315 $j->[0]{-relation_chain_depth}++;
3321 unless ($already_joined) {
3322 push @$from, $source->_resolve_join(
3330 $seen->{-relation_chain_depth}++;
3332 return {%$attrs, from => $from, seen_join => $seen};
3335 sub _resolved_attrs {
3337 return $self->{_attrs} if $self->{_attrs};
3339 my $attrs = { %{ $self->{attrs} || {} } };
3340 my $source = $self->result_source;
3341 my $alias = $attrs->{alias};
3343 $self->throw_exception("Specifying distinct => 1 in conjunction with collapse => 1 is unsupported")
3344 if $attrs->{collapse} and $attrs->{distinct};
3346 # default selection list
3347 $attrs->{columns} = [ $source->columns ]
3348 unless List::Util::first { exists $attrs->{$_} } qw/columns cols select as/;
3350 # merge selectors together
3351 for (qw/columns select as/) {
3352 $attrs->{$_} = $self->_merge_attr($attrs->{$_}, delete $attrs->{"+$_"})
3353 if $attrs->{$_} or $attrs->{"+$_"};
3356 # disassemble columns
3358 if (my $cols = delete $attrs->{columns}) {
3359 for my $c (ref $cols eq 'ARRAY' ? @$cols : $cols) {
3360 if (ref $c eq 'HASH') {
3361 for my $as (sort keys %$c) {
3362 push @sel, $c->{$as};
3373 # when trying to weed off duplicates later do not go past this point -
3374 # everything added from here on is unbalanced "anyone's guess" stuff
3375 my $dedup_stop_idx = $#as;
3377 push @as, @{ ref $attrs->{as} eq 'ARRAY' ? $attrs->{as} : [ $attrs->{as} ] }
3379 push @sel, @{ ref $attrs->{select} eq 'ARRAY' ? $attrs->{select} : [ $attrs->{select} ] }
3380 if $attrs->{select};
3382 # assume all unqualified selectors to apply to the current alias (legacy stuff)
3383 $_ = (ref $_ or $_ =~ /\./) ? $_ : "$alias.$_" for @sel;
3385 # disqualify all $alias.col as-bits (inflate-map mandated)
3386 $_ = ($_ =~ /^\Q$alias.\E(.+)$/) ? $1 : $_ for @as;
3388 # de-duplicate the result (remove *identical* select/as pairs)
3389 # and also die on duplicate {as} pointing to different {select}s
3390 # not using a c-style for as the condition is prone to shrinkage
3393 while ($i <= $dedup_stop_idx) {
3394 if ($seen->{"$sel[$i] \x00\x00 $as[$i]"}++) {
3399 elsif ($seen->{$as[$i]}++) {
3400 $self->throw_exception(
3401 "inflate_result() alias '$as[$i]' specified twice with different SQL-side {select}-ors"
3409 $attrs->{select} = \@sel;
3410 $attrs->{as} = \@as;
3412 $attrs->{from} ||= [{
3414 -alias => $self->{attrs}{alias},
3415 $self->{attrs}{alias} => $source->from,
3418 if ( $attrs->{join} || $attrs->{prefetch} ) {
3420 $self->throw_exception ('join/prefetch can not be used with a custom {from}')
3421 if ref $attrs->{from} ne 'ARRAY';
3423 my $join = (delete $attrs->{join}) || {};
3425 if ( defined $attrs->{prefetch} ) {
3426 $join = $self->_merge_joinpref_attr( $join, $attrs->{prefetch} );
3429 $attrs->{from} = # have to copy here to avoid corrupting the original
3431 @{ $attrs->{from} },
3432 $source->_resolve_join(
3435 { %{ $attrs->{seen_join} || {} } },
3436 ( $attrs->{seen_join} && keys %{$attrs->{seen_join}})
3437 ? $attrs->{from}[-1][0]{-join_path}
3444 if ( defined $attrs->{order_by} ) {
3445 $attrs->{order_by} = (
3446 ref( $attrs->{order_by} ) eq 'ARRAY'
3447 ? [ @{ $attrs->{order_by} } ]
3448 : [ $attrs->{order_by} || () ]
3452 if ($attrs->{group_by} and ref $attrs->{group_by} ne 'ARRAY') {
3453 $attrs->{group_by} = [ $attrs->{group_by} ];
3457 # generate selections based on the prefetch helper
3458 my ($prefetch, @prefetch_select, @prefetch_as);
3459 $prefetch = $self->_merge_joinpref_attr( {}, delete $attrs->{prefetch} )
3460 if defined $attrs->{prefetch};
3464 $self->throw_exception("Unable to prefetch, resultset contains an unnamed selector $attrs->{_dark_selector}{string}")
3465 if $attrs->{_dark_selector};
3467 $self->throw_exception("Specifying prefetch in conjunction with an explicit collapse => 0 is unsupported")
3468 if defined $attrs->{collapse} and ! $attrs->{collapse};
3470 $attrs->{collapse} = 1;
3472 # this is a separate structure (we don't look in {from} directly)
