X-Git-Url: http://git.shadowcat.co.uk/gitweb/gitweb.cgi?a=blobdiff_plain;f=lib%2FDBIx%2FClass%2FResultSet.pm;h=c799da79395c6a2c6a4ac714c2e66e8fe3a71c8a;hb=83d2991997f6070366d3d2bcd3f1bcc07562b930;hp=aa6477484df5864fc65418e2d19a1441c18ab7e0;hpb=3904d3c3bd6e3dda5dbc9dc49f8cc778eef114e2;p=dbsrgits%2FDBIx-Class.git diff --git a/lib/DBIx/Class/ResultSet.pm b/lib/DBIx/Class/ResultSet.pm index aa64774..c799da7 100644 --- a/lib/DBIx/Class/ResultSet.pm +++ b/lib/DBIx/Class/ResultSet.pm @@ -2,25 +2,35 @@ package DBIx::Class::ResultSet; use strict; use warnings; +use base qw/DBIx::Class/; +use DBIx::Class::Carp; +use DBIx::Class::ResultSetColumn; +use Scalar::Util qw/blessed weaken reftype/; +use DBIx::Class::_Util 'fail_on_internal_wantarray'; +use Try::Tiny; +use Data::Compare (); # no imports!!! guard against insane architecture + +# not importing first() as it will clash with our own method +use List::Util (); + +BEGIN { + # De-duplication in _merge_attr() is disabled, but left in for reference + # (the merger is used for other things that ought not to be de-duped) + *__HM_DEDUP = sub () { 0 }; +} + +use namespace::clean; + use overload '0+' => "count", 'bool' => "_bool", fallback => 1; -use Carp::Clan qw/^DBIx::Class/; -use DBIx::Class::Exception; -use Data::Page; -use Storable; -use DBIx::Class::ResultSetColumn; -use DBIx::Class::ResultSourceHandle; -use List::Util (); -use Scalar::Util (); - -use base qw/DBIx::Class/; -#use Test::Deep::NoTest (qw/eq_deeply/); -use Data::Dumper::Concise; +# this is real - CDBICompat overrides it with insanity +# yes, prototype won't matter, but that's for now ;) +sub _bool () { 1 } -__PACKAGE__->mk_group_accessors('simple' => qw/_result_class _source_handle/); +__PACKAGE__->mk_group_accessors('simple' => qw/_result_class result_source/); =head1 NAME @@ -28,8 +38,12 @@ DBIx::Class::ResultSet - Represents a query used for fetching a set of results. =head1 SYNOPSIS - my $users_rs = $schema->resultset('User'); - my $registered_users_rs = $schema->resultset('User')->search({ registered => 1 }); + my $users_rs = $schema->resultset('User'); + while( $user = $users_rs->next) { + print $user->username; + } + + my $registered_users_rs = $schema->resultset('User')->search({ registered => 1 }); my @cds_in_2005 = $schema->resultset('CD')->search({ year => 2005 })->all(); =head1 DESCRIPTION @@ -57,7 +71,40 @@ represents. The query that the ResultSet represents is B executed against the database when these methods are called: -L L L L L L +L, L, L, L, L, L. + +If a resultset is used in a numeric context it returns the L. +However, if it is used in a boolean context it is B true. So if +you want to check if a resultset has any results, you must use C. + +=head1 CUSTOM ResultSet CLASSES THAT USE Moose + +If you want to make your custom ResultSet classes with L, use a template +similar to: + + package MyApp::Schema::ResultSet::User; + + use Moose; + use namespace::autoclean; + use MooseX::NonMoose; + extends 'DBIx::Class::ResultSet'; + + sub BUILDARGS { $_[2] } + + ...your code... + + __PACKAGE__->meta->make_immutable; + + 1; + +The L is necessary so that the L constructor does not +clash with the regular ResultSet constructor. Alternatively, you can use: + + __PACKAGE__->meta->make_immutable(inline_constructor => 0); + +The L is necessary because the +signature of the ResultSet C is C<< ->new($source, \%args) >>. =head1 EXAMPLES @@ -72,14 +119,14 @@ another. sub get_data { my $self = shift; my $request = $self->get_request; # Get a request object somehow. - my $schema = $self->get_schema; # Get the DBIC schema object somehow. + my $schema = $self->result_source->schema; my $cd_rs = $schema->resultset('CD')->search({ title => $request->param('title'), year => $request->param('year'), }); - $self->apply_security_policy( $cd_rs ); + $cd_rs = $self->apply_security_policy( $cd_rs ); return $cd_rs->all(); } @@ -95,13 +142,17 @@ another. =head3 Resolving conditions and attributes -When a resultset is chained from another resultset, conditions and -attributes with the same keys need resolving. +When a resultset is chained from another resultset (e.g.: +C<< my $new_rs = $old_rs->search(\%extra_cond, \%attrs) >>), conditions +and attributes with the same keys need resolving. -L, L, L, L attributes are merged -into the existing ones from the original resultset. +If any of L, L, L are present, they reset the +original selection, and start the selection "clean". -The L, L attribute, and any search conditions are +The L, L, L, L, L attributes +are merged into the existing ones from the original resultset. + +The L and L attributes, and any search conditions, are merged with an SQL C to the existing condition from the original resultset. @@ -142,22 +193,15 @@ Which is the same as: See: L, L, L, L, L. -=head1 OVERLOADING - -If a resultset is used in a numeric context it returns the L. -However, if it is used in a boolean context it is always true. So if -you want to check if a resultset has any results use C. -C will always be true. - =head1 METHODS =head2 new =over 4 -=item Arguments: $source, \%$attrs +=item Arguments: L<$source|DBIx::Class::ResultSource>, L<\%attrs?|/ATTRIBUTES> -=item Return Value: $rs +=item Return Value: L<$resultset|/search> =back @@ -166,16 +210,31 @@ L) and an attribute hash (see L below). Does not perform any queries -- these are executed as needed by the other methods. -Generally you won't need to construct a resultset manually. You'll -automatically get one from e.g. a L called in scalar context: +Generally you never construct a resultset manually. Instead you get one +from e.g. a +C<< $schema->L('$source_name') >> +or C<< $another_resultset->L(...) >> (the later called in +scalar context): my $rs = $schema->resultset('CD')->search({ title => '100th Window' }); -IMPORTANT: If called on an object, proxies to new_result instead so +=over + +=item WARNING + +If called on an object, proxies to L instead, so my $cd = $schema->resultset('CD')->new({ title => 'Spoon' }); -will return a CD object, not a ResultSet. +will return a CD object, not a ResultSet, and is equivalent to: + + my $cd = $schema->resultset('CD')->new_result({ title => 'Spoon' }); + +Please also keep in mind that many internals call L directly, +so overloading this method with the idea of intercepting new result object +creation B. See also warning pertaining to L. + +=back =cut @@ -184,9 +243,11 @@ sub new { return $class->new_result(@_) if ref $class; my ($source, $attrs) = @_; - $source = $source->handle - unless $source->isa('DBIx::Class::ResultSourceHandle'); + $source = $source->resolve + if $source->isa('DBIx::Class::ResultSourceHandle'); + $attrs = { %{$attrs||{}} }; + delete @{$attrs}{qw(_last_sqlmaker_alias_map _related_results_construction)}; if ($attrs->{page}) { $attrs->{rows} ||= 10; @@ -194,32 +255,33 @@ sub new { $attrs->{alias} ||= 'me'; - # Creation of {} and bless separated to mitigate RH perl bug - # see https://bugzilla.redhat.com/show_bug.cgi?id=196836 - my $self = { - _source_handle => $source, + my $self = bless { + result_source => $source, cond => $attrs->{where}, - count => undef, pager => undef, - attrs => $attrs - }; + attrs => $attrs, + }, $class; - bless $self, $class; + # if there is a dark selector, this means we are already in a + # chain and the cleanup/sanification was taken care of by + # _search_rs already + $self->_normalize_selection($attrs) + unless $attrs->{_dark_selector}; $self->result_class( - $attrs->{result_class} || $source->resolve->result_class + $attrs->{result_class} || $source->result_class ); - return $self; + $self; } =head2 search =over 4 -=item Arguments: $cond, \%attrs? +=item Arguments: L<$cond|DBIx::Class::SQLMaker> | undef, L<\%attrs?|/ATTRIBUTES> -=item Return Value: $resultset (scalar context), @row_objs (list context) +=item Return Value: $resultset (scalar context) | L<@result_objs|DBIx::Class::Manual::ResultClass> (list context) =back @@ -229,6 +291,10 @@ sub new { my $new_rs = $cd_rs->search([ { year => 2005 }, { year => 2004 } ]); # year = 2005 OR year = 2004 +In list context, C<< ->all() >> is called implicitly on the resultset, thus +returning a list of L objects instead. +To avoid that, use L. + If you need to pass in additional attributes but no additional condition, call it as C. @@ -239,26 +305,55 @@ call it as C. For a list of attributes that can be passed to C, see L. For more examples of using this function, see -L. For a complete -documentation for the first argument, see L. +L. For a complete +documentation for the first argument, see L +and its extension L. For more help on using joins with search, see L. +=head3 CAVEAT + +Note that L does not process/deflate any of the values passed in the +L-compatible search condition structure. This is unlike other +condition-bound methods L, L and L. The user must ensure +manually that any value passed to this method will stringify to something the +RDBMS knows how to deal with. A notable example is the handling of L +objects, for more info see: +L. + =cut sub search { my $self = shift; my $rs = $self->search_rs( @_ ); - return (wantarray ? $rs->all : $rs); + + if (wantarray) { + DBIx::Class::_ENV_::ASSERT_NO_INTERNAL_WANTARRAY and my $sog = fail_on_internal_wantarray($rs); + return $rs->all; + } + elsif (defined wantarray) { + return $rs; + } + else { + # we can be called by a relationship helper, which in + # turn may be called in void context due to some braindead + # overload or whatever else the user decided to be clever + # at this particular day. Thus limit the exception to + # external code calls only + $self->throw_exception ('->search is *not* a mutator, calling it in void context makes no sense') + if (caller)[0] !~ /^\QDBIx::Class::/; + + return (); + } } =head2 search_rs =over 4 -=item Arguments: $cond, \%attrs? +=item Arguments: L<$cond|DBIx::Class::SQLMaker>, L<\%attrs?|/ATTRIBUTES> -=item Return Value: $resultset +=item Return Value: L<$resultset|/search> =back @@ -270,115 +365,298 @@ always return a resultset, even in list context. sub search_rs { my $self = shift; - # Special-case handling for (undef, undef). - if ( @_ == 2 && !defined $_[1] && !defined $_[0] ) { - pop(@_); pop(@_); - } + my $rsrc = $self->result_source; + my ($call_cond, $call_attrs); + + # Special-case handling for (undef, undef) or (undef) + # Note that (foo => undef) is valid deprecated syntax + @_ = () if not scalar grep { defined $_ } @_; - my $attrs = {}; - $attrs = pop(@_) if @_ > 1 and ref $_[$#_] eq 'HASH'; - my $our_attrs = { %{$self->{attrs}} }; - my $having = delete $our_attrs->{having}; - my $where = delete $our_attrs->{where}; + # just a cond + if (@_ == 1) { + $call_cond = shift; + } + # fish out attrs in the ($condref, $attr) case + elsif (@_ == 2 and ( ! defined $_[0] or (ref $_[0]) ne '') ) { + ($call_cond, $call_attrs) = @_; + } + elsif (@_ % 2) { + $self->throw_exception('Odd number of arguments to search') + } + # legacy search + elsif (@_) { + carp_unique 'search( %condition ) is deprecated, use search( \%condition ) instead' + unless $rsrc->result_class->isa('DBIx::Class::CDBICompat'); + + for my $i (0 .. $#_) { + next if $i % 2; + $self->throw_exception ('All keys in condition key/value pairs must be plain scalars') + if (! defined $_[$i] or ref $_[$i] ne ''); + } - my $rows; + $call_cond = { @_ }; + } + # see if we can keep the cache (no $rs changes) + my $cache; my %safe = (alias => 1, cache => 1); + if ( ! List::Util::first { !$safe{$_} } keys %$call_attrs and ( + ! defined $call_cond + or + ref $call_cond eq 'HASH' && ! keys %$call_cond + or + ref $call_cond eq 'ARRAY' && ! @$call_cond + )) { + $cache = $self->get_cache; + } - unless ( - (@_ && defined($_[0])) # @_ == () or (undef) - || - (keys %$attrs # empty attrs or only 'safe' attrs - && List::Util::first { !$safe{$_} } keys %$attrs) - ) { - # no search, effectively just a clone - $rows = $self->get_cache; - } - - # reset the selector list - if (List::Util::first { exists $attrs->{$_} } qw{columns select as}) { - delete @{$our_attrs}{qw{select as columns +select +as +columns include_columns}}; - } - - my $new_attrs = { %{$our_attrs}, %{$attrs} }; - - # merge new attrs into inherited - foreach my $key (qw/join prefetch +select +as +columns include_columns bind/) { - next unless exists $attrs->{$key}; - $new_attrs->{$key} = $self->_merge_attr($our_attrs->{$key}, $attrs->{$key}); - } - - my $cond = (@_ - ? ( - (@_ == 1 || ref $_[0] eq "HASH") - ? ( - (ref $_[0] eq 'HASH') - ? ( - (keys %{ $_[0] } > 0) - ? shift - : undef - ) - : shift - ) - : ( - (@_ % 2) - ? $self->throw_exception("Odd number of arguments to search") - : {@_} - ) - ) - : undef - ); + my $old_attrs = { %{$self->{attrs}} }; + my ($old_having, $old_where) = delete @{$old_attrs}{qw(having where)}; - if (defined $where) { - $new_attrs->{where} = ( - defined $new_attrs->{where} - ? { '-and' => [ - map { - ref $_ eq 'ARRAY' ? [ -or => $_ ] : $_ - } $where, $new_attrs->{where} - ] - } - : $where); + my $new_attrs = { %$old_attrs }; + + # take care of call attrs (only if anything is changing) + if ($call_attrs and keys %$call_attrs) { + + # copy for _normalize_selection + $call_attrs = { %$call_attrs }; + + my @selector_attrs = qw/select as columns cols +select +as +columns include_columns/; + + # reset the current selector list if new selectors are supplied + if (List::Util::first { exists $call_attrs->{$_} } qw/columns cols select as/) { + delete @{$old_attrs}{(@selector_attrs, '_dark_selector')}; + } + + # Normalize the new selector list (operates on the passed-in attr structure) + # Need to do it on every chain instead of only once on _resolved_attrs, in + # order to allow detection of empty vs partial 'as' + $call_attrs->{_dark_selector} = $old_attrs->{_dark_selector} + if $old_attrs->{_dark_selector}; + $self->_normalize_selection ($call_attrs); + + # start with blind overwriting merge, exclude selector attrs + $new_attrs = { %{$old_attrs}, %{$call_attrs} }; + delete @{$new_attrs}{@selector_attrs}; + + for (@selector_attrs) { + $new_attrs->{$_} = $self->_merge_attr($old_attrs->{$_}, $call_attrs->{$_}) + if ( exists $old_attrs->{$_} or exists $call_attrs->{$_} ); + } + + # older deprecated name, use only if {columns} is not there + if (my $c = delete $new_attrs->{cols}) { + carp_unique( "Resultset attribute 'cols' is deprecated, use 'columns' instead" ); + if ($new_attrs->{columns}) { + carp "Resultset specifies both the 'columns' and the legacy 'cols' attributes - ignoring 'cols'"; + } + else { + $new_attrs->{columns} = $c; + } + } + + + # join/prefetch use their own crazy merging heuristics + foreach my $key (qw/join prefetch/) { + $new_attrs->{$key} = $self->_merge_joinpref_attr($old_attrs->{$key}, $call_attrs->{$key}) + if