3473 # as the resolver needs to shift things off the lists to work
3474 # properly (identical-prefetches on different branches)
3476 if (ref $attrs->{from} eq 'ARRAY') {
3478 my $start_depth = $attrs->{seen_join}{-relation_chain_depth} || 0;
3480 for my $j ( @{$attrs->{from}}[1 .. $#{$attrs->{from}} ] ) {
3481 next unless $j->[0]{-alias};
3482 next unless $j->[0]{-join_path};
3483 next if ($j->[0]{-relation_chain_depth} || 0) < $start_depth;
3485 my @jpath = map { keys %$_ } @{$j->[0]{-join_path}};
3488 $p = $p->{$_} ||= {} for @jpath[ ($start_depth/2) .. $#jpath]; #only even depths are actual jpath boundaries
3489 push @{$p->{-join_aliases} }, $j->[0]{-alias};
3493 my @prefetch = $source->_resolve_prefetch( $prefetch, $alias, $join_map );
3495 # save these for after distinct resolution
3496 @prefetch_select = map { $_->[0] } @prefetch;
3497 @prefetch_as = map { $_->[1] } @prefetch;
3500 # run through the resulting joinstructure (starting from our current slot)
3501 # and unset collapse if proven unnecessary
3503 # also while we are at it find out if the current root source has
3504 # been premultiplied by previous related_source chaining
3506 # this allows to predict whether a root object with all other relation
3507 # data set to NULL is in fact unique
3508 if ($attrs->{collapse}) {
3510 if (ref $attrs->{from} eq 'ARRAY') {
3512 if (@{$attrs->{from}} == 1) {
3513 # no joins - no collapse
3514 $attrs->{collapse} = 0;
3517 # find where our table-spec starts
3518 my @fromlist = @{$attrs->{from}};
3520 my $t = shift @fromlist;
3523 # me vs join from-spec distinction - a ref means non-root
3524 if (ref $t eq 'ARRAY') {
3526 $is_multi ||= ! $t->{-is_single};
3528 last if ($t->{-alias} && $t->{-alias} eq $alias);
3529 $attrs->{_main_source_premultiplied} ||= $is_multi;
3532 # no non-singles remaining, nor any premultiplication - nothing to collapse
3534 ! $attrs->{_main_source_premultiplied}
3536 ! List::Util::first { ! $_->[0]{-is_single} } @fromlist
3538 $attrs->{collapse} = 0;
3544 # if we can not analyze the from - err on the side of safety
3545 $attrs->{_main_source_premultiplied} = 1;
3549 # generate the distinct induced group_by before injecting the prefetched select/as parts
3550 if (delete $attrs->{distinct}) {
3551 if ($attrs->{group_by}) {
3552 carp_unique ("Useless use of distinct on a grouped resultset ('distinct' is ignored when a 'group_by' is present)");
3555 $attrs->{_grouped_by_distinct} = 1;
3556 # distinct affects only the main selection part, not what prefetch may add below
3557 ($attrs->{group_by}, my $new_order) = $source->storage->_group_over_selection($attrs);
3559 # FIXME possibly ignore a rewritten order_by (may turn out to be an issue)
3560 # The thinking is: if we are collapsing the subquerying prefetch engine will
3561 # rip stuff apart for us anyway, and we do not want to have a potentially
3562 # function-converted external order_by
3563 # ( there is an explicit if ( collapse && _grouped_by_distinct ) check in DBIHacks )
3564 $attrs->{order_by} = $new_order unless $attrs->{collapse};
3568 # inject prefetch-bound selection (if any)
3569 push @{$attrs->{select}}, @prefetch_select;
3570 push @{$attrs->{as}}, @prefetch_as;
3572 # whether we can get away with the dumbest (possibly DBI-internal) collapser
3573 if ( List::Util::first { $_ =~ /\./ } @{$attrs->{as}} ) {
3574 $attrs->{_related_results_construction} = 1;
3577 # if both page and offset are specified, produce a combined offset
3578 # even though it doesn't make much sense, this is what pre 081xx has
3580 if (my $page = delete $attrs->{page}) {
3582 ($attrs->{rows} * ($page - 1))
3584 ($attrs->{offset} || 0)
3588 return $self->{_attrs} = $attrs;
3592 my ($self, $attr) = @_;
3594 if (ref $attr eq 'HASH') {
3595 return $self->_rollout_hash($attr);
3596 } elsif (ref $attr eq 'ARRAY') {
3597 return $self->_rollout_array($attr);
3603 sub _rollout_array {
3604 my ($self, $attr) = @_;
3607 foreach my $element (@{$attr}) {
3608 if (ref $element eq 'HASH') {
3609 push( @rolled_array, @{ $self->_rollout_hash( $element ) } );
3610 } elsif (ref $element eq 'ARRAY') {
3611 # XXX - should probably recurse here
3612 push( @rolled_array, @{$self->_rollout_array($element)} );
3614 push( @rolled_array, $element );
3617 return \@rolled_array;
3621 my ($self, $attr) = @_;
3624 foreach my $key (keys %{$attr}) {
3625 push( @rolled_array, { $key => $attr->{$key} } );
3627 return \@rolled_array;
3630 sub _calculate_score {
3631 my ($self, $a, $b) = @_;
3633 if (defined $a xor defined $b) {
3636 elsif (not defined $a) {
3640 if (ref $b eq 'HASH') {
3641 my ($b_key) = keys %{$b};
3642 if (ref $a eq 'HASH') {
3643 my ($a_key) = keys %{$a};
3644 if ($a_key eq $b_key) {
3645 return (1 + $self->_calculate_score( $a->{$a_key}, $b->{$b_key} ));
3650 return ($a eq $b_key) ? 1 : 0;
3653 if (ref $a eq 'HASH') {
3654 my ($a_key) = keys %{$a};
3655 return ($b eq $a_key) ? 1 : 0;
3657 return ($b eq $a) ? 1 : 0;
3662 sub _merge_joinpref_attr {
3663 my ($self, $orig, $import) = @_;
3665 return $import unless defined($orig);
3666 return $orig unless defined($import);
3668 $orig = $self->_rollout_attr($orig);
3669 $import = $self->_rollout_attr($import);
3672 foreach my $import_element ( @{$import} ) {
3673 # find best candidate from $orig to merge $b_element into
3674 my $best_candidate = { position => undef, score => 0 }; my $position = 0;
3675 foreach my $orig_element ( @{$orig} ) {
3676 my $score = $self->_calculate_score( $orig_element, $import_element );
3677 if ($score > $best_candidate->{score}) {
3678 $best_candidate->{position} = $position;
3679 $best_candidate->{score} = $score;
3683 my ($import_key) = ( ref $import_element eq 'HASH' ) ? keys %{$import_element} : ($import_element);
3684 $import_key = '' if not defined $import_key;
3686 if ($best_candidate->{score} == 0 || exists $seen_keys->{$import_key}) {
3687 push( @{$orig}, $import_element );
3689 my $orig_best = $orig->[$best_candidate->{position}];
3690 # merge orig_best and b_element together and replace original with merged
3691 if (ref $orig_best ne 'HASH') {
3692 $orig->[$best_candidate->{position}] = $import_element;
3693 } elsif (ref $import_element eq 'HASH') {
3694 my ($key) = keys %{$orig_best};
3695 $orig->[$best_candidate->{position}] = { $key => $self->_merge_joinpref_attr($orig_best->{$key}, $import_element->{$key}) };
3698 $seen_keys->{$import_key} = 1; # don't merge the same key twice
3701 return @$orig ? $orig : ();
3709 require Hash::Merge;
3710 my $hm = Hash::Merge->new;
3712 $hm->specify_behavior({
3715 my ($defl, $defr) = map { defined $_ } (@_[0,1]);
3717 if ($defl xor $defr) {
3718 return [ $defl ? $_[0] : $_[1] ];
3723 elsif (__HM_DEDUP and $_[0] eq $_[1]) {
3727 return [$_[0], $_[1]];
3731 return $_[1] if !defined $_[0];
3732 return $_[1] if __HM_DEDUP and List::Util::first { $_ eq $_[0] } @{$_[1]};
3733 return [$_[0], @{$_[1]}]
3736 return [] if !defined $_[0] and !keys %{$_[1]};
3737 return [ $_[1] ] if !defined $_[0];
3738 return [ $_[0] ] if !keys %{$_[1]};
3739 return [$_[0], $_[1]]
3744 return $_[0] if !defined $_[1];
3745 return $_[0] if __HM_DEDUP and List::Util::first { $_ eq $_[1] } @{$_[0]};
3746 return [@{$_[0]}, $_[1]]
3749 my @ret = @{$_[0]} or return $_[1];
3750 return [ @ret, @{$_[1]} ] unless __HM_DEDUP;
3751 my %idx = map { $_ => 1 } @ret;
3752 push @ret, grep { ! defined $idx{$_} } (@{$_[1]});
3756 return [ $_[1] ] if ! @{$_[0]};
3757 return $_[0] if !keys %{$_[1]};
3758 return $_[0] if __HM_DEDUP and List::Util::first { $_ eq $_[1] } @{$_[0]};
3759 return [ @{$_[0]}, $_[1] ];
3764 return [] if !keys %{$_[0]} and !defined $_[1];
3765 return [ $_[0] ] if !defined $_[1];
3766 return [ $_[1] ] if !keys %{$_[0]};
3767 return [$_[0], $_[1]]
3770 return [] if !keys %{$_[0]} and !@{$_[1]};
3771 return [ $_[0] ] if !@{$_[1]};
3772 return $_[1] if !keys %{$_[0]};
3773 return $_[1] if __HM_DEDUP and List::Util::first { $_ eq $_[0] } @{$_[1]};
3774 return [ $_[0], @{$_[1]} ];