exists $call_attrs->{$key}; + } + + # stack binds together + $new_attrs->{bind} = [ @{ $old_attrs->{bind} || [] }, @{ $call_attrs->{bind} || [] } ]; } - if (defined $cond) { - $new_attrs->{where} = ( - defined $new_attrs->{where} - ? { '-and' => [ - map { - ref $_ eq 'ARRAY' ? [ -or => $_ ] : $_ - } $cond, $new_attrs->{where} - ] - } - : $cond); + + for ($old_where, $call_cond) { + if (defined $_) { + $new_attrs->{where} = $self->_stack_cond ( + $_, $new_attrs->{where} + ); + } } - if (defined $having) { - $new_attrs->{having} = ( - defined $new_attrs->{having} - ? { '-and' => [ - map { - ref $_ eq 'ARRAY' ? [ -or => $_ ] : $_ - } $having, $new_attrs->{having} - ] - } - : $having); + if (defined $old_having) { + $new_attrs->{having} = $self->_stack_cond ( + $old_having, $new_attrs->{having} + ) } - my $rs = (ref $self)->new($self->result_source, $new_attrs); + my $rs = (ref $self)->new($rsrc, $new_attrs); - $rs->set_cache($rows) if ($rows); + $rs->set_cache($cache) if ($cache); return $rs; } +my $dark_sel_dumper; +sub _normalize_selection { + my ($self, $attrs) = @_; + + # legacy syntax + if ( exists $attrs->{include_columns} ) { + carp_unique( "Resultset attribute 'include_columns' is deprecated, use '+columns' instead" ); + $attrs->{'+columns'} = $self->_merge_attr( + $attrs->{'+columns'}, delete $attrs->{include_columns} + ); + } + + # columns are always placed first, however + + # Keep the X vs +X separation until _resolved_attrs time - this allows to + # delay the decision on whether to use a default select list ($rsrc->columns) + # allowing stuff like the remove_columns helper to work + # + # select/as +select/+as pairs need special handling - the amount of select/as + # elements in each pair does *not* have to be equal (think multicolumn + # selectors like distinct(foo, bar) ). If the selector is bare (no 'as' + # supplied at all) - try to infer the alias, either from the -as parameter + # of the selector spec, or use the parameter whole if it looks like a column + # name (ugly legacy heuristic). If all fails - leave the selector bare (which + # is ok as well), but make sure no more additions to the 'as' chain take place + for my $pref ('', '+') { + + my ($sel, $as) = map { + my $key = "${pref}${_}"; + + my $val = [ ref $attrs->{$key} eq 'ARRAY' + ? @{$attrs->{$key}} + : $attrs->{$key} || () + ]; + delete $attrs->{$key}; + $val; + } qw/select as/; + + if (! @$as and ! @$sel ) { + next; + } + elsif (@$as and ! @$sel) { + $self->throw_exception( + "Unable to handle ${pref}as specification (@$as) without a corresponding ${pref}select" + ); + } + elsif( ! @$as ) { + # no as part supplied at all - try to deduce (unless explicit end of named selection is declared) + # if any @$as has been supplied we assume the user knows what (s)he is doing + # and blindly keep stacking up pieces + unless ($attrs->{_dark_selector}) { + SELECTOR: + for (@$sel) { + if ( ref $_ eq 'HASH' and exists $_->{-as} ) { + push @$as, $_->{-as}; + } + # assume any plain no-space, no-parenthesis string to be a column spec + # FIXME - this is retarded but is necessary to support shit like 'count(foo)' + elsif ( ! ref $_ and $_ =~ /^ [^\s\(\)]+ $/x) { + push @$as, $_; + } + # if all else fails - raise a flag that no more aliasing will be allowed + else { + $attrs->{_dark_selector} = { + plus_stage => $pref, + string => ($dark_sel_dumper ||= do { + require Data::Dumper::Concise; + Data::Dumper::Concise::DumperObject()->Indent(0); + })->Values([$_])->Dump + , + }; + last SELECTOR; + } + } + } + } + elsif (@$as < @$sel) { + $self->throw_exception( + "Unable to handle an ${pref}as specification (@$as) with less elements than the corresponding ${pref}select" + ); + } + elsif ($pref and $attrs->{_dark_selector}) { + $self->throw_exception( + "Unable to process named '+select', resultset contains an unnamed selector $attrs->{_dark_selector}{string}" + ); + } + + + # merge result + $attrs->{"${pref}select"} = $self->_merge_attr($attrs->{"${pref}select"}, $sel); + $attrs->{"${pref}as"} = $self->_merge_attr($attrs->{"${pref}as"}, $as); + } +} + +sub _stack_cond { + my ($self, $left, $right) = @_; + + # collapse single element top-level conditions + # (single pass only, unlikely to need recursion) + for ($left, $right) { + if (ref $_ eq 'ARRAY') { + if (@$_ == 0) { + $_ = undef; + } + elsif (@$_ == 1) { + $_ = $_->[0]; + } + } + elsif (ref $_ eq 'HASH') { + my ($first, $more) = keys %$_; + + # empty hash + if (! defined $first) { + $_ = undef; + } + # one element hash + elsif (! defined $more) { + if ($first eq '-and' and ref $_->{'-and'} eq 'HASH') { + $_ = $_->{'-and'}; + } + elsif ($first eq '-or' and ref $_->{'-or'} eq 'ARRAY') { + $_ = $_->{'-or'}; + } + } + } + } + + # merge hashes with weeding out of duplicates (simple cases only) + if (ref $left eq 'HASH' and ref $right eq 'HASH') { + + # shallow copy to destroy + $right = { %$right }; + for (grep { exists $right->{$_} } keys %$left) { + # the use of eq_deeply here is justified - the rhs of an + # expression can contain a lot of twisted weird stuff + delete $right->{$_} if Data::Compare::Compare( $left->{$_}, $right->{$_} ); + } + + $right = undef unless keys %$right; + } + + + if (defined $left xor defined $right) { + return defined $left ? $left : $right; + } + elsif (! defined $left) { + return undef; + } + else { + return { -and => [ $left, $right ] }; + } +} + =head2 search_literal +B: C is provided for Class::DBI compatibility and +should only be used in that context. C is a convenience +method. It is equivalent to calling C<< $schema->search(\[]) >>, but if you +want to ensure columns are bound correctly, use L. + +See L and +L for searching techniques that do not +require C. + =over 4 -=item Arguments: $sql_fragment, @bind_values +=item Arguments: $sql_fragment, @standalone_bind_values -=item Return Value: $resultset (scalar context), @row_objs (list context) +=item Return Value: L<$resultset|/search> (scalar context) | L<@result_objs|DBIx::Class::Manual::ResultClass> (list context) =back @@ -388,21 +666,11 @@ sub search_rs { Pass a literal chunk of SQL to be added to the conditional part of the resultset query. -CAVEAT: C is provided for Class::DBI compatibility and should -only be used in that context. C is a convenience method. -It is equivalent to calling $schema->search(\[]), but if you want to ensure -columns are bound correctly, use C. - Example of how to use C instead of C my @cds = $cd_rs->search_literal('cdid = ? AND (artist = ? OR artist = ?)', (2, 1, 2)); my @cds = $cd_rs->search(\[ 'cdid = ? AND (artist = ? OR artist = ?)', [ 'cdid', 2 ], [ 'artist', 1 ], [ 'artist', 2 ] ]); - -See L and -L for searching techniques that do not -require C. - =cut sub search_literal { @@ -411,32 +679,63 @@ sub search_literal { if ( @bind && ref($bind[-1]) eq 'HASH' ) { $attr = pop @bind; } - return $self->search(\[ $sql, map [ __DUMMY__ => $_ ], @bind ], ($attr || () )); + return $self->search(\[ $sql, map [ {} => $_ ], @bind ], ($attr || () )); } =head2 find =over 4 -=item Arguments: @values | \%cols, \%attrs? +=item Arguments: \%columns_values | @pk_values, { key => $unique_constraint, L<%attrs|/ATTRIBUTES> }? -=item Return Value: $row_object | undef +=item Return Value: L<$result|DBIx::Class::Manual::ResultClass> | undef =back -Finds a row based on its primary key or unique constraint. For example, to find -a row by its primary key: +Finds and returns a single row based on supplied criteria. Takes either a +hashref with the same format as L (including inference of foreign +keys from related objects), or a list of primary key values in the same +order as the L +declaration on the L. + +In either case an attempt is made to combine conditions already existing on +the resultset with the condition passed to this method. + +To aid with preparing the correct query for the storage you may supply the +C attribute, which is the name of a +L (the +unique constraint corresponding to the +L is always named +C). If the C attribute has been supplied, and DBIC is unable +to construct a query that satisfies the named unique constraint fully ( +non-NULL values for each column member of the constraint) an exception is +thrown. + +If no C is specified, the search is carried over all unique constraints +which are fully defined by the available condition. + +If no such constraint is found, C currently defaults to a simple +C<< search->(\%column_values) >> which may or may not do what you expect. +Note that this fallback behavior may be deprecated in further versions. If +you need to search with arbitrary conditions - use L. If the query +resulting from this fallback produces more than one row, a warning to the +effect is issued, though only the first row is constructed and returned as +C<$result_object>. - my $cd = $schema->resultset('CD')->find(5); +In addition to C, L recognizes and applies standard +L in the same way as L does. -You can also find a row by a specific unique constraint using the C -attribute. For example: +Note that if you have extra concerns about the correctness of the resulting +query you need to specify the C attribute and supply the entire condition +as an argument to find (since it is not always possible to perform the +combination of the resultset condition with the supplied one, especially if +the resultset condition contains literal sql). - my $cd = $schema->resultset('CD')->find('Massive Attack', 'Mezzanine', { - key => 'cd_artist_title' - }); +For example, to find a row by its primary key: -Additionally, you can specify the columns explicitly by name: + my $cd = $schema->resultset('CD')->find(5); + +You can also find a row by a specific unique constraint: my $cd = $schema->resultset('CD')->find( { @@ -446,24 +745,7 @@ Additionally, you can specify the columns explicitly by name: { key => 'cd_artist_title' } ); -If the C is specified as C, it searches only on the primary key. - -If no C is specified, it searches on all unique constraints defined on the -source for which column data is provided, including the primary key. - -If your table does not have a primary key, you B provide a value for the -C attribute matching one of the unique constraints on the source. - -In addition to C, L recognizes and applies standard -L in the same way as L does. - -Note: If your query does not return only one row, a warning is generated: - - Query returned more than one row - -See also L and L. For information on how to -declare unique constraints, see -L. +See also L and L. =cut @@ -471,57 +753,75 @@ sub find { my $self = shift; my $attrs = (@_ > 1 && ref $_[$#_] eq 'HASH' ? pop(@_) : {}); - # Default to the primary key, but allow a specific key - my @cols = exists $attrs->{key} - ? $self->result_source->unique_constraint_columns($attrs->{key}) - : $self->result_source->primary_columns; - $self->throw_exception( - "Can't find unless a primary key is defined or unique constraint is specified" - ) unless @cols; + my $rsrc = $self->result_source; - # Parse out a hashref from input - my $input_query; - if (ref $_[0] eq 'HASH') { - $input_query = { %{$_[0]} }; + my $constraint_name; + if (exists $attrs->{key}) { + $constraint_name = defined $attrs->{key} + ? $attrs->{key} + : $self->throw_exception("An undefined 'key' resultset attribute makes no sense") + ; } - elsif (@_ == @cols) { - $input_query = {}; - @{$input_query}{@cols} = @_; + + # Parse out the condition from input + my $call_cond; + + if (ref $_[0] eq 'HASH') { + $call_cond = { %{$_[0]} }; } else { - # Compatibility: Allow e.g. find(id => $value) - carp "Find by key => value deprecated; please use a hashref instead"; - $input_query = {@_}; - } - - my (%related, $info); - - KEY: foreach my $key (keys %$input_query) { - if (ref($input_query->{$key}) - && ($info = $self->result_source->relationship_info($key))) { - my $val = delete $input_query->{$key}; - next KEY if (ref($val) eq 'ARRAY'); # has_many for multi_create - my $rel_q = $self->result_source->_resolve_condition( - $info->{cond}, $val, $key - ); - die "Can't handle OR join condition in find" if ref($rel_q) eq 'ARRAY'; + # if only values are supplied we need to default to 'primary' + $constraint_name = 'primary' unless defined $constraint_name; + + my @c_cols = $rsrc->unique_constraint_columns($constraint_name); + + $self->throw_exception( + "No constraint columns, maybe a malformed '$constraint_name' constraint?" + ) unless @c_cols; + + $self->throw_exception ( + 'find() expects either a column/value hashref, or a list of values ' + . "corresponding to the columns of the specified unique constraint '$constraint_name'" + ) unless @c_cols == @_; + + $call_cond = {}; + @{$call_cond}{@c_cols} = @_; + } + + my %related; + for my $key (keys %$call_cond) { + if ( + my $keyref = ref($call_cond->{$key}) + and + my $relinfo = $rsrc->relationship_info($key) + ) { + my $val = delete $call_cond->{$key}; + + next if $keyref eq 'ARRAY'; # has_many for multi_create + + my $rel_q = $rsrc->_resolve_condition( + $relinfo->{cond}, $val, $key, $key + ); + die "Can't handle complex relationship conditions in find" if ref($rel_q) ne 'HASH'; @related{keys %$rel_q} = values %$rel_q; } } - if (my @keys = keys %related) { - @{$input_query}{@keys} = values %related; - } + # relationship conditions take precedence (?) + @{$call_cond}{keys %related} = values %related; - # Build the final query: Default to the disjunction of the unique queries, - # but allow the input query in case the ResultSet defines the query or the - # user is abusing find my $alias = exists $attrs->{alias} ? $attrs->{alias} : $self->{attrs}{alias}; - my $query; - if (exists $attrs->{key}) { - my @unique_cols = $self->result_source->unique_constraint_columns($attrs->{key}); - my $unique_query = $self->_build_unique_query($input_query, \@unique_cols); - $query = $self->_add_alias($unique_query, $alias); + my $final_cond; + if (defined $constraint_name) { + $final_cond = $self->_qualify_cond_columns ( + + $self->_build_unique_cond ( + $constraint_name, + $call_cond, + ), + + $alias, + ); } elsif ($self->{attrs}{accessor} and $self->{attrs}{accessor} eq 'single') { # This means that we got here after a merger of relationship conditions @@ -532,14 +832,28 @@ sub find { # relationship } else { - my @unique_queries = $self->_unique_queries($input_query, $attrs); - $query = @unique_queries - ? [ map { $self->_add_alias($_, $alias) } @unique_queries ] - : $self->_add_alias($input_query, $alias); + # no key was specified - fall down to heuristics mode: + # run through all unique queries registered on the resultset, and + # 'OR' all qualifying queries together + my (@unique_queries, %seen_column_combinations); + for my $c_name ($rsrc->unique_constraint_names) { + next if $seen_column_combinations{ + join "\x00", sort $rsrc->unique_constraint_columns($c_name) + }++; + + push @unique_queries, try { + $self->_build_unique_cond ($c_name, $call_cond, 'croak_on_nulls') + } || (); + } + + $final_cond = @unique_queries + ? [ map { $self->_qualify_cond_columns($_, $alias) } @unique_queries ] + : $self->_non_unique_find_fallback ($call_cond, $attrs) + ; } - # Run the query - my $rs = $self->search ($query, {result_class => $self->result_class, %$attrs}); + # Run the query, passing the result_class since it should propagate for find + my $rs = $self->search ($final_cond, {result_class => $self->result_class, %$attrs}); if ($rs->_resolved_attrs->{collapse}) { my $row = $rs->next; carp "Query returned more than one row" if $rs->next; @@ -550,80 +864,97 @@ sub find { } } -# _add_alias +# This is a stop-gap method as agreed during the discussion on find() cleanup: +# http://lists.scsys.co.uk/pipermail/dbix-class/2010-October/009535.html +# +# It is invoked when find() is called in legacy-mode with insufficiently-unique +# condition. It is provided for overrides until a saner way forward is devised # -# Add the specified alias to the specified query hash. A copy is made so the -# original query is not modified. +# *NOTE* This is not a public method, and it's *GUARANTEED* to disappear down +# the road. Please adjust your tests accordingly to catch this situation early +# DBIx::Class::ResultSet->can('_non_unique_find_fallback') is reasonable +# +# The method will not be removed without an adequately complete replacement +# for strict-mode enforcement +sub _non_unique_find_fallback { + my ($self, $cond, $attrs) = @_; + + return $self->_qualify_cond_columns( + $cond, + exists $attrs->{alias} + ? $attrs->{alias} + : $self->{attrs}{alias} + ); +} -sub _add_alias { - my ($self, $query, $alias) = @_; - my %aliased = %$query; - foreach my $col (grep { ! m/\./ } keys %aliased) { - $aliased{"$alias.