3777 return [] if !keys %{$_[0]} and !keys %{$_[1]};
3778 return [ $_[0] ] if !keys %{$_[1]};
3779 return [ $_[1] ] if !keys %{$_[0]};
3780 return [ $_[0] ] if $_[0] eq $_[1];
3781 return [ $_[0], $_[1] ];
3784 } => 'DBIC_RS_ATTR_MERGER');
3788 return $hm->merge ($_[1], $_[2]);
3792 sub STORABLE_freeze {
3793 my ($self, $cloning) = @_;
3794 my $to_serialize = { %$self };
3796 # A cursor in progress can't be serialized (and would make little sense anyway)
3797 # the parser can be regenerated (and can't be serialized)
3798 delete @{$to_serialize}{qw/cursor _row_parser _result_inflator/};
3800 # nor is it sensical to store a not-yet-fired-count pager
3801 if ($to_serialize->{pager} and ref $to_serialize->{pager}{total_entries} eq 'CODE') {
3802 delete $to_serialize->{pager};
3805 Storable::nfreeze($to_serialize);
3808 # need this hook for symmetry
3810 my ($self, $cloning, $serialized) = @_;
3812 %$self = %{ Storable::thaw($serialized) };
3818 =head2 throw_exception
3820 See L<DBIx::Class::Schema/throw_exception> for details.
3824 sub throw_exception {
3827 if (ref $self and my $rsrc = $self->result_source) {
3828 $rsrc->throw_exception(@_)
3831 DBIx::Class::Exception->throw(@_);
3839 # XXX: FIXME: Attributes docs need clearing up
3843 Attributes are used to refine a ResultSet in various ways when
3844 searching for data. They can be passed to any method which takes an
3845 C<\%attrs> argument. See L</search>, L</search_rs>, L</find>,
3848 Default attributes can be set on the result class using
3849 L<DBIx::Class::ResultSource/resultset_attributes>. (Please read
3850 the CAVEATS on that feature before using it!)
3852 These are in no particular order:
3858 =item Value: ( $order_by | \@order_by | \%order_by )
3862 Which column(s) to order the results by.
3864 [The full list of suitable values is documented in
3865 L<SQL::Abstract/"ORDER BY CLAUSES">; the following is a summary of
3868 If a single column name, or an arrayref of names is supplied, the
3869 argument is passed through directly to SQL. The hashref syntax allows
3870 for connection-agnostic specification of ordering direction:
3872 For descending order:
3874 order_by => { -desc => [qw/col1 col2 col3/] }
3876 For explicit ascending order:
3878 order_by => { -asc => 'col' }
3880 The old scalarref syntax (i.e. order_by => \'year DESC') is still
3881 supported, although you are strongly encouraged to use the hashref
3882 syntax as outlined above.
3888 =item Value: \@columns | \%columns | $column
3892 Shortcut to request a particular set of columns to be retrieved. Each
3893 column spec may be a string (a table column name), or a hash (in which
3894 case the key is the C<as> value, and the value is used as the C<select>
3895 expression). Adds C<me.> onto the start of any column without a C<.> in
3896 it and sets C<select> from that, then auto-populates C<as> from
3897 C<select> as normal. (You may also use the C<cols> attribute, as in
3898 earlier versions of DBIC, but this is deprecated.)
3900 Essentially C<columns> does the same as L</select> and L</as>.
3902 columns => [ 'foo', { bar => 'baz' } ]
3906 select => [qw/foo baz/],
3913 =item Value: \@columns
3917 Indicates additional columns to be selected from storage. Works the same as
3918 L</columns> but adds columns to the selection. (You may also use the
3919 C<include_columns> attribute, as in earlier versions of DBIC, but this is
3920 deprecated). For example:-
3922 $schema->resultset('CD')->search(undef, {
3923 '+columns' => ['artist.name'],
3927 would return all CDs and include a 'name' column to the information
3928 passed to object inflation. Note that the 'artist' is the name of the
3929 column (or relationship) accessor, and 'name' is the name of the column
3930 accessor in the related table.
3932 B<NOTE:> You need to explicitly quote '+columns' when defining the attribute.
3933 Not doing so causes Perl to incorrectly interpret +columns as a bareword with a
3934 unary plus operator before it.