$col"} = delete $aliased{$col}; +sub _qualify_cond_columns { + my ($self, $cond, $alias) = @_; + + my %aliased = %$cond; + for (keys %aliased) { + $aliased{"$alias.$_"} = delete $aliased{$_} + if $_ !~ /\./; } return \%aliased; } -# _unique_queries -# -# Build a list of queries which satisfy unique constraints. - -sub _unique_queries { - my ($self, $query, $attrs) = @_; - - my @constraint_names = exists $attrs->{key} - ? ($attrs->{key}) - : $self->result_source->unique_constraint_names; - - my $where = $self->_collapse_cond($self->{attrs}{where} || {}); - my $num_where = scalar keys %$where; - - my (@unique_queries, %seen_column_combinations); - foreach my $name (@constraint_names) { - my @constraint_cols = $self->result_source->unique_constraint_columns($name); +sub _build_unique_cond { + my ($self, $constraint_name, $extra_cond, $croak_on_null) = @_; - my $constraint_sig = join "\x00", sort @constraint_cols; - next if $seen_column_combinations{$constraint_sig}++; + my @c_cols = $self->result_source->unique_constraint_columns($constraint_name); - my $unique_query = $self->_build_unique_query($query, \@constraint_cols); + # combination may fail if $self->{cond} is non-trivial + my ($final_cond) = try { + $self->_merge_with_rscond ($extra_cond) + } catch { + +{ %$extra_cond } + }; - my $num_cols = scalar @constraint_cols; - my $num_query = scalar keys %$unique_query; + # trim out everything not in $columns + $final_cond = { map { + exists $final_cond->{$_} + ? ( $_ => $final_cond->{$_} ) + : () + } @c_cols }; - my $total = $num_query + $num_where; - if ($num_query && ($num_query == $num_cols || $total == $num_cols)) { - # The query is either unique on its own or is unique in combination with - # the existing where clause - push @unique_queries, $unique_query; - } + if (my @missing = grep + { ! ($croak_on_null ? defined $final_cond->{$_} : exists $final_cond->{$_}) } + (@c_cols) + ) { + $self->throw_exception( sprintf ( "Unable to satisfy requested constraint '%s', no values for column(s): %s", + $constraint_name, + join (', ', map { "'$_'" } @missing), + ) ); } - return @unique_queries; -} - -# _build_unique_query -# -# Constrain the specified query hash based on the specified column names. - -sub _build_unique_query { - my ($self, $query, $unique_cols) = @_; + if ( + !$croak_on_null + and + !$ENV{DBIC_NULLABLE_KEY_NOWARN} + and + my @undefs = sort grep { ! defined $final_cond->{$_} } (keys %$final_cond) + ) { + carp_unique ( sprintf ( + "NULL/undef values supplied for requested unique constraint '%s' (NULL " + . 'values in column(s): %s). This is almost certainly not what you wanted, ' + . 'though you can set DBIC_NULLABLE_KEY_NOWARN to disable this warning.', + $constraint_name, + join (', ', map { "'$_'" } @undefs), + )); + } - return { - map { $_ => $query->{$_} } - grep { exists $query->{$_} } - @$unique_cols - }; + return $final_cond; } =head2 search_related =over 4 -=item Arguments: $rel, $cond, \%attrs? +=item Arguments: $rel_name, $cond?, L<\%attrs?|/ATTRIBUTES> -=item Return Value: $new_resultset +=item Return Value: L<$resultset|/search> (scalar context) | L<@result_objs|DBIx::Class::Manual::ResultClass> (list context) =back @@ -634,6 +965,11 @@ sub _build_unique_query { Searches the specified relationship, optionally specifying a condition and attributes for matching records. See L for more information. +In list context, C<< ->all() >> is called implicitly on the resultset, thus +returning a list of result objects instead. To avoid that, use L. + +See also L. + =cut sub search_related { @@ -657,7 +993,7 @@ sub search_related_rs { =item Arguments: none -=item Return Value: $cursor +=item Return Value: L<$cursor|DBIx::Class::Cursor> =back @@ -667,30 +1003,31 @@ L for more information. =cut sub cursor { - my ($self) = @_; - - my $attrs = $self->_resolved_attrs_copy; + my $self = shift; - return $self->{cursor} - ||= $self->result_source->storage->select($attrs->{from}, $attrs->{select}, - $attrs->{where},$attrs); + return $self->{cursor} ||= do { + my $attrs = $self->_resolved_attrs; + $self->result_source->storage->select( + $attrs->{from}, $attrs->{select}, $attrs->{where}, $attrs + ); + }; } =head2 single =over 4 -=item Arguments: $cond? +=item Arguments: L<$cond?|DBIx::Class::SQLMaker> -=item Return Value: $row_object? +=item Return Value: L<$result|DBIx::Class::Manual::ResultClass> | undef =back my $cd = $schema->resultset('CD')->single({ year => 2001 }); Inflates the first result without creating a cursor if the resultset has -any records in it; if not returns nothing. Used by L as a lean version of -L. +any records in it; if not returns C. Used by L as a lean version +of L. While this method can take an optional search condition (just like L) being a fast-code-path it does not recognize search attributes. If you need to @@ -725,13 +1062,11 @@ sub single { $self->throw_exception('single() only takes search conditions, no attributes. You want ->search( $cond, $attrs )->single()'); } - my $attrs = $self->_resolved_attrs_copy; + my $attrs = { %{$self->_resolved_attrs} }; - if ($attrs->{collapse}) { - $self->throw_exception( - 'single() can not be used on resultsets prefetching has_many. Use find( \%cond ) or next() instead' - ); - } + $self->throw_exception( + 'single() can not be used on resultsets collapsing a has_many. Use find( \%cond ) or next() instead' + ) if $attrs->{collapse}; if ($where) { if (defined $attrs->{where}) { @@ -745,56 +1080,17 @@ sub single { } } -# XXX: Disabled since it doesn't infer uniqueness in all cases -# unless ($self->_is_unique_query($attrs->{where})) { -# carp "Query not guaranteed to return a single row" -# . "; please declare your unique constraints or use search instead"; -# } - - my @data = $self->result_source->storage->select_single( + my $data = [ $self->result_source->storage->select_single( $attrs->{from}, $attrs->{select}, $attrs->{where}, $attrs - ); + )]; - return @data - ? ($self->_construct_objects(@data))[0] - : undef - ; + return undef unless @$data; + $self->{_stashed_rows} = [ $data ]; + $self->_construct_results->[0]; } -# _is_unique_query -# -# Try to determine if the specified query is guaranteed to be unique, based on -# the declared unique constraints. - -sub _is_unique_query { - my ($self, $query) = @_; - - my $collapsed = $self->_collapse_query($query); - my $alias = $self->{attrs}{alias}; - - foreach my $name ($self->result_source->unique_constraint_names) { - my @unique_cols = map { - "$alias.$_" - } $self->result_source->unique_constraint_columns($name); - - # Count the values for each unique column - my %seen = map { $_ => 0 } @unique_cols; - - foreach my $key (keys %$collapsed) { - my $aliased = $key =~ /\./ ? $key : "$alias.$key"; - next unless exists $seen{$aliased}; # Additional constraints are okay - $seen{$aliased} = scalar keys %{ $collapsed->{$key} }; - } - - # If we get 0 or more than 1 value for a column, it's not necessarily unique - return 1 unless grep { $_ != 1 } values %seen; - } - - return 0; -} - # _collapse_query # # Recursively collapse the query, accumulating values for each column. @@ -831,9 +1127,9 @@ sub _collapse_query { =over 4 -=item Arguments: $cond? +=item Arguments: L<$cond?|DBIx::Class::SQLMaker> -=item Return Value: $resultsetcolumn +=item Return Value: L<$resultsetcolumn|DBIx::Class::ResultSetColumn> =back @@ -853,9 +1149,9 @@ sub get_column { =over 4 -=item Arguments: $cond, \%attrs? +=item Arguments: L<$cond|DBIx::Class::SQLMaker>, L<\%attrs?|/ATTRIBUTES> -=item Return Value: $resultset (scalar context), @row_objs (list context) +=item Return Value: L<$resultset|/search> (scalar context) | L<@result_objs|DBIx::Class::Manual::ResultClass> (list context) =back @@ -881,7 +1177,7 @@ instead. An example conversion is: sub search_like { my $class = shift; - carp ( + carp_unique ( 'search_like() is deprecated and will be removed in DBIC version 0.09.' .' Instead use ->search({ x => { -like => "y%" } })' .' (note the outer pair of {}s - they are important!)' @@ -898,7 +1194,7 @@ sub search_like { =item Arguments: $first, $last -=item Return Value: $resultset (scalar context), @row_objs (list context) +=item Return Value: L<$resultset|/search> (scalar context) | L<@result_objs|DBIx::Class::Manual::ResultClass> (list context) =back @@ -916,9 +1212,7 @@ sub slice { $attrs->{offset} = $self->{attrs}{offset} || 0; $attrs->{offset} += $min; $attrs->{rows} = ($max ? ($max - $min + 1) : 1); - return $self->search(undef(), $attrs); - #my $slice = (ref $self)->new($self->result_source, $attrs); - #return (wantarray ? $slice->all : $slice); + return $self->search(undef, $attrs); } =head2 next @@ -927,7 +1221,7 @@ sub slice { =item Arguments: none -=item Return Value: $result? +=item Return Value: L<$result|DBIx::Class::Manual::ResultClass> | undef =back @@ -948,319 +1242,292 @@ first record from the resultset. sub next { my ($self) = @_; + if (my $cache = $self->get_cache) { $self->{all_cache_position} ||= 0; return $cache->[$self->{all_cache_position}++]; } + if ($self->{attrs}{cache}) { + delete $self->{pager}; $self->{all_cache_position} = 1; return ($self->all)[0]; } - if ($self->{stashed_objects}) { - my $obj = shift(@{$self->{stashed_objects}}); - delete $self->{stashed_objects} unless @{$self->{stashed_objects}}; - return $obj; - } - my @row = ( - exists $self->{stashed_row} - ? @{delete $self->{stashed_row}} - : $self->cursor->next - ); - return undef unless (@row); - my ($row, @more) = $self->_construct_objects(@row); - $self->{stashed_objects} = \@more if @more; - return $row; -} - -# takes a single DBI-row of data and coinstructs as many objects -# as the resultset attributes call for. -# This can be a bit of an action at a distance - it takes as an argument -# the *current* cursor-row (already taken off the $sth), but if -# collapsing is requested it will keep advancing the cursor either -# until the current row-object is assembled (the collapser was able to -# order the result sensibly) OR until the cursor is exhausted (an -# unordered collapsing resultset effectively triggers ->all) - -# FIXME: why the *FUCK* do we pass around DBI data by copy?! Sadly needs -# assessment before changing... -# -sub _construct_objects { - my ($self, @row) = @_; - - my $attrs = $self->_resolved_attrs; - my $keep_collapsing = $attrs->{collapse}; - my $res_index; -=begin - do { - my $me_pref_col = $attrs->{_row_parser}->($row_ref); + return shift(@{$self->{_stashed_results}}) if @{ $self->{_stashed_results}||[] }; - my $container; - if ($keep_collapsing) { + $self->{_stashed_results} = $self->_construct_results + or return undef; - # FIXME - we should be able to remove these 2 checks after the design validates - $self->throw_exception ('Collapsing without a top-level collapse-set... can not happen') - unless @{$me_ref_col->[2]}; - $self->throw_exception ('Top-level collapse-set contains a NULL-value... can not happen') - if grep { ! defined $_ } @{$me_pref_col->[2]}; - - my $main_ident = join "\x00", @{$me_pref_col->[2]}; + return shift @{$self->{_stashed_results}}; +} - if (! $res_index->{$main_ident}) { - # this is where we bail out IFF we are ordered, and the $main_ident changes +# Constructs as many results as it can in one pass while respecting +# cursor laziness. Several modes of operation: +# +# * Always builds everything present in @{$self->{_stashed_rows}} +# * If called with $fetch_all true - pulls everything off the cursor and +# builds all result structures (or objects) in one pass +# * If $self->_resolved_attrs->{collapse} is true, checks the order_by +# and if the resultset is ordered properly by the left side: +# * Fetches stuff off the cursor until the "master object" changes, +# and saves the last extra row (if any) in @{$self->{_stashed_rows}} +# OR +# * Just fetches, and collapses/constructs everything as if $fetch_all +# was requested (there is no other way to collapse except for an +# eager cursor) +# * If no collapse is requested - just get the next row, construct and +# return +sub _construct_results { + my ($self, $fetch_all) = @_; - $res_index->{$main_ident} = { - all_me_pref => [, - index => scalar keys %$res_index, - }; - } - } + my $rsrc = $self->result_source; + my $attrs = $self->_resolved_attrs; - - - $container = $res_index->{$main_ident}{container}; - }; - - push @$container, [ @{$me_pref_col}[0,1] ]; - - - - } while ( - $keep_collapsing - && - do { $row_ref = [$self->cursor->next]; $self->{stashed_row} = $row_ref if @$row_ref; scalar @$row_ref } - ); - - # attempt collapse all rows with same collapse identity - if (@to_collapse > 1) { - my @collapsed; - while (@to_collapse) { - $self->_merge_result(\@collapsed, shift @to_collapse); - } + if ( + ! $fetch_all + and + ! $attrs->{order_by} + and + $attrs->{collapse} + and + my @pcols = $rsrc->primary_columns + ) { + # default order for collapsing unless the user asked for something + $attrs->{order_by} = [ map { join '.', $attrs->{alias}, $_} @pcols ]; + $attrs->{_ordered_for_collapse} = 1; + $attrs->{_order_is_artificial} = 1; } -=cut - my $mepref_structs = $self->_collapse_result(\@row) - or return (); + # this will be used as both initial raw-row collector AND as a RV of + # _construct_results. Not regrowing the array twice matters a lot... + # a surprising amount actually + my $rows = delete $self->{_stashed_rows}; - my $rsrc = $self->result_source; - my $res_class = $self->result_class; - my $inflator = $res_class->can ('inflate_result'); + my $cursor; # we may not need one at all - my @objs = map { - $res_class->$inflator ($rsrc, @$_) - } (@$mepref_structs); + my $did_fetch_all = $fetch_all; - if (my $f = $attrs->{record_filter}) { - @objs = map { $f->($_) } @objs; + if ($fetch_all) { + # FIXME SUBOPTIMAL - we can do better, cursor->next/all (well diff. methods) should return a ref + $rows = [ ($rows ? @$rows : ()), $self->cursor->all ]; } - - return @objs; -} - -=begin - -# two arguments: $as_proto is an arrayref of column names, -# $row_ref is an arrayref of the data. If none of the row data -# is defined we return undef (that's copied from the old -# _collapse_result). Next we decide whether we need to collapse -# the resultset (i.e. we prefetch something) or not. $collapse -# indicates that. The do-while loop will run once if we do not need -# to collapse the result and will run as long as _merge_result returns -# a true value. It will return undef if the current added row does not -# match the previous row. A bit of stashing and cursor magic is -# required so that the cursor is not mixed up. - -# "$rows" is a bit misleading. In the end, there should only be one -# element in this arrayref. - -sub _collapse_result { - my ( $self, $as_proto, $row_ref ) = @_; - my $has_def; - for (@$row_ref) { - if ( defined $_ ) { - $has_def++; - last; - } + elsif( $attrs->{collapse} ) { + + # a cursor will need to be closed over in case of collapse + $cursor = $self->cursor; + + $attrs->{_ordered_for_collapse} = ( + ( + $attrs->{order_by} + and + $rsrc->schema + ->storage + ->_main_source_order_by_portion_is_stable($rsrc, $attrs->{order_by}, $attrs->{where}) + ) ? 1 : 0 + ) unless defined $attrs->{_ordered_for_collapse}; + + if (! $attrs->{_ordered_for_collapse}) { + $did_fetch_all = 1; + + # instead of looping over ->next, use ->all in stealth mode + # *without* calling a ->reset afterwards + # FIXME ENCAPSULATION - encapsulation breach, cursor method additions pending + if (! $cursor->{_done}) { + $rows = [ ($rows ? @$rows : ()), $cursor->all ]; + $cursor->{_done} = 1; + } } - return undef unless $has_def; - - my $collapse = $self->_resolved_attrs->{collapse}; - my $rows = []; - my @row = @$row_ref; - do { - my $i = 0; - my $row = { map { $_ => $row[ $i++ ] } @$as_proto }; - $row = $self->result_source->_parse_row($row, $collapse); - unless ( scalar @$rows ) { - push( @$rows, $row ); - } - $collapse = undef unless ( $self->_merge_result( $rows, $row ) ); - } while ( - $collapse - && do { @row = $self->cursor->next; $self->{stashed_row} = \@row if @row; } - ); - - return $rows->[0]; - -} + } -# _merge_result accepts an arrayref of rows objects (again, an arrayref of two elements) -# and a row object which should be merged into the first object. -# First we try to find out whether $row is already in $rows. If this is the case -# we try to merge them by iteration through their relationship data. We call -# _merge_result again on them, so they get merged. + if (! $did_fetch_all and ! @{$rows||[]} ) { + # FIXME SUBOPTIMAL - we can do better, cursor->next/all (well diff. methods) should return a ref + $cursor ||= $self->cursor; + if (scalar (my @r = $cursor->next) ) { + $rows = [ \@r ]; + } + } -# If we don't find the $row in $rows, we append it to $rows and return undef. -# _merge_result returns 1 otherwise (i.e. $row has been found in $rows). + return undef unless @{$rows||[]}; -sub _merge_result { - my ( $self, $rows, $row ) = @_; - my ( $columns, $rels ) = @$row; - my $found = undef; - foreach my $seen (@$rows) { - my $match = 1; - foreach my $column ( keys %$columns ) { - if ( defined $seen->[0]->{$column} ^ defined $columns->{$column} - or defined $columns->{$column} - && $seen->[0]->{$column} ne $columns->{$column} ) - { + # sanity check - people are too clever for their own good + if ($attrs->{collapse} and my $aliastypes = $attrs->{_last_sqlmaker_alias_map} ) { - $match = 0; - last; - } - } - if ($match) { - $found = $seen; - last; - } + my $multiplied_selectors; + for my $sel_alias ( grep { $_ ne $attrs->{alias} } keys %{ $aliastypes->{selecting} } ) { + if ( + $aliastypes->{multiplying}{$sel_alias} + or + $aliastypes->{premultiplied}{$sel_alias} + ) { + $multiplied_selectors->{$_} = 1 for values %{$aliastypes->{selecting}{$sel_alias}{-seen_columns}} + } } - if ($found) { - foreach my $rel ( keys %$rels ) { - my $old_rows = $found->[1]->{$rel}; - $self->_merge_result( - ref $found->[1]->{$rel}->[0] eq 'HASH' ? [ $found->[1]->{$rel} ] - : $found->[1]->{$rel}, - ref $rels->{$rel}->[0] eq 'HASH' ? [ $rels->{$rel}->[0], $rels->{$rel}->[1] ] - : $rels->{$rel}->[0] - ); - - my $attrs = $self->_resolved_attrs; - my ($keep_collapsing, $set_ident) = @{$attrs}{qw/collapse _collapse_ident/}; - - # FIXME this is temporary, need to calculate in _resolved_attrs - $set_ident ||= { me => [ $self->result_source->_pri_cols ], pref => {} }; - - my @cur_row = @$row_ref; - my (@to_collapse, $last_ident); - - do { - my $row_hr = { map { $as_proto->[$_] => $cur_row[$_] } (0 .. $#$as_proto) }; - - # see if we are switching to another object - # this can be turned off and things will still work - # since _merge_prefetch knows about _collapse_ident -# my $cur_ident = [ @{$row_hr}{@$set_ident} ]; - my $cur_ident = []; - $last_ident ||= $cur_ident; - -# if ($keep_collapsing = Test::Deep::eq_deeply ($cur_ident, $last_ident)) { -# push @to_collapse, $self->result_source->_parse_row ( -# $row_hr, -# ); -# } - } while ( - $keep_collapsing - && - do { @cur_row = $self->cursor->next; $self->{stashed_row} = \@cur_row if @cur_row; } - ); - die Dumper \@to_collapse; + for my $i (0 .. $#{$attrs->{as}} ) { + my $sel = $attrs->{select}[$i]; + if (ref $sel eq 'SCALAR') { + $sel = $$sel; + } + elsif( ref $sel eq 'REF' and ref $$sel eq 'ARRAY' ) { + $sel = $$sel->[0]; + } - # attempt collapse all rows with same collapse identity - if (@to_collapse > 1) { - my @collapsed; - while (@to_collapse) { - $self->_merge_result(\@collapsed, shift @to_collapse); + $self->throw_exception( + 'Result collapse not possible - selection from a has_many source redirected to the main object' + ) if ($multiplied_selectors->{$sel} and $attrs->{as}[$i] !~ /\./); } - @to_collapse = @collapsed; } - # still didn't fully collapse - $self->throw_exception ('Resultset collapse failed (theoretically impossible). Maybe a wrong collapse_ident...?') - if (@to_collapse > 1); - - return $to_collapse[0]; -} - + # hotspot - skip the setter + my $res_class = $self->_result_class; -# two arguments: $as_proto is an arrayref of 'as' column names, -# $row_ref is an arrayref of the data. The do-while loop will run -# once if we do not need to collapse the result and will run as long as -# _merge_result returns a true value. It will return undef if the -# current added row does not match the previous row, which in turn -# means we need to stash the row for the subsequent ->next call -sub _collapse_result { - my ( $self, $as_proto, $row_ref ) = @_; + my $inflator_cref = $self->{_result_inflator}{cref} ||= do { + $res_class->can ('inflate_result') + or $self->throw_exception("Inflator $res_class does not provide an inflate_result() method"); + }; - my $attrs = $self->_resolved_attrs; - my ($keep_collapsing, $set_ident) = @{$attrs}{qw/collapse _collapse_ident/}; + my $infmap = $attrs->{as}; - die Dumper [$as_proto, $row_ref, $keep_collapsing, $set_ident ]; + $self->{_result_inflator}{is_core_row} = ( ( + $inflator_cref + == + ( \&DBIx::Class::Row::inflate_result || die "No ::Row::inflate_result() - can't happen" ) + ) ? 1 : 0 ) unless defined $self->{_result_inflator}{is_core_row}; + $self->{_result_inflator}{is_hri} = ( ( + ! $self->{_result_inflator}{is_core_row} + and + $inflator_cref == ( + require DBIx::Class::ResultClass::HashRefInflator + && + DBIx::Class::ResultClass::HashRefInflator->can('inflate_result') + ) + ) ? 1 : 0 ) unless defined $self->{_result_inflator}{is_hri}; + + + if (! $attrs->{_related_results_construction}) { + # construct a much simpler array->hash folder for the one-table cases right here + if ($self->{_result_inflator}{is_hri}) { + for my $r (@$rows) { + $r = { map { $infmap->[$_] => $r->[$_] } 0..$#$infmap }; + } + } + # FIXME SUBOPTIMAL this is a very very very hot spot + # while rather optimal we can *still* do much better, by + # building a smarter Row::inflate_result(), and + # switch to feeding it data via a much leaner interface + # + # crude unscientific benchmarking indicated the shortcut eval is not worth it for + # this particular resultset size + elsif (@$rows < 60) { + for my $r (@$rows) { + $r = $inflator_cref->($res_class, $rsrc, { map { $infmap->[$_] => $r->[$_] } (0..$#$infmap) } ); + } + } + else { + eval sprintf ( + '$_ = $inflator_cref->($res_class, $rsrc, { %s }) for @$rows', + join (', ', map { "\$infmap->[$_] => \$_->[$_]" } 0..$#$infmap ) + ); + } + } + else { + my $parser_type = + $self->{_result_inflator}{is_hri} ? 'hri' + : $self->{_result_inflator}{is_core_row} ? 'classic_pruning' + : 'classic_nonpruning' + ; - my @cur_row = @$row_ref; - my (@to_collapse, $last_ident); + # $args and $attrs to _mk_row_parser are separated to delineate what is + # core collapser stuff and what is dbic $rs specific + @{$self->{_row_parser}{$parser_type}}{qw(cref nullcheck)} = $rsrc->_mk_row_parser({ + eval => 1, + inflate_map => $infmap, + collapse => $attrs->{collapse}, + premultiplied => $attrs->{_main_source_premultiplied}, + hri_style => $self->{_result_inflator}{is_hri}, + prune_null_branches => $self->{_result_inflator}{is_hri} || $self->{_result_inflator}{is_core_row}, + }, $attrs) unless $self->{_row_parser}{$parser_type}{cref}; + + # column_info metadata historically hasn't been too reliable. + # We need to start fixing this somehow (the collapse resolver + # can't work without it). Add an explicit check for the *main* + # result, hopefully this will gradually weed out such errors + # + # FIXME - this is a temporary kludge that reduces performance + # It is however necessary for the time being + my ($unrolled_non_null_cols_to_check, $err); + + if (my $check_non_null_cols = $self->{_row_parser}{$parser_type}{nullcheck} ) { + + $err = + 'Collapse aborted due to invalid ResultSource metadata - the following ' + . 'selections are declared non-nullable but NULLs were retrieved: ' + ; - do { - my $row_hr = { map { $as_proto->[$_] => $cur_row[$_] } (0 .. $#$as_proto) }; + my @violating_idx; + COL: for my $i (@$check_non_null_cols) { + ! defined $_->[$i] and push @violating_idx, $i and next COL for @$rows; + } - # see if we are switching to another object - # this can be turned off and things will still work - # since _merge_prefetch knows about _collapse_ident -# my $cur_ident = [ @{$row_hr}{@$set_ident} ]; - my $cur_ident = []; - $last_ident ||= $cur_ident; + $self->throw_exception( $err . join (', ', map { "'$infmap->[$_]'" } @violating_idx ) ) + if @violating_idx; -# if ($keep_collapsing = eq_deeply ($cur_ident, $last_ident)) { -# push @to_collapse, $self->result_source->_parse_row ( -# $row_hr, -# ); -# } - } while ( - $keep_collapsing - && - do { @cur_row = $self->cursor->next; $self->{stashed_row} = \@cur_row if @cur_row; } - ); + $unrolled_non_null_cols_to_check = join (',', @$check_non_null_cols); + } - # attempt collapse all rows with same collapse identity + my $next_cref = + ($did_fetch_all or ! $attrs->{collapse}) ? undef + : defined $unrolled_non_null_cols_to_check ? eval sprintf <<'EOS', $unrolled_non_null_cols_to_check +sub { + # FIXME SUBOPTIMAL - we can do better, cursor->next/all (well diff. methods) should return a ref + my @r = $cursor->next or return; + if (my @violating_idx = grep { ! defined $r[$_] } (%s) ) { + $self->throw_exception( $err . join (', ', map { "'$infmap->[$_]'" } @violating_idx ) ) + } + \@r } -=cut +EOS + : sub { + # FIXME SUBOPTIMAL - we can do better, cursor->next/all (well diff. methods) should return a ref + my @r = $cursor->next or return; + \@r + } + ; -# Takes an arrayref of me/pref pairs and a new me/pref pair that should -# be merged on a preexisting matching me (or should be pushed into $merged -# as a new me/pref pair for further invocations). It should be possible to -# use this function to collapse complete ->all results, provided _collapse_result() is adjusted -# to provide everything to this sub not to barf when $merged contains more than one -# arrayref) -sub _merge_prefetch { - my ($self, $merged, $next_row) = @_; + $self->{_row_parser}{$parser_type}{cref}->( + $rows, + $next_cref ? ( $next_cref, $self->{_stashed_rows} = [] ) : (), + ); - unless (@$merged) { - push @$merged, $next_row; - return; + # Special-case multi-object HRI - there is no $inflator_cref pass + unless ($self->{_result_inflator}{is_hri}) { + $_ = $inflator_cref->($res_class, $rsrc, @$_) for @$rows + } } + # The @$rows check seems odd at first - why wouldn't we want to warn + # regardless? The issue is things like find() etc, where the user + # *knows* only one result will come back. In these cases the ->all + # is not a pessimization, but rather something we actually want + carp_unique( + 'Unable to properly collapse has_many results in iterator mode due ' + . 'to order criteria - performed an eager cursor slurp underneath. ' + . 'Consider using ->all() instead' + ) if ( ! $fetch_all and @$rows > 1 ); + + return $rows; } =head2 result_source =over 4 -=item Arguments: $result_source? +=item Arguments: L<$result_source?|DBIx::Class::ResultSource> -=item Return Value: $result_source +=item Return Value: L<$result_source|DBIx::Class::ResultSource> =back @@ -1277,7 +1544,7 @@ is derived. =back -An accessor for the class to use when creating row objects. Defaults to +An accessor for the class to use when creating result objects. Defaults to C<< result_source->result_class >> - which in most cases is the name of the L<"table"|DBIx::Class::Manual::Glossary/"ResultSource"> class. @@ -1291,9 +1558,22 @@ in the original source class will not run. sub result_class { my ($self, $result_class) = @_; if ($result_class) { - $self->ensure_class_loaded($result_class); + + # don't fire this for an object + $self->ensure_class_loaded($result_class) + unless ref($result_class); + + if ($self->get_cache) { + carp_unique('Changing the result_class of a ResultSet instance with cached results is a noop - the cache contents will not be altered'); + } + # FIXME ENCAPSULATION - encapsulation breach, cursor method additions pending + elsif ($self->{cursor} && $self->{cursor}{_pos}) { + $self->throw_exception('Changing the result_class of a ResultSet instance with an active cursor is not supported'); + } + $self->_result_class($result_class); - $self->{attrs}{result_class} = $result_class if ref $self; + + delete $self->{_result_inflator}; } $self->_result_class; } @@ -1302,7 +1582,7 @@ sub result_class { =over 4 -=item Arguments: $cond, \%attrs?? +=item Arguments: L<$cond|DBIx::Class::SQLMaker>, L<\%attrs?|/ATTRIBUTES> =item Return Value: $count @@ -1319,12 +1599,11 @@ sub count { return $self->search(@_)->count if @_ and defined $_[0]; return scalar @{ $self->get_cache } if $self->get_cache; - my $attrs = $self->_resolved_attrs_copy; + my $attrs = { %{ $self->_resolved_attrs } }; # this is a little optimization - it is faster to do the limit # adjustments in software, instead of a subquery - my $rows = delete $attrs->{rows}; - my $offset = delete $attrs->{offset}; + my ($rows, $offset) = delete @{$attrs}{qw/rows offset/}; my $crs; if ($self->_has_resolved_attr (qw/collapse group_by/)) { @@ -1346,9 +1625,9 @@ sub count { =over 4 -=item Arguments: $cond, \%attrs?? +=item Arguments: L<$cond|DBIx::Class::SQLMaker>, L<\%attrs?|/ATTRIBUTES> -=item Return Value: $count_rs +=item Return Value: L<$count_rs|DBIx::Class::ResultSetColumn> =back @@ -1372,10 +1651,10 @@ sub count_rs { # software based limiting can not be ported if this $rs is to be used # in a subquery itself (i.e. ->as_query) if ($self->_has_resolved_attr (qw/collapse group_by offset rows/)) { - return $self->_count_subq_rs; + return $self->_count_subq_rs($self->{_attrs}); } else { - return $self->_count_rs; + return $self->_count_rs($self->{_attrs}); } } @@ -1386,20 +1665,17 @@ sub _count_rs { my ($self, $attrs) = @_; my $rsrc = $self->result_source; - $attrs ||= $self->_resolved_attrs; my $tmp_attrs = { %$attrs }; - - # take off any limits, record_filter is cdbi, and no point of ordering a count - delete $tmp_attrs->{$_} for (qw/select as rows offset order_by record_filter/); + # take off any limits, record_filter is cdbi, and no point of ordering nor locking a count + delete @{$tmp_attrs}{qw/rows offset order_by record_filter for/}; # overwrite the selector (supplied by the storage) - $tmp_attrs->{select} = $rsrc->storage->_count_select ($rsrc, $tmp_attrs); - $tmp_attrs->{as} = 'count'; - - my $tmp_rs = $rsrc->resultset_class->new($rsrc, $tmp_attrs)->get_column ('count'); - - return $tmp_rs; + $rsrc->resultset_class->new($rsrc, { + %$tmp_attrs, + select => $rsrc->storage->_count_select ($rsrc, $attrs), + as => 'count', + })->get_column ('count'); } # @@ -1409,48 +1685,105 @@ sub _count_subq_rs { my ($self, $attrs) = @_; my $rsrc = $self->result_source; - $attrs ||= $self->_resolved_attrs_copy; my $sub_attrs = { %$attrs }; + # extra selectors do not go in the subquery and there is no point of ordering it, nor locking it + delete @{$sub_attrs}{qw/collapse columns as select order_by for/}; - # extra selectors do not go in the subquery and there is no point of ordering it - delete $sub_attrs->{$_} for qw/collapse select _prefetch_select as order_by/; - - # if we multi-prefetch we group_by primary keys only as this is what we would + # if we multi-prefetch we group_by something unique, as this is what we would # get out of the rs via ->next/->all. We *DO WANT* to clobber old group_by regardless if ( $attrs->{collapse} ) { - $sub_attrs->{group_by} = [ map { "$attrs->{alias}.$_" } ($rsrc->_pri_cols) ] + $sub_attrs->{group_by} = [ map { "$attrs->{alias}.$_" } @{ + $rsrc->_identifying_column_set || $self->throw_exception( + 'Unable to construct a unique group_by criteria properly collapsing the ' + . 'has_many prefetch before count()' + ); + } ] } - $sub_attrs->{select} = $rsrc->storage->_subq_count_select ($rsrc, $attrs); + # Calculate subquery selector + if (my $g = $sub_attrs->{group_by}) { - # this is so that the query can be simplified e.g. - # * ordering can be thrown away in things like Top limit - $sub_attrs->{-for_count_only} = 1; + my $sql_maker = $rsrc->storage->sql_maker; - my $sub_rs = $rsrc->resultset_class->new ($rsrc, $sub_attrs); + # necessary as the group_by may refer to aliased functions + my $sel_index; + for my $sel (@{$attrs->{select}}) { + $sel_index->{$sel->{-as}} = $sel + if (ref $sel eq 'HASH' and $sel->{-as}); + } - $attrs->{from} = [{ - -alias => 'count_subq', - -source_handle => $rsrc->handle, - count_subq => $sub_rs->as_query, - }]; + # anything from the original select mentioned on the group-by needs to make it to the inner selector + # also look for named aggregates referred in the having clause + # having often contains scalarrefs - thus parse it out entirely + my @parts = @$g; + if ($attrs->{having}) { + local $sql_maker->{having_bind}; + local $sql_maker->{quote_char} = $sql_maker->{quote_char}; + local $sql_maker->{name_sep} = $sql_maker->{name_sep}; + unless (defined $sql_maker->{quote_char} and length $sql_maker->{quote_char}) { + $sql_maker->{quote_char} = [ "\x00", "\xFF" ]; + # if we don't unset it we screw up retarded but unfortunately working + # 'MAX(foo.bar)' => { '>', 3 } + $sql_maker->{name_sep} = ''; + } - # the subquery replaces this - delete $attrs->{$_} for qw/where bind collapse group_by having having_bind rows offset/; + my ($lquote, $rquote, $sep) = map { quotemeta $_ } ($sql_maker->_quote_chars, $sql_maker->name_sep); + + my $having_sql = $sql_maker->_parse_rs_attrs ({ having => $attrs->{having} }); + my %seen_having; + + # search for both a proper quoted qualified string, for a naive unquoted scalarref + # and if all fails for an utterly naive quoted scalar-with-function + while ($having_sql =~ / + $rquote $sep $lquote (.+?) $rquote + | + [\s,] \w+ \. (\w+) [\s,] + | + [\s,] $lquote (.+?) $rquote [\s,] + /gx) { + my $part = $1 || $2 || $3; # one of them matched if we got here + unless ($seen_having{$part}++) { + push @parts, $part; + } + } + } - return $self->_count_rs ($attrs); -} + for (@parts) { + my $colpiece = $sel_index->{$_} || $_; -sub _bool { - return 1; + # unqualify join-based group_by's. Arcane but possible query + # also horrible horrible hack to alias a column (not a func.) + # (probably need to introduce SQLA syntax) + if ($colpiece =~ /\./ && $colpiece !~ /^$attrs->{alias}\./) { + my $as = $colpiece; + $as =~ s/\./__/; + $colpiece = \ sprintf ('%s AS %s', map { $sql_maker->_quote ($_) } ($colpiece, $as) ); + } + push @{$sub_attrs->{select}}, $colpiece; + } + } + else { + my @pcols = map { "$attrs->{alias}.$_" } ($rsrc->primary_columns); + $sub_attrs->{select} = @pcols ? \@pcols : [ 1 ]; + } + + return $rsrc->resultset_class + ->new ($rsrc, $sub_attrs) + ->as_subselect_rs + ->search ({}, { columns => { count => $rsrc->storage->_count_select ($rsrc, $attrs) } }) + ->get_column ('count'); } + =head2 count_literal +B: C is provided for Class::DBI compatibility and +should only be used in that context. See L for further info. + =over 4 -=item Arguments: $sql_fragment, @bind_values +=item Arguments: $sql_fragment, @standalone_bind_values =item Return Value: $count @@ -1469,47 +1802,33 @@ sub count_literal { shift->search_literal(@_)->count; } =item Arguments: none -=item Return Value: @objects +=item Return Value: L<@result_objs|DBIx::Class::Manual::ResultClass> =back -Returns all elements in the resultset. Called implicitly if the resultset -is returned in list context. +Returns all elements in the resultset. =cut sub all { my $self = shift; if(@_) { - $self->throw_exception("all() doesn't take any arguments, you probably wanted ->search(...)->all()"); + $self->throw_exception("all() doesn't take any arguments, you probably wanted ->search(...)->all()"); } + delete @{$self}{qw/_stashed_rows _stashed_results/}; + if (my $c = $self->get_cache) { return @$c; } - my @objects; + $self->cursor->reset; - if ($self->_resolved_attrs->{collapse}) { - # Using $self->cursor->all is really just an optimisation. - # If we're collapsing has_many prefetches it probably makes - # very little difference, and this is cleaner than hacking - # _construct_objects to survive the approach - $self->cursor->reset; - my @row = $self->cursor->next; - while (@row) { - push(@objects, $self->_construct_objects(@row)); - @row = (exists $self->{stashed_row} - ? @{delete $self->{stashed_row}} - : $self->cursor->next); - } - } else { - @objects = map { $self->_construct_objects($_) } $self->cursor->all; - } + my $objs = $self->_construct_results('fetch_all') || []; - $self->set_cache(\@objects) if $self->{attrs}{cache}; + $self->set_cache($objs) if $self->{attrs}{cache}; - return @objects; + return @$objs; } =head2 reset @@ -1530,7 +1849,8 @@ another query. sub reset { my ($self) = @_; - delete $self->{_attrs} if exists $self->{_attrs}; + + delete @{$self}{qw/_stashed_rows _stashed_results/}; $self->{all_cache_position} = 0; $self->cursor->reset; return $self; @@ -1542,12 +1862,12 @@ sub reset { =item Arguments: none -=item Return Value: $object? +=item Return Value: L<$result|DBIx::Class::Manual::ResultClass> | undef =back -Resets the resultset and returns an object for the first result (if the -resultset returns anything). +L the resultset (causing a fresh query to storage) and returns +an object for the first result (or C if the resultset is empty). =cut @@ -1566,63 +1886,142 @@ sub _rs_update_delete { my ($self, $op, $values) = @_; my $rsrc = $self->result_source; + my $storage = $rsrc->schema->storage; + + my $attrs = { %{$self->_resolved_attrs} }; + + my $join_classifications; + my ($existing_group_by) = delete @{$attrs}{qw(group_by _grouped_by_distinct)}; + + # do we need a subquery for any reason? + my $needs_subq = ( + defined $existing_group_by + or + # if {from} is unparseable wrap a subq + ref($attrs->{from}) ne 'ARRAY' + or + # limits call for a subq + $self->_has_resolved_attr(qw/rows offset/) + ); - # if a condition exists we need to strip all table qualifiers - # if this is not possible we'll force a subquery below - my $cond = $rsrc->schema->storage->_strip_cond_qualifiers ($self->{cond}); + # simplify the joinmap, so we can further decide if a subq is necessary + if (!$needs_subq and @{$attrs->{from}} > 1) { - my $needs_group_by_subq = $self->_has_resolved_attr (qw/collapse group_by -join/); - my $needs_subq = $needs_group_by_subq || (not defined $cond) || $self->_has_resolved_attr(qw/row offset/); + ($attrs->{from}, $join_classifications) = + $storage->_prune_unused_joins ($attrs); - if ($needs_group_by_subq or $needs_subq) { + # any non-pruneable non-local restricting joins imply subq + $needs_subq = defined List::Util::first { $_ ne $attrs->{alias} } keys %{ $join_classifications->{restricting} || {} }; + } - # make a new $rs selecting only the PKs (that's all we really need) - my $attrs = $self->_resolved_attrs_copy; + # check if the head is composite (by now all joins are thrown out unless $needs_subq) + $needs_subq ||= ( + (ref $attrs->{from}[0]) ne 'HASH' + or + ref $attrs->{from}[0]{ $attrs->{from}[0]{-alias} } + ); - delete $attrs->{$_} for qw/collapse select as/; - $attrs->{columns} = [ map { "$attrs->{alias}.$_" } ($self->result_source->_pri_cols) ]; + my ($cond, $guard); + # do we need anything like a subquery? + if (! $needs_subq) { + # Most databases do not allow aliasing of tables in UPDATE/DELETE. Thus + # a condition containing 'me' or other table prefixes will not work + # at all. Tell SQLMaker to dequalify idents via a gross hack. + $cond = do { + my $sqla = $rsrc->storage->sql_maker; + local $sqla->{_dequalify_idents} = 1; + \[ $sqla->_recurse_where($self->{cond}) ]; + }; + } + else { + # we got this far - means it is time to wrap a subquery + my $idcols = $rsrc->_identifying_column_set || $self->throw_exception( + sprintf( + "Unable to perform complex resultset %s() without an identifying set of columns on source '%s'", + $op, + $rsrc->source_name, + ) + ); - if ($needs_group_by_subq) { - # make sure no group_by was supplied, or if there is one - make sure it matches - # the columns compiled above perfectly. Anything else can not be sanely executed - # on most databases so croak right then and there + # make a new $rs selecting only the PKs (that's all we really need for the subq) + delete $attrs->{$_} for qw/select as collapse/; + $attrs->{columns} = [ map { "$attrs->{alias}.$_" } @$idcols ]; - if (my $g = $attrs->{group_by}) { - my @current_group_by = map - { $_ =~ /\./ ? $_ : "$attrs->{alias}.$_" } - @$g - ; + # this will be consumed by the pruner waaaaay down the stack + $attrs->{_force_prune_multiplying_joins} = 1; - if ( - join ("\x00", sort @current_group_by) - ne - join ("\x00", sort @{$attrs->{columns}} ) - ) { - $self->throw_exception ( - "You have just attempted a $op operation on a resultset which does group_by" - . ' on columns other than the primary keys, while DBIC internally needs to retrieve' - . ' the primary keys in a subselect. All sane RDBMS engines do not support this' - . ' kind of queries. Please retry the operation with a modified group_by or' - . ' without using one at all.' - ); + my $subrs = (ref $self)->new($rsrc, $attrs); + + if (@$idcols == 1) { + $cond = { $idcols->[0] => { -in => $subrs->as_query } }; + } + elsif ($storage->_use_multicolumn_in) { + # no syntax for calling this properly yet + # !!! EXPERIMENTAL API !!! WILL CHANGE !!! + $cond = $storage->sql_maker->_where_op_multicolumn_in ( + $idcols, # how do I convey a list of idents...? can binds reside on lhs? + $subrs->as_query + ), + } + else { + # if all else fails - get all primary keys and operate over a ORed set + # wrap in a transaction for consistency + # this is where the group_by/multiplication starts to matter + if ( + $existing_group_by + or + # we do not need to check pre-multipliers, since if the premulti is there, its + # parent (who is multi) will be there too + keys %{ $join_classifications->{multiplying} || {} } + ) { + # make sure if there is a supplied group_by it matches the columns compiled above + # perfectly. Anything else can not be sanely executed on most databases so croak + # right then and there + if ($existing_group_by) { + my @current_group_by = map + { $_ =~ /\./ ? $_ : "$attrs->{alias}.