3936 =head2 include_columns
3940 =item Value: \@columns
3944 Deprecated. Acts as a synonym for L</+columns> for backward compatibility.
3950 =item Value: \@select_columns
3954 Indicates which columns should be selected from the storage. You can use
3955 column names, or in the case of RDBMS back ends, function or stored procedure
3958 $rs = $schema->resultset('Employee')->search(undef, {
3961 { count => 'employeeid' },
3962 { max => { length => 'name' }, -as => 'longest_name' }
3967 SELECT name, COUNT( employeeid ), MAX( LENGTH( name ) ) AS longest_name FROM employee
3969 B<NOTE:> You will almost always need a corresponding L</as> attribute when you
3970 use L</select>, to instruct DBIx::Class how to store the result of the column.
3971 Also note that the L</as> attribute has nothing to do with the SQL-side 'AS'
3972 identifier aliasing. You can however alias a function, so you can use it in
3973 e.g. an C<ORDER BY> clause. This is done via the C<-as> B<select function
3974 attribute> supplied as shown in the example above.
3976 B<NOTE:> You need to explicitly quote '+select'/'+as' when defining the attributes.
3977 Not doing so causes Perl to incorrectly interpret them as a bareword with a
3978 unary plus operator before it.
3984 Indicates additional columns to be selected from storage. Works the same as
3985 L</select> but adds columns to the default selection, instead of specifying
3994 =item Value: \@inflation_names
3998 Indicates column names for object inflation. That is L</as> indicates the
3999 slot name in which the column value will be stored within the
4000 L<Row|DBIx::Class::Row> object. The value will then be accessible via this
4001 identifier by the C<get_column> method (or via the object accessor B<if one
4002 with the same name already exists>) as shown below. The L</as> attribute has
4003 B<nothing to do> with the SQL-side C<AS>. See L</select> for details.
4005 $rs = $schema->resultset('Employee')->search(undef, {
4008 { count => 'employeeid' },
4009 { max => { length => 'name' }, -as => 'longest_name' }
4018 If the object against which the search is performed already has an accessor
4019 matching a column name specified in C<as>, the value can be retrieved using
4020 the accessor as normal:
4022 my $name = $employee->name();
4024 If on the other hand an accessor does not exist in the object, you need to
4025 use C<get_column> instead:
4027 my $employee_count = $employee->get_column('employee_count');
4029 You can create your own accessors if required - see
4030 L<DBIx::Class::Manual::Cookbook> for details.
4036 Indicates additional column names for those added via L</+select>. See L</as>.
4044 =item Value: ($rel_name | \@rel_names | \%rel_names)
4048 Contains a list of relationships that should be joined for this query. For
4051 # Get CDs by Nine Inch Nails
4052 my $rs = $schema->resultset('CD')->search(
4053 { 'artist.name' => 'Nine Inch Nails' },
4054 { join => 'artist' }
4057 Can also contain a hash reference to refer to the other relation's relations.
4060 package MyApp::Schema::Track;
4061 use base qw/DBIx::Class/;
4062 __PACKAGE__->table('track');
4063 __PACKAGE__->add_columns(qw/trackid cd position title/);
4064 __PACKAGE__->set_primary_key('trackid');
4065 __PACKAGE__->belongs_to(cd => 'MyApp::Schema::CD');
4068 # In your application
4069 my $rs = $schema->resultset('Artist')->search(
4070 { 'track.title' => 'Teardrop' },
4072 join => { cd => 'track' },
4073 order_by => 'artist.name',
4077 You need to use the relationship (not the table) name in conditions,
4078 because they are aliased as such. The current table is aliased as "me", so
4079 you need to use me.column_name in order to avoid ambiguity. For example:
4081 # Get CDs from 1984 with a 'Foo' track
4082 my $rs = $schema->resultset('CD')->search(
4085 'tracks.name' => 'Foo'
4087 { join => 'tracks' }
4090 If the same join is supplied twice, it will be aliased to <rel>_2 (and
4091 similarly for a third time). For e.g.
4093 my $rs = $schema->resultset('Artist')->search({
4094 'cds.title' => 'Down to Earth',
4095 'cds_2.title' => 'Popular',
4097 join => [ qw/cds cds/ ],
4100 will return a set of all artists that have both a cd with title 'Down
4101 to Earth' and a cd with title 'Popular'.
4103 If you want to fetch related objects from other tables as well, see L</prefetch>
4106 NOTE: An internal join-chain pruner will discard certain joins while
4107 constructing the actual SQL query, as long as the joins in question do not
4108 affect the retrieved result. This for example includes 1:1 left joins
4109 that are not part of the restriction specification (WHERE/HAVING) nor are
4110 a part of the query selection.
4112 For more help on using joins with search, see L<DBIx::Class::Manual::Joining>.
4118 =item Value: (0 | 1)
4122 When set to a true value, indicates that any rows fetched from joined has_many
4123 relationships are to be aggregated into the corresponding "parent" object. For
4124 example, the resultset:
4126 my $rs = $schema->resultset('CD')->search({}, {
4127 '+columns' => [ qw/ tracks.title tracks.position / ],
4132 While executing the following query:
4134 SELECT me.*, tracks.title, tracks.position
4136 LEFT JOIN track tracks
4137 ON tracks.cdid = me.cdid
4139 Will return only as many objects as there are rows in the CD source, even
4140 though the result of the query may span many rows. Each of these CD objects
4141 will in turn have multiple "Track" objects hidden behind the has_many
4142 generated accessor C<tracks>. Without C<< collapse => 1 >>, the return values
4143 of this resultset would be as many CD objects as there are tracks (a "Cartesian
4144 product"), with each CD object containing exactly one of all fetched Track data.
4146 When a collapse is requested on a non-ordered resultset, an order by some
4147 unique part of the main source (the left-most table) is inserted automatically.
4148 This is done so that the resultset is allowed to be "lazy" - calling
4149 L<< $rs->next|/next >> will fetch only as many rows as it needs to build the next
4150 object with all of its related data.
4152 If an L</order_by> is already declared, and orders the resultset in a way that
4153 makes collapsing as described above impossible (e.g. C<< ORDER BY
4154 has_many_rel.column >> or C<ORDER BY RANDOM()>), DBIC will automatically
4155 switch to "eager" mode and slurp the entire resultset before constructing the
4156 first object returned by L</next>.
4158 Setting this attribute on a resultset that does not join any has_many
4159 relations is a no-op.
4161 For a more in-depth discussion, see L</PREFETCHING>.
4167 =item Value: ($rel_name | \@rel_names | \%rel_names)
4171 This attribute is a shorthand for specifying a L</join> spec, adding all
4172 columns from the joined related sources as L</+columns> and setting
4173 L</collapse> to a true value. For example, the following two queries are
4176 my $rs = $schema->resultset('Artist')->search({}, {
4177 prefetch => { cds => ['genre', 'tracks' ] },
4182 my $rs = $schema->resultset('Artist')->search({}, {
4183 join => { cds => ['genre', 'tracks' ] },
4187 { +{ "cds.$_" => "cds.$_" } }
4188 $schema->source('Artist')->related_source('cds')->columns
4191 { +{ "cds.genre.$_" => "genre.$_" } }
4192 $schema->source('Artist')->related_source('cds')->related_source('genre')->columns
4195 { +{ "cds.tracks.$_" => "tracks.$_" } }
4196 $schema->source('Artist')->related_source('cds')->related_source('tracks')->columns
4201 Both producing the following SQL:
4203 SELECT me.artistid, me.name, me.rank, me.charfield,
4204 cds.cdid, cds.artist, cds.title, cds.year, cds.genreid, cds.single_track,
4205 genre.genreid, genre.name,
4206 tracks.trackid, tracks.cd, tracks.position, tracks.title, tracks.last_updated_on, tracks.last_updated_at
4209 ON cds.artist = me.artistid
4210 LEFT JOIN genre genre
4211 ON genre.genreid = cds.genreid
4212 LEFT JOIN track tracks
4213 ON tracks.cd = cds.cdid
4214 ORDER BY me.artistid
4216 While L</prefetch> implies a L</join>, it is ok to mix the two together, as
4217 the arguments are properly merged and generally do the right thing. For
4218 example, you may want to do the following:
4220 my $artists_and_cds_without_genre = $schema->resultset('Artist')->search(
4221 { 'genre.genreid' => undef },
4223 join => { cds => 'genre' },
4228 Which generates the following SQL:
4230 SELECT me.artistid, me.name, me.rank, me.charfield,
4231 cds.cdid, cds.artist, cds.title, cds.year, cds.genreid, cds.single_track
4234 ON cds.artist = me.artistid
4235 LEFT JOIN genre genre
4236 ON genre.genreid = cds.genreid
4237 WHERE genre.genreid IS NULL
4238 ORDER BY me.artistid
4240 For a more in-depth discussion, see L</PREFETCHING>.
4246 =item Value: $source_alias
4250 Sets the source alias for the query. Normally, this defaults to C<me>, but
4251 nested search queries (sub-SELECTs) might need specific aliases set to
4252 reference inner queries. For example:
4255 ->related_resultset('CDs')
4256 ->related_resultset('Tracks')
4258 'track.id' => { -ident => 'none_search.id' },
4262 my $ids = $self->search({
4265 alias => 'none_search',
4266 group_by => 'none_search.id',
4267 })->get_column('id')->as_query;
4269 $self->search({ id => { -in => $ids } })
4271 This attribute is directly tied to L</current_source_alias>.
4281 Makes the resultset paged and specifies the page to retrieve. Effectively
4282 identical to creating a non-pages resultset and then calling ->page($page)
4285 If L</rows> attribute is not specified it defaults to 10 rows per page.
4287 When you have a paged resultset, L</count> will only return the number
4288 of rows in the page. To get the total, use the L</pager> and call
4289 C<total_entries> on it.
4299 Specifies the maximum number of rows for direct retrieval or the number of
4300 rows per page if the page attribute or method is used.
4306 =item Value: $offset
4310 Specifies the (zero-based) row number for the first row to be returned, or the
4311 of the first row of the first page if paging is used.
4313 =head2 software_limit
4317 =item Value: (0 | 1)
4321 When combined with L</rows> and/or L</offset> the generated SQL will not
4322 include any limit dialect stanzas. Instead the entire result will be selected
4323 as if no limits were specified, and DBIC will perform the limit locally, by
4324 artificially advancing and finishing the resulting L</cursor>.
4326 This is the recommended way of performing resultset limiting when no sane RDBMS
4327 implementation is available (e.g.
4328 L<Sybase ASE|DBIx::Class::Storage::DBI::Sybase::ASE> using the
4329 L<Generic Sub Query|DBIx::Class::SQLMaker::LimitDialects/GenericSubQ> hack)
4335 =item Value: \@columns
4339 A arrayref of columns to group by. Can include columns of joined tables.
4341 group_by => [qw/ column1 column2 ... /]
4347 =item Value: $condition
4351 HAVING is a select statement attribute that is applied between GROUP BY and
4352 ORDER BY. It is applied to the after the grouping calculations have been
4355 having => { 'count_employee' => { '>=', 100 } }
4357 or with an in-place function in which case literal SQL is required:
4359 having => \[ 'count(employee) >= ?', [ count => 100 ] ]
4365 =item Value: (0 | 1)
4369 Set to 1 to automatically generate a L</group_by> clause based on the selection
4370 (including intelligent handling of L</order_by> contents). Note that the group
4371 criteria calculation takes place over the B<final> selection. This includes
4372 any L</+columns>, L</+select> or L</order_by> additions in subsequent
4373 L</search> calls, and standalone columns selected via
4374 L<DBIx::Class::ResultSetColumn> (L</get_column>). A notable exception are the
4375 extra selections specified via L</prefetch> - such selections are explicitly
4376 excluded from group criteria calculations.
4378 If the final ResultSet also explicitly defines a L</group_by> attribute, this
4379 setting is ignored and an appropriate warning is issued.
4385 Adds to the WHERE clause.
4387 # only return rows WHERE deleted IS NULL for all searches
4388 __PACKAGE__->resultset_attributes({ where => { deleted => undef } });
4390 Can be overridden by passing C<< { where => undef } >> as an attribute
4393 For more complicated where clauses see L<SQL::Abstract/WHERE CLAUSES>.
4399 Set to 1 to cache search results. This prevents extra SQL queries if you
4400 revisit rows in your ResultSet:
4402 my $resultset = $schema->resultset('Artist')->search( undef, { cache => 1 } );
4404 while( my $artist = $resultset->next ) {
4408 $rs->first; # without cache, this would issue a query
4410 By default, searches are not cached.
4412 For more examples of using these attributes, see
4413 L<DBIx::Class::Manual::Cookbook>.
4419 =item Value: ( 'update' | 'shared' | \$scalar )
4423 Set to 'update' for a SELECT ... FOR UPDATE or 'shared' for a SELECT
4424 ... FOR SHARED. If \$scalar is passed, this is taken directly and embedded in the
4429 DBIx::Class supports arbitrary related data prefetching from multiple related
4430 sources. Any combination of relationship types and column sets are supported.
4431 If L<collapsing|/collapse> is requested, there is an additional requirement of
4432 selecting enough data to make every individual object uniquely identifiable.
4434 Here are some more involved examples, based on the following relationship map:
4437 My::Schema::CD->belongs_to( artist => 'My::Schema::Artist' );
4438 My::Schema::CD->might_have( liner_note => 'My::Schema::LinerNotes' );
4439 My::Schema::CD->has_many( tracks => 'My::Schema::Track' );
4441 My::Schema::Artist->belongs_to( record_label => 'My::Schema::RecordLabel' );
4443 My::Schema::Track->has_many( guests => 'My::Schema::Guest' );
4447 my $rs = $schema->resultset('Tag')->search(
4456 The initial search results in SQL like the following:
4458 SELECT tag.*, cd.*, artist.* FROM tag
4459 JOIN cd ON tag.cd = cd.cdid
4460 JOIN artist ON cd.artist = artist.artistid
4462 L<DBIx::Class> has no need to go back to the database when we access the
4463 C<cd> or C<artist> relationships, which saves us two SQL statements in this
4466 Simple prefetches will be joined automatically, so there is no need
4467 for a C<join> attribute in the above search.
4469 The L</prefetch> attribute can be used with any of the relationship types
4470 and multiple prefetches can be specified together. Below is a more complex
4471 example that prefetches a CD's artist, its liner notes (if present),
4472 the cover image, the tracks on that CD, and the guests on those
4475 my $rs = $schema->resultset('CD')->search(
4479 { artist => 'record_label'}, # belongs_to => belongs_to
4480 'liner_note', # might_have
4481 'cover_image', # has_one
4482 { tracks => 'guests' }, # has_many => has_many
4487 This will produce SQL like the following:
4489 SELECT cd.*, artist.*, record_label.*, liner_note.*, cover_image.*,
4493 ON artist.artistid = me.artistid
4494 JOIN record_label record_label
4495 ON record_label.labelid = artist.labelid
4496 LEFT JOIN track tracks
4497 ON tracks.cdid = me.cdid
4498 LEFT JOIN guest guests
4499 ON guests.trackid = track.trackid
4500 LEFT JOIN liner_notes liner_note
4501 ON liner_note.cdid = me.cdid
4502 JOIN cd_artwork cover_image
4503 ON cover_image.cdid = me.cdid
4506 Now the C<artist>, C<record_label>, C<liner_note>, C<cover_image>,
4507 C<tracks>, and C<guests> of the CD will all be available through the
4508 relationship accessors without the need for additional queries to the
4513 Prefetch does a lot of deep magic. As such, it may not behave exactly
4514 as you might expect.
4520 Prefetch uses the L</cache> to populate the prefetched relationships. This
4521 may or may not be what you want.
4525 If you specify a condition on a prefetched relationship, ONLY those
4526 rows that match the prefetched condition will be fetched into that relationship.
4527 This means that adding prefetch to a search() B<may alter> what is returned by
4528 traversing a relationship. So, if you have C<< Artist->has_many(CDs) >> and you do
4530 my $artist_rs = $schema->resultset('Artist')->search({
4536 my $count = $artist_rs->first->cds->count;
4538 my $artist_rs_prefetch = $artist_rs->search( {}, { prefetch => 'cds' } );
4540 my $prefetch_count = $artist_rs_prefetch->first->cds->count;
4542 cmp_ok( $count, '==', $prefetch_count, "Counts should be the same" );
4544 That cmp_ok() may or may not pass depending on the datasets involved. In other
4545 words the C<WHERE> condition would apply to the entire dataset, just like
4546 it would in regular SQL. If you want to add a condition only to the "right side"
4547 of a C<LEFT JOIN> - consider declaring and using a L<relationship with a custom
4548 condition|DBIx::Class::Relationship::Base/condition>
4552 =head1 DBIC BIND VALUES
4554 Because DBIC may need more information to bind values than just the column name
4555 and value itself, it uses a special format for both passing and receiving bind
4556 values. Each bind value should be composed of an arrayref of
4557 C<< [ \%args => $val ] >>. The format of C<< \%args >> is currently:
4563 If present (in any form), this is what is being passed directly to bind_param.
4564 Note that different DBD's expect different bind args. (e.g. DBD::SQLite takes
4565 a single numerical type, while DBD::Pg takes a hashref if bind options.)
4567 If this is specified, all other bind options described below are ignored.
4571 If present, this is used to infer the actual bind attribute by passing to
4572 C<< $resolved_storage->bind_attribute_by_data_type() >>. Defaults to the
4573 "data_type" from the L<add_columns column info|DBIx::Class::ResultSource/add_columns>.
4575 Note that the data type is somewhat freeform (hence the sqlt_ prefix);
4576 currently drivers are expected to "Do the Right Thing" when given a common
4577 datatype name. (Not ideal, but that's what we got at this point.)
4581 Currently used to correctly allocate buffers for bind_param_inout().
4582 Defaults to "size" from the L<add_columns column info|DBIx::Class::ResultSource/add_columns>,
4583 or to a sensible value based on the "data_type".
4587 Used to fill in missing sqlt_datatype and sqlt_size attributes (if they are
4588 explicitly specified they are never overridden). Also used by some weird DBDs,
4589 where the column name should be available at bind_param time (e.g. Oracle).
4593 For backwards compatibility and convenience, the following shortcuts are
4596 [ $name => $val ] === [ { dbic_colname => $name }, $val ]
4597 [ \$dt => $val ] === [ { sqlt_datatype => $dt }, $val ]
4598 [ undef, $val ] === [ {}, $val ]
4599 $val === [ {}, $val ]
4601 =head1 AUTHOR AND CONTRIBUTORS
4603 See L<AUTHOR|DBIx::Class/AUTHOR> and L<CONTRIBUTORS|DBIx::Class/CONTRIBUTORS> in DBIx::Class
4607 You may distribute this code under the same terms as Perl itself.