$_" } + @$existing_group_by + ; + + if ( + join ("\x00", sort @current_group_by) + ne + join ("\x00", sort @{$attrs->{columns}} ) + ) { + $self->throw_exception ( + "You have just attempted a $op operation on a resultset which does group_by" + . ' on columns other than the primary keys, while DBIC internally needs to retrieve' + . ' the primary keys in a subselect. All sane RDBMS engines do not support this' + . ' kind of queries. Please retry the operation with a modified group_by or' + . ' without using one at all.' + ); + } } + + $subrs = $subrs->search({}, { group_by => $attrs->{columns} }); } - else { - $attrs->{group_by} = $attrs->{columns}; + + $guard = $storage->txn_scope_guard; + + $cond = []; + for my $row ($subrs->cursor->all) { + push @$cond, { map + { $idcols->[$_] => $row->[$_] } + (0 .. $#$idcols) + }; } } + } - my $subrs = (ref $self)->new($rsrc, $attrs); + my $res = $storage->$op ( + $rsrc, + $op eq 'update' ? $values : (), + $cond, + ); - return $self->result_source->storage->_subq_update_delete($subrs, $op, $values); - } - else { - return $rsrc->storage->$op( - $rsrc, - $op eq 'update' ? $values : (), - $cond, - ); - } + $guard->commit if $guard; + + return $res; } =head2 update @@ -1631,13 +2030,30 @@ sub _rs_update_delete { =item Arguments: \%values -=item Return Value: $storage_rv +=item Return Value: $underlying_storage_rv =back Sets the specified columns in the resultset to the supplied values in a -single query. Return value will be true if the update succeeded or false -if no records were updated; exact type of success value is storage-dependent. +single query. Note that this will not run any accessor/set_column/update +triggers, nor will it update any result object instances derived from this +resultset (this includes the contents of the L +if any). See L if you need to execute any on-update +triggers or cascades defined either by you or a +L. + +The return value is a pass through of what the underlying +storage backend returned, and may vary. See L for the most +common case. + +=head3 CAVEAT + +Note that L does not process/deflate any of the values passed in. +This is unlike the corresponding L. The user must +ensure manually that any value passed to this method will stringify to +something the RDBMS knows how to deal with. A notable example is the +handling of L objects, for more info see: +L. =cut @@ -1659,8 +2075,9 @@ sub update { =back -Fetches all objects and updates them one at a time. Note that C -will run DBIC cascade triggers, while L will not. +Fetches all objects and updates them one at a time via +L. Note that C will run DBIC defined +triggers, while L will not. =cut @@ -1668,9 +2085,10 @@ sub update_all { my ($self, $values) = @_; $self->throw_exception('Values for update_all must be a hash') unless ref $values eq 'HASH'; - foreach my $obj ($self->all) { - $obj->set_columns($values)->update; - } + + my $guard = $self->result_source->schema->txn_scope_guard; + $_->update({%$values}) for $self->all; # shallow copy - update will mangle it + $guard->commit; return 1; } @@ -1680,16 +2098,20 @@ sub update_all { =item Arguments: none -=item Return Value: $storage_rv +=item Return Value: $underlying_storage_rv =back -Deletes the contents of the resultset from its result source. Note that this -will not run DBIC cascade triggers. See L if you need triggers -to run. See also L. +Deletes the rows matching this resultset in a single query. Note that this +will not run any delete triggers, nor will it alter the +L status of any result object instances +derived from this resultset (this includes the contents of the +L if any). See L if you need to +execute any on-delete triggers or cascades defined either by you or a +L. -Return value will be the amount of rows deleted; exact type of return value -is storage-dependent. +The return value is a pass through of what the underlying storage backend +returned, and may vary. See L for the most common case. =cut @@ -1711,8 +2133,9 @@ sub delete { =back -Fetches all objects and deletes them one at a time. Note that C -will run DBIC cascade triggers, while L will not. +Fetches all objects and deletes them one at a time via +L. Note that C will run DBIC defined +triggers, while L will not. =cut @@ -1721,7 +2144,9 @@ sub delete_all { $self->throw_exception('delete_all does not accept any arguments') if @_; + my $guard = $self->result_source->schema->txn_scope_guard; $_->delete for $self->all; + $guard->commit; return 1; } @@ -1729,28 +2154,55 @@ sub delete_all { =over 4 -=item Arguments: \@data; +=item Arguments: [ \@column_list, \@row_values+ ] | [ \%col_data+ ] + +=item Return Value: L<\@result_objects|DBIx::Class::Manual::ResultClass> (scalar context) | L<@result_objects|DBIx::Class::Manual::ResultClass> (list context) =back -Accepts either an arrayref of hashrefs or alternatively an arrayref of arrayrefs. -For the arrayref of hashrefs style each hashref should be a structure suitable -forsubmitting to a $resultset->create(...) method. +Accepts either an arrayref of hashrefs or alternatively an arrayref of +arrayrefs. + +=over + +=item NOTE -In void context, C in L is used -to insert the data, as this is a faster method. +The context of this method call has an important effect on what is +submitted to storage. In void context data is fed directly to fastpath +insertion routines provided by the underlying storage (most often +L), bypassing the L and +L calls on the +L class, including any +augmentation of these methods provided by components. For example if you +are using something like L to create primary +keys for you, you will find that your PKs are empty. In this case you +will have to explicitly force scalar or list context in order to create +those values. -Otherwise, each set of data is inserted into the database using -L, and the resulting objects are -accumulated into an array. The array itself, or an array reference -is returned depending on scalar or list context. +=back + +In non-void (scalar or list) context, this method is simply a wrapper +for L. Depending on list or scalar context either a list of +L objects or an arrayref +containing these objects is returned. + +When supplying data in "arrayref of arrayrefs" invocation style, the +first element should be a list of column names and each subsequent +element should be a data value in the earlier specified column order. +For example: -Example: Assuming an Artist Class that has many CDs Classes relating: + $schema->resultset("Artist")->populate([ + [ qw( artistid name ) ], + [ 100, 'A Formally Unknown Singer' ], + [ 101, 'A singer that jumped the shark two albums ago' ], + [ 102, 'An actually cool singer' ], + ]); - my $Artist_rs = $schema->resultset("Artist"); +For the arrayref of hashrefs style each hashref should be a structure +suitable for passing to L. Multi-create is also permitted with +this syntax. - ## Void Context Example - $Artist_rs->populate([ + $schema->resultset("Artist")->populate([ { artistid => 4, name => 'Manufactured Crap', cds => [ { title => 'My First CD', year => 2006 }, { title => 'Yet More Tweeny-Pop crap', year => 2007 }, @@ -1764,37 +2216,11 @@ Example: Assuming an Artist Class that has many CDs Classes relating: }, ]); - ## Array Context Example - my ($ArtistOne, $ArtistTwo, $ArtistThree) = $Artist_rs->populate([ - { name => "Artist One"}, - { name => "Artist Two"}, - { name => "Artist Three", cds=> [ - { title => "First CD", year => 2007}, - { title => "Second CD", year => 2008}, - ]} - ]); - - print $ArtistOne->name; ## response is 'Artist One' - print $ArtistThree->cds->count ## reponse is '2' - -For the arrayref of arrayrefs style, the first element should be a list of the -fieldsnames to which the remaining elements are rows being inserted. For -example: - - $Arstist_rs->populate([ - [qw/artistid name/], - [100, 'A Formally Unknown Singer'], - [101, 'A singer that jumped the shark two albums ago'], - [102, 'An actually cool singer'], - ]); - -Please note an important effect on your data when choosing between void and -wantarray context. Since void context goes straight to C in -L this will skip any component that is overriding -C. So if you are using something like L to -create primary keys for you, you will find that your PKs are empty. In this -case you will have to use the wantarray context in order to create those -values. +If you attempt a void-context multi-create as in the example above (each +Artist also has the related list of CDs), and B supply the +necessary autoinc foreign key information, this method will proxy to the +less efficient L, and then throw the Result objects away. In this +case there are obviously no benefits to using this method over L. =cut @@ -1804,27 +2230,29 @@ sub populate { # cruft placed in standalone method my $data = $self->_normalize_populate_args(@_); + return unless @$data; + if(defined wantarray) { - my @created; - foreach my $item (@$data) { - push(@created, $self->create($item)); - } + my @created = map { $self->create($_) } @$data; return wantarray ? @created : \@created; - } else { + } + else { my $first = $data->[0]; # if a column is a registered relationship, and is a non-blessed hash/array, consider # it relationship data my (@rels, @columns); + my $rsrc = $self->result_source; + my $rels = { map { $_ => $rsrc->relationship_info($_) } $rsrc->relationships }; for (keys %$first) { my $ref = ref $first->{$_}; - $self->result_source->has_relationship($_) && ($ref eq 'ARRAY' or $ref eq 'HASH') + $rels->{$_} && ($ref eq 'ARRAY' or $ref eq 'HASH') ? push @rels, $_ : push @columns, $_ ; } - my @pks = $self->result_source->primary_columns; + my @pks = $rsrc->primary_columns; ## do the belongs_to relationships foreach my $index (0..$#$data) { @@ -1842,11 +2270,12 @@ sub populate { foreach my $rel (@rels) { next unless ref $data->[$index]->{$rel} eq "HASH"; my $result = $self->related_resultset($rel)->create($data->[$index]->{$rel}); - my ($reverse) = keys %{$self->result_source->reverse_relationship_info($rel)}; + my ($reverse_relname, $reverse_relinfo) = %{$rsrc->reverse_relationship_info($rel)}; my $related = $result->result_source->_resolve_condition( - $result->result_source->relationship_info($reverse)->{cond}, + $reverse_relinfo->{cond}, $self, $result, + $rel, ); delete $data->[$index]->{$rel}; @@ -1857,33 +2286,33 @@ sub populate { } ## inherit the data locked in the conditions of the resultset - my ($rs_data) = $self->_merge_cond_with_data({}); + my ($rs_data) = $self->_merge_with_rscond({}); delete @{$rs_data}{@columns}; - my @inherit_cols = keys %$rs_data; - my @inherit_data = values %$rs_data; ## do bulk insert on current row - $self->result_source->storage->insert_bulk( - $self->result_source, - [@columns, @inherit_cols], - [ map { [ @$_{@columns}, @inherit_data ] } @$data ], + $rsrc->storage->insert_bulk( + $rsrc, + [@columns, keys %$rs_data], + [ map { [ @$_{@columns}, values %$rs_data ] } @$data ], ); ## do the has_many relationships foreach my $item (@$data) { + my $main_row; + foreach my $rel (@rels) { - next unless $item->{$rel} && ref $item->{$rel} eq "ARRAY"; + next unless ref $item->{$rel} eq "ARRAY" && @{ $item->{$rel} }; - my $parent = $self->find({map { $_ => $item->{$_} } @pks}) - || $self->throw_exception('Cannot find the relating object.'); + $main_row ||= $self->new_result({map { $_ => $item->{$_} } @pks}); - my $child = $parent->$rel; + my $child = $main_row->$rel; my $related = $child->result_source->_resolve_condition( - $parent->result_source->relationship_info($rel)->{cond}, + $rels->{$rel}{cond}, $child, - $parent, + $main_row, + $rel, ); my @rows_to_add = ref $item->{$rel} eq 'ARRAY' ? @{$item->{$rel}} : ($item->{$rel}); @@ -1896,13 +2325,16 @@ sub populate { } -# populate() argumnets went over several incarnations +# populate() arguments went over several incarnations # What we ultimately support is AoH sub _normalize_populate_args { my ($self, $arg) = @_; if (ref $arg eq 'ARRAY') { - if (ref $arg->[0] eq 'HASH') { + if (!@$arg) { + return []; + } + elsif (ref $arg->[0] eq 'HASH') { return $arg; } elsif (ref $arg->[0] eq 'ARRAY') { @@ -1924,11 +2356,11 @@ sub _normalize_populate_args { =item Arguments: none -=item Return Value: $pager +=item Return Value: L<$pager|Data::Page> =back -Return Value a L object for the current resultset. Only makes +Returns a L object for the current resultset. Only makes sense for queries with a C attribute. To get the full count of entries for a paged resultset, call @@ -1942,20 +2374,26 @@ sub pager { return $self->{pager} if $self->{pager}; my $attrs = $self->{attrs}; - $self->throw_exception("Can't create pager for non-paged rs") - unless $self->{attrs}{page}; + if (!defined $attrs->{page}) { + $self->throw_exception("Can't create pager for non-paged rs"); + } + elsif ($attrs->{page} <= 0) { + $self->throw_exception('Invalid page number (page-numbers are 1-based)'); + } $attrs->{rows} ||= 10; # throw away the paging flags and re-run the count (possibly # with a subselect) to get the real total count my $count_attrs = { %$attrs }; - delete $count_attrs->{$_} for qw/rows offset page pager/; - my $total_count = (ref $self)->new($self->result_source, $count_attrs)->count; + delete @{$count_attrs}{qw/rows offset page pager/}; - return $self->{pager} = Data::Page->new( - $total_count, + my $total_rs = (ref $self)->new($self->result_source, $count_attrs); + + require DBIx::Class::ResultSet::Pager; + return $self->{pager} = DBIx::Class::ResultSet::Pager->new( + sub { $total_rs->count }, #lazy-get the total $attrs->{rows}, - $self->{attrs}{page} + $self->{attrs}{page}, ); } @@ -1965,7 +2403,7 @@ sub pager { =item Arguments: $page_number -=item Return Value: $rs +=item Return Value: L<$resultset|/search> =back @@ -1984,16 +2422,16 @@ sub page { =over 4 -=item Arguments: \%vals +=item Arguments: \%col_data -=item Return Value: $rowobject +=item Return Value: L<$result|DBIx::Class::Manual::ResultClass> =back -Creates a new row object in the resultset's result class and returns +Creates a new result object in the resultset's result class and returns it. The row is not inserted into the database at this point, call L to do that. Calling L -will tell you whether the row object has been inserted or not. +will tell you whether the result object has been inserted or not. Passes the hashref of input on to L. @@ -2001,30 +2439,47 @@ Passes the hashref of input on to L. sub new_result { my ($self, $values) = @_; - $self->throw_exception( "new_result needs a hash" ) + + $self->throw_exception( "new_result takes only one argument - a hashref of values" ) + if @_ > 2; + + $self->throw_exception( "new_result expects a hashref" ) unless (ref $values eq 'HASH'); - my ($merged_cond, $cols_from_relations) = $self->_merge_cond_with_data($values); + my ($merged_cond, $cols_from_relations) = $self->_merge_with_rscond($values); - my %new = ( + my $new = $self->result_class->new({ %$merged_cond, - @$cols_from_relations + ( @$cols_from_relations ? (-cols_from_relations => $cols_from_relations) - : (), - -source_handle => $self->_source_handle, + : () + ), -result_source => $self->result_source, # DO NOT REMOVE THIS, REQUIRED - ); + }); + + if ( + reftype($new) eq 'HASH' + and + ! keys %$new + and + blessed($new) + ) { + carp_unique (sprintf ( + "%s->new returned a blessed empty hashref - a strong indicator something is wrong with its inheritance chain", + $self->result_class, + )); + } - return $self->result_class->new(\%new); + $new; } -# _merge_cond_with_data +# _merge_with_rscond # # Takes a simple hash of K/V data and returns its copy merged with the # condition already present on the resultset. Additionally returns an # arrayref of value/condition names, which were inferred from related # objects (this is needed for in-memory related objects) -sub _merge_cond_with_data { +sub _merge_with_rscond { my ($self, $data) = @_; my (%new_data, @cols_from_relations); @@ -2044,17 +2499,25 @@ sub _merge_cond_with_data { ); } else { - # precendence must be given to passed values over values inherited from + # precedence must be given to passed values over values inherited from # the cond, so the order here is important. my $collapsed_cond = $self->_collapse_cond($self->{cond}); my %implied = %{$self->_remove_alias($collapsed_cond, $alias)}; while ( my($col, $value) = each %implied ) { - if (ref($value) eq 'HASH' && keys(%$value) && (keys %$value)[0] eq '=') { + my $vref = ref $value; + if ( + $vref eq 'HASH' + and + keys(%$value) == 1 + and + (keys %$value)[0] eq '=' + ) { $new_data{$col} = $value->{'='}; - next; } - $new_data{$col} = $value if $self->_is_deterministic_value($value); + elsif( !$vref or $vref eq 'SCALAR' or blessed($value) ) { + $new_data{$col} = $value; + } } } @@ -2066,26 +2529,12 @@ sub _merge_cond_with_data { return (\%new_data, \@cols_from_relations); } -# _is_deterministic_value -# -# Make an effor to strip non-deterministic values from the condition, -# to make sure new_result chokes less - -sub _is_deterministic_value { - my $self = shift; - my $value = shift; - my $ref_type = ref $value; - return 1 if $ref_type eq '' || $ref_type eq 'SCALAR'; - return 1 if Scalar::Util::blessed($value); - return 0; -} - # _has_resolved_attr # # determines if the resultset defines at least one # of the attributes supplied # -# used to determine if a subquery is neccessary +# used to determine if a subquery is necessary # # supports some virtual attributes: # -join @@ -2194,7 +2643,7 @@ sub _remove_alias { =item Arguments: none -=item Return Value: \[ $sql, @bind ] +=item Return Value: \[ $sql, L<@bind_values|/DBIC BIND VALUES> ] =back @@ -2207,27 +2656,22 @@ This is generally used as the RHS for a subquery. sub as_query { my $self = shift; - my $attrs = $self->_resolved_attrs_copy; + my $attrs = { %{ $self->_resolved_attrs } }; - # For future use: - # - # in list ctx: - # my ($sql, \@bind, \%dbi_bind_attrs) = _select_args_to_query (...) - # $sql also has no wrapping parenthesis in list ctx - # - my $sqlbind = $self->result_source->storage - ->_select_args_to_query ($attrs->{from}, $attrs->{select}, $attrs->{where}, $attrs); + my $aq = $self->result_source->storage->_select_args_to_query ( + $attrs->{from}, $attrs->{select}, $attrs->{where}, $attrs + ); - return $sqlbind; + $aq; } =head2 find_or_new =over 4 -=item Arguments: \%vals, \%attrs? +=item Arguments: \%col_data, { key => $unique_constraint, L<%attrs|/ATTRIBUTES> }? -=item Return Value: $rowobject +=item Return Value: L<$result|DBIx::Class::Manual::ResultClass> =back @@ -2235,19 +2679,20 @@ sub as_query { { artist => 'fred' }, { key => 'artists' }); $cd->cd_to_producer->find_or_new({ producer => $producer }, - { key => 'primary }); + { key => 'primary' }); + +Find an existing record from this resultset using L. if none exists, +instantiate a new result object and return it. The object will not be saved +into your storage until you call L on it. -Find an existing record from this resultset, based on its primary -key, or a unique constraint. If none exists, instantiate a new result -object and return it. The object will not be saved into your storage -until you call L on it. +You most likely want this method when looking for existing rows using a unique +constraint that is not the primary key, or looking for related rows. -You most likely want this method when looking for existing rows using -a unique constraint that is not the primary key, or looking for -related rows. +If you want objects to be saved immediately, use L instead. -If you want objects to be saved immediately, use L -instead. +B: Make sure to read the documentation of L and understand the +significance of the C attribute, as its lack may skew your search, and +subsequently result in spurious new objects. B: Take care when using C with a table having columns with default values that you intend to be automatically @@ -2271,9 +2716,9 @@ sub find_or_new { =over 4 -=item Arguments: \%vals +=item Arguments: \%col_data -=item Return Value: a L $object +=item Return Value: L<$result|DBIx::Class::Manual::ResultClass> =back @@ -2297,12 +2742,11 @@ This can be applied recursively, and will work correctly for a structure with an arbitrary depth and width, as long as the relationships actually exists and the correct column data has been supplied. - Instead of hashrefs of plain related data (key/value pairs), you may also pass new or inserted objects. New objects (not inserted yet, see -L), will be inserted into their appropriate tables. +L), will be inserted into their appropriate tables. -Effectively a shortcut for C<< ->new_result(\%vals)->insert >>. +Effectively a shortcut for C<< ->new_result(\%col_data)->insert >>. Example of creating a new row. @@ -2323,7 +2767,7 @@ or C resultset. Note Arrayref. ); Example of creating a new row and also creating a row in a related -Cresultset. Note Hashref. +C resultset. Note Hashref. $cd_rs->create({ title=>"Music for Silly Walks", @@ -2340,28 +2784,29 @@ Cresultset. Note Hashref. When subclassing ResultSet never attempt to override this method. Since it is a simple shortcut for C<< $self->new_result($attrs)->insert >>, a lot of the internals simply never call it, so your override will be -bypassed more often than not. Override either L -or L depending on how early in the -L process you need to intervene. +bypassed more often than not. Override either L +or L depending on how early in the +L process you need to intervene. See also warning pertaining to +L. =back =cut sub create { - my ($self, $attrs) = @_; + my ($self, $col_data) = @_; $self->throw_exception( "create needs a hashref" ) - unless ref $attrs eq 'HASH'; - return $self->new_result($attrs)->insert; + unless ref $col_data eq 'HASH'; + return $self->new_result($col_data)->insert; } =head2 find_or_create =over 4 -=item Arguments: \%vals, \%attrs? +=item Arguments: \%col_data, { key => $unique_constraint, L<%attrs|/ATTRIBUTES> }? -=item Return Value: $rowobject +=item Return Value: L<$result|DBIx::Class::Manual::ResultClass> =back @@ -2389,6 +2834,10 @@ constraint. For example: { key => 'cd_artist_title' } ); +B: Make sure to read the documentation of L and understand the +significance of the C attribute, as its lack may skew your search, and +subsequently result in spurious row creation. + B: Because find_or_create() reads from the database and then possibly inserts based on the result, this method is subject to a race condition. Another process could create a record in the table after @@ -2404,6 +2853,23 @@ all in the call to C, even when set to C. See also L and L. For information on how to declare unique constraints, see L. +If you need to know if an existing row was found or a new one created use +L and L instead. Don't forget +to call L to save the newly created row to the +database! + + my $cd = $schema->resultset('CD')->find_or_new({ + cdid => 5, + artist => 'Massive Attack', + title => 'Mezzanine', + year => 2005, + }); + + if( !$cd->in_storage ) { + # do some stuff + $cd->insert; + } + =cut sub find_or_create { @@ -2420,18 +2886,17 @@ sub find_or_create { =over 4 -=item Arguments: \%col_values, { key => $unique_constraint }? +=item Arguments: \%col_data, { key => $unique_constraint, L<%attrs|/ATTRIBUTES> }? -=item Return Value: $rowobject +=item Return Value: L<$result|DBIx::Class::Manual::ResultClass> =back $resultset->update_or_create({ col => $val, ... }); -First, searches for an existing row matching one of the unique constraints -(including the primary key) on the source of this resultset. If a row is -found, updates it with the other given column values. Otherwise, creates a new -row. +Like L, but if a row is found it is immediately updated via +C<< $found_row->update (\%col_data) >>. + Takes an optional C attribute to search on a specific unique constraint. For example: @@ -2450,17 +2915,12 @@ For example: producer => $producer, name => 'harry', }, { - key => 'primary, + key => 'primary', }); - -If no C is specified, it searches on all unique constraints defined on the -source, including the primary key. - -If the C is specified as C, it searches only on the primary key. - -See also L and L. For information on how to declare -unique constraints, see L. +B: Make sure to read the documentation of L and understand the +significance of the C attribute, as its lack may skew your search, and +subsequently result in spurious row creation. B: Take care when using C with a table having columns with default values that you intend to be automatically @@ -2468,6 +2928,14 @@ supplied by the database (e.g. an auto_increment primary key column). In normal usage, the value of such columns should NOT be included at all in the call to C, even when set to C. +See also L and L. For information on how to declare +unique constraints, see L. + +If you need to know if an existing row was updated or a new one created use +L and L instead. Don't forget +to call L to save the newly created row to the +database! + =cut sub update_or_create { @@ -2488,21 +2956,17 @@ sub update_or_create { =over 4 -=item Arguments: \%col_values, { key => $unique_constraint }? +=item Arguments: \%col_data, { key => $unique_constraint, L<%attrs|/ATTRIBUTES> }? -=item Return Value: $rowobject +=item Return Value: L<$result|DBIx::Class::Manual::ResultClass> =back $resultset->update_or_new({ col => $val, ... }); -First, searches for an existing row matching one of the unique constraints -(including the primary key) on the source of this resultset. If a row is -found, updates it with the other given column values. Otherwise, instantiate -a new result object and return it. The object will not be saved into your storage -until you call L on it. +Like L but if a row is found it is immediately updated via +C<< $found_row->update (\%col_data) >>. -Takes an optional C attribute to search on a specific unique constraint. For example: # In your application @@ -2523,6 +2987,10 @@ For example: $cd->insert; } +B: Make sure to read the documentation of L and understand the +significance of the C attribute, as its lack may skew your search, and +subsequently result in spurious new objects. + B: Take care when using C with a table having columns with default values that you intend to be automatically supplied by the database (e.g. an auto_increment primary key column). @@ -2553,7 +3021,7 @@ sub update_or_new { =item Arguments: none -=item Return Value: \@cache_objects? +=item Return Value: L<\@result_objs|DBIx::Class::Manual::ResultClass> | undef =back @@ -2572,15 +3040,15 @@ sub get_cache { =over 4 -=item Arguments: \@cache_objects +=item Arguments: L<\@result_objs|DBIx::Class::Manual::ResultClass> -=item Return Value: \@cache_objects +=item Return Value: L<\@result_objs|DBIx::Class::Manual::ResultClass> =back Sets the contents of the cache for the resultset. Expects an arrayref of objects of the same class as those produced by the resultset. Note that -if the cache is set the resultset will return the cached objects rather +if the cache is set, the resultset will return the cached objects rather than re-querying the database even if the cache attr is not set. The contents of the cache can also be populated by using the @@ -2601,7 +3069,7 @@ sub set_cache { =item Arguments: none -=item Return Value: [] +=item Return Value: undef =back @@ -2644,16 +3112,16 @@ sub is_paged { sub is_ordered { my ($self) = @_; - return scalar $self->result_source->storage->_parse_order_by($self->{attrs}{order_by}); + return scalar $self->result_source->storage->_extract_order_criteria($self->{attrs}{order_by}); } =head2 related_resultset =over 4 -=item Arguments: $relationship_name +=item Arguments: $rel_name -=item Return Value: $resultset +=item Return Value: L<$resultset|/search> =back @@ -2666,8 +3134,10 @@ Returns a related resultset for the supplied relationship name. sub related_resultset { my ($self, $rel) = @_; - $self->{related_resultsets} ||= {}; - return $self->{related_resultsets}{$rel} ||= do { + return $self->{related_resultsets}{$rel} + if defined $self->{related_resultsets}{$rel}; + + return $self->{related_resultsets}{$rel} = do { my $rsrc = $self->result_source; my $rel_info = $rsrc->relationship_info($rel); @@ -2687,21 +3157,12 @@ sub related_resultset { # (the select/as attrs were deleted in the beginning), we need to flip all # left joins to inner, so we get the expected results # read the comment on top of the actual function to see what this does - $attrs->{from} = $rsrc->schema->storage->_straight_join_to_node ($attrs->{from}, $alias); + $attrs->{from} = $rsrc->schema->storage->_inner_join_to_node ($attrs->{from}, $alias); #XXX - temp fix for result_class bug. There likely is a more elegant fix -groditi delete @{$attrs}{qw(result_class alias)}; - my $new_cache; - - if (my $cache = $self->get_cache) { - if ($cache->[0] && $cache->[0]->related_resultset($rel)->get_cache) { - $new_cache = [ map { @{$_->related_resultset($rel)->get_cache} } - @$cache ]; - } - } - my $rel_source = $rsrc->related_source($rel); my $new = do { @@ -2722,7 +3183,16 @@ sub related_resultset { where => $attrs->{where}, }); }; - $new->set_cache($new_cache) if $new_cache; + + if (my $cache = $self->get_cache) { + my @related_cache = map + { $_->related_resultset($rel)->get_cache || () } + @$cache + ; + + $new->set_cache([ map @$_, @related_cache ]) if @related_cache == @$cache; + } + $new; }; } @@ -2757,17 +3227,15 @@ source alias of the current result set: my $me = $self->current_source_alias; - return $self->search( + return $self->search({ "$me.modified" => $user->id, - ); + }); } =cut sub current_source_alias { - my ($self) = @_; - - return ($self->{attrs} || {})->{alias} || 'me'; + return (shift->{attrs} || {})->{alias} || 'me'; } =head2 as_subselect_rs @@ -2776,7 +3244,7 @@ sub current_source_alias { =item Arguments: none -=item Return Value: $resultset +=item Return Value: L<$resultset|/search> =back @@ -2820,16 +3288,26 @@ but because we isolated the group by into a subselect the above works. =cut sub as_subselect_rs { - my $self = shift; + my $self = shift; + + my $attrs = $self->_resolved_attrs; + + my $fresh_rs = (ref $self)->new ( + $self->result_source + ); - return $self->result_source->resultset->search( undef, { - alias => $self->current_source_alias, - from => [{ - $self->current_source_alias => $self->as_query, - -alias => $self->current_source_alias, - -source_handle => $self->result_source->handle, - }] - }); + # these pieces will be locked in the subquery + delete $fresh_rs->{cond}; + delete @{$fresh_rs->{attrs}}{qw/where bind/}; + + return $fresh_rs->search( {}, { + from => [{ + $attrs->{alias} => $self->as_query, + -alias => $attrs->{alias}, + -rsrc => $self->result_source, + }], + alias => $attrs->{alias}, + }); } # This code is called by search_related, and makes sure there @@ -2852,9 +3330,9 @@ sub _chain_relationship { # we need to take the prefetch the attrs into account before we # ->_resolve_join as otherwise they get lost - captainL - my $join = $self->_merge_attr( $attrs->{join}, $attrs->{prefetch} ); + my $join = $self->_merge_joinpref_attr( $attrs->{join}, $attrs->{prefetch} ); - delete @{$attrs}{qw/join prefetch collapse distinct select as columns +select +as +columns/}; + delete @{$attrs}{qw/join prefetch collapse group_by distinct _grouped_by_distinct select as columns +select +as +columns/}; my $seen = { %{ (delete $attrs->{seen_join}) || {} } }; @@ -2870,17 +3348,17 @@ sub _chain_relationship { # are resolved (prefetch is useless - we are wrapping # a subquery anyway). my $rs_copy = $self->search; - $rs_copy->{attrs}{join} = $self->_merge_attr ( + $rs_copy->{attrs}{join} = $self->_merge_joinpref_attr ( $rs_copy->{attrs}{join}, delete $rs_copy->{attrs}{prefetch}, ); $from = [{ - -source_handle => $source->handle, - -alias => $attrs->{alias}, + -rsrc => $source, + -alias => $attrs->{alias}, $attrs->{alias} => $rs_copy->as_query, }]; - delete @{$attrs}{@force_subq_attrs, 'where'}; + delete @{$attrs}{@force_subq_attrs, qw/where bind/}; $seen->{-relation_chain_depth} = 0; } elsif ($attrs->{from}) { #shallow copy suffices @@ -2888,7 +3366,7 @@ sub _chain_relationship { } else { $from = [{ - -source_handle => $source->handle, + -rsrc => $source, -alias => $attrs->{alias}, $attrs->{alias} => $source->from, }]; @@ -2939,12 +3417,6 @@ sub _chain_relationship { return {%$attrs, from => $from, seen_join => $seen}; } -# too many times we have to do $attrs = { %{$self->_resolved_attrs} } -sub _resolved_attrs_copy { - my $self = shift; - return { %{$self->_resolved_attrs (@_)} }; -} - sub _resolved_attrs { my $self = shift; return $self->{_attrs} if $self->{_attrs}; @@ -2953,100 +3425,78 @@ sub _resolved_attrs { my $source = $self->result_source; my $alias = $attrs->{alias}; - $attrs->{columns} ||= delete $attrs->{cols} if exists $attrs->{cols}; - my @colbits; + $self->throw_exception("Specifying distinct => 1 in conjunction with collapse => 1 is unsupported") + if $attrs->{collapse} and $attrs->{distinct}; - # build columns (as long as select isn't set) into a set of as/select hashes - unless ( $attrs->{select} ) { + # default selection list + $attrs->{columns} = [ $source->columns ] + unless List::Util::first { exists $attrs->{$_} } qw/columns cols select as/; - my @cols; - if ( ref $attrs->{columns} eq 'ARRAY' ) { - @cols = @{ delete $attrs->{columns}} - } elsif ( defined $attrs->{columns} ) { - @cols = delete $attrs->{columns} - } else { - @cols = $source->columns - } - - for (@cols) { - if ( ref $_ eq 'HASH' ) { - push @colbits, $_ - } else { - my $key = /^\Q${alias}.\E(.+)$/ - ? "$1" - : "$_"; - my $value = /\./ - ? "$_" - : "${alias}.$_"; - push @colbits, { $key => $value }; - } - } + # merge selectors together + for (qw/columns select as/) { + $attrs->{$_} = $self->_merge_attr($attrs->{$_}, delete $attrs->{"+$_"}) + if $attrs->{$_} or $attrs->{"+$_"}; } - # add the additional columns on - foreach (qw{include_columns +columns}) { - if ( $attrs->{$_} ) { - my @list = ( ref($attrs->{$_}) eq 'ARRAY' ) - ? @{ delete $attrs->{$_} } - : delete $attrs->{$_}; - for (@list) { - if ( ref($_) eq 'HASH' ) { - push @colbits, $_ - } else { - my $key = ( split /\./, $_ )[-1]; - my $value = ( /\./ ? $_ : "$alias.$_" ); - push @colbits, { $key => $value }; + # disassemble columns + my (@sel, @as); + if (my $cols = delete $attrs->{columns}) { + for my $c (ref $cols eq 'ARRAY' ? @$cols : $cols) { + if (ref $c eq 'HASH') { + for my $as (sort keys %$c) { + push @sel, $c->{$as}; + push @as, $as; } } + else { + push @sel, $c; + push @as, $c; + } } } - # start with initial select items - if ( $attrs->{select} ) { - $attrs->{select} = - ( ref $attrs->{select} eq 'ARRAY' ) - ? [ @{ $attrs->{select} } ] - : [ $attrs->{select} ]; - - if ( $attrs->{as} ) { - $attrs->{as} = - ( - ref $attrs->{as} eq 'ARRAY' - ? [ @{ $attrs->{as} } ] - : [ $attrs->{as} ] - ) - } else { - $attrs->{as} = [ map { - m/^\Q${alias}.\E(.+)$/ - ? $1 - : $_ - } @{ $attrs->{select} } - ] + # when trying to weed off duplicates later do not go past this point - + # everything added from here on is unbalanced "anyone's guess" stuff + my $dedup_stop_idx = $#as; + + push @as, @{ ref $attrs->{as} eq 'ARRAY' ? $attrs->{as} : [ $attrs->{as} ] } + if $attrs->{as}; + push @sel, @{ ref $attrs->{select} eq 'ARRAY' ? $attrs->{select} : [ $attrs->{select} ] } + if $attrs->{select}; + + # assume all unqualified selectors to apply to the current alias (legacy stuff) + $_ = (ref $_ or $_ =~ /\./) ? $_ : "$alias.$_" for @sel; + + # disqualify all $alias.col as-bits (inflate-map mandated) + $_ = ($_ =~ /^\Q$alias.\E(.+)$/) ? $1 : $_ for @as; + + # de-duplicate the result (remove *identical* select/as pairs) + # and also die on duplicate {as} pointing to different {select}s + # not using a c-style for as the condition is prone to shrinkage + my $seen; + my $i = 0; + while ($i <= $dedup_stop_idx) { + if ($seen->{"$sel[$i] \x00\x00 $as[$i]"}++) { + splice @sel, $i, 1; + splice @as, $i, 1; + $dedup_stop_idx--; + } + elsif ($seen->{$as[$i]}++) { + $self->throw_exception( + "inflate_result() alias '$as[$i]' specified twice with different SQL-side {select}-ors" + ); + } + else { + $i++; } } - else { - # otherwise we intialise select & as to empty - $attrs->{select} = []; - $attrs->{as} = []; - } - - # now add colbits to select/as - push @{ $attrs->{select} }, map values %{$_}, @colbits; - push @{ $attrs->{as} }, map keys %{$_}, @colbits; - if ( my $adds = delete $attrs->{'+select'} ) { - $adds = [$adds] unless ref $adds eq 'ARRAY'; - push @{ $attrs->{select} }, - map { /\./ || ref $_ ? $_ : "$alias.$_" } @$adds; - } - if ( my $adds = delete $attrs->{'+as'} ) { - $adds = [$adds] unless ref $adds eq 'ARRAY'; - push @{ $attrs->{as} }, @$adds; - } + $attrs->{select} = \@sel; + $attrs->{as} = \@as; $attrs->{from} ||= [{ - -source_handle => $source->handle, - -alias => $self->{attrs}{alias}, + -rsrc => $source, + -alias => $self->{attrs}{alias}, $self->{attrs}{alias} => $source->from, }]; @@ -3055,10 +3505,10 @@ sub _resolved_attrs { $self->throw_exception ('join/prefetch can not be used with a custom {from}') if ref $attrs->{from} ne 'ARRAY'; - my $join = delete $attrs->{join} || {}; + my $join = (delete $attrs->{join}) || {}; if ( defined $attrs->{prefetch} ) { - $join = $self->_merge_attr( $join, $attrs->{prefetch} ); + $join = $self->_merge_joinpref_attr( $join, $attrs->{prefetch} ); } $attrs->{from} = # have to copy here to avoid corrupting the original @@ -3088,34 +3538,20 @@ sub _resolved_attrs { $attrs->{group_by} = [ $attrs->{group_by} ]; } - # generate the distinct induced group_by early, as prefetch will be carried via a - # subquery (since a group_by is present) - if (delete $attrs->{distinct}) { - if ($attrs->{group_by}) { - carp ("Useless use of distinct on a grouped resultset ('distinct' is ignored when a 'group_by' is present)"); - } - else { - $attrs->{group_by} = [ grep { !ref($_) || (ref($_) ne 'HASH') } @{$attrs->{select}} ]; - # add any order_by parts that are not already present in the group_by - # we need to be careful not to add any named functions/aggregates - # i.e. select => [ ... { count => 'foo', -as 'foocount' } ... ] - my %already_grouped = map { $_ => 1 } (@{$attrs->{group_by}}); + # generate selections based on the prefetch helper + my ($prefetch, @prefetch_select, @prefetch_as); + $prefetch = $self->_merge_joinpref_attr( {}, delete $attrs->{prefetch} ) + if defined $attrs->{prefetch}; - my $storage = $self->result_source->schema->storage; + if ($prefetch) { - my $rs_column_list = $storage->_resolve_column_info ($attrs->{from}); + $self->throw_exception("Unable to prefetch, resultset contains an unnamed selector $attrs->{_dark_selector}{string}") + if $attrs->{_dark_selector}; - for my $chunk ($storage->_parse_order_by($attrs->{order_by})) { - if ($rs_column_list->{$chunk} && not $already_grouped{$chunk}++) { - push @{$attrs->{group_by}}, $chunk; - } - } - } - } + $self->throw_exception("Specifying prefetch in conjunction with an explicit collapse => 0 is unsupported") + if defined $attrs->{collapse} and ! $attrs->{collapse}; - # generate selections based on the prefetch helper - if ( my $prefetch = delete $attrs->{prefetch} ) { $attrs->{collapse} = 1; # this is a separate structure (we don't look in {from} directly) @@ -3141,43 +3577,87 @@ sub _resolved_attrs { my @prefetch = $source->_resolve_prefetch( $prefetch, $alias, $join_map ); - # we need to somehow mark which columns came from prefetch - $attrs->{_prefetch_select} = [ map { $_->[0] } @prefetch ]; - - push @{ $attrs->{select} }, @{$attrs->{_prefetch_select}}; - push @{ $attrs->{as} }, (map { $_->[1] } @prefetch); + # save these for after distinct resolution + @prefetch_select = map { $_->[0] } @prefetch; + @prefetch_as = map { $_->[1] } @prefetch; } # run through the resulting joinstructure (starting from our current slot) - # and unset collapse if proven unnesessary - if ($attrs->{collapse} && ref $attrs->{from} eq 'ARRAY') { + # and unset collapse if proven unnecessary + # + # also while we are at it find out if the current root source has + # been premultiplied by previous related_source chaining + # + # this allows to predict whether a root object with all other relation + # data set to NULL is in fact unique + if ($attrs->{collapse}) { - if (@{$attrs->{from}} > 1) { + if (ref $attrs->{from} eq 'ARRAY') { - # find where our table-spec starts and consider only things after us - my @fromlist = @{$attrs->{from}}; - while (@fromlist) { - my $t = shift @fromlist; - $t = $t->[0] if ref $t eq 'ARRAY'; #me vs join from-spec mismatch - last if ($t->{-alias} && $t->{-alias} eq $alias); + if (@{$attrs->{from}} == 1) { + # no joins - no collapse + $attrs->{collapse} = 0; } + else { + # find where our table-spec starts + my @fromlist = @{$attrs->{from}}; + while (@fromlist) { + my $t = shift @fromlist; + + my $is_multi; + # me vs join from-spec distinction - a ref means non-root + if (ref $t eq 'ARRAY') { + $t = $t->[0]; + $is_multi ||= ! $t->{-is_single}; + } + last if ($t->{-alias} && $t->{-alias} eq $alias); + $attrs->{_main_source_premultiplied} ||= $is_multi; + } - for (@fromlist) { - $attrs->{collapse} = ! $_->[0]{-is_single} - and last; + # no non-singles remaining, nor any premultiplication - nothing to collapse + if ( + ! $attrs->{_main_source_premultiplied} + and + ! List::Util::first { ! $_->[0]{-is_single} } @fromlist + ) { + $attrs->{collapse} = 0; + } } } + else { - # no joins - no collapse - $attrs->{collapse} = 0; + # if we can not analyze the from - err on the side of safety + $attrs->{_main_source_premultiplied} = 1; } } - # the row parser generates differently depending on whether collapsing is requested - # the need to look at {select} is temporary - $attrs->{_row_parser} = $source->_mk_row_parser ( - @{$attrs}{qw/as collapse select/} - ); + # generate the distinct induced group_by before injecting the prefetched select/as parts + if (delete $attrs->{distinct}) { + if ($attrs->{group_by}) { + carp_unique ("Useless use of distinct on a grouped resultset ('distinct' is ignored when a 'group_by' is present)"); + } + else { + $attrs->{_grouped_by_distinct} = 1; + # distinct affects only the main selection part, not what prefetch may add below + ($attrs->{group_by}, my $new_order) = $source->storage->_group_over_selection($attrs); + + # FIXME possibly ignore a rewritten order_by (may turn out to be an issue) + # The thinking is: if we are collapsing the subquerying prefetch engine will + # rip stuff apart for us anyway, and we do not want to have a potentially + # function-converted external order_by + # ( there is an explicit if ( collapse && _grouped_by_distinct ) check in DBIHacks ) + $attrs->{order_by} = $new_order unless $attrs->{collapse}; + } + } + + # inject prefetch-bound selection (if any) + push @{$attrs->{select}}, @prefetch_select; + push @{$attrs->{as}}, @prefetch_as; + + # whether we can get away with the dumbest (possibly DBI-internal) collapser + if ( List::Util::first { $_ =~ /\./ } @{$attrs->{as}} ) { + $attrs->{_related_results_construction} = 1; + } # if both page and offset are specified, produce a combined offset # even though it doesn't make much sense, this is what pre 081xx has @@ -3264,7 +3744,7 @@ sub _calculate_score { } } -sub _merge_attr { +sub _merge_joinpref_attr { my ($self, $orig, $import) = @_; return $import unless defined($orig); @@ -3286,6 +3766,7 @@ sub _merge_attr { $position++; } my ($import_key) = ( ref $import_element eq 'HASH' ) ? keys %{$import_element} : ($import_element); + $import_key = '' if not defined $import_key; if ($best_candidate->{score} == 0 || exists $seen_keys->{$import_key}) { push( @{$orig}, $import_element ); @@ -3296,25 +3777,129 @@ sub _merge_attr { $orig->[$best_candidate->{position}] = $import_element; } elsif (ref $import_element eq 'HASH') { my ($key) = keys %{$orig_best}; - $orig->[$best_candidate->{position}] = { $key => $self->_merge_attr($orig_best->{$key}, $import_element->{$key}) }; + $orig->[$best_candidate->{position}] = { $key => $self->_merge_joinpref_attr($orig_best->{$key}, $import_element->{$key}) }; } } $seen_keys->{$import_key} = 1; # don't merge the same key twice } - return $orig; + return @$orig ? $orig : (); } -sub result_source { - my $self = shift; +{ + my $hm; - if (@_) { - $self->_source_handle($_[0]->handle); - } else { - $self->_source_handle->resolve; - } + sub _merge_attr { + $hm ||= do { + require Hash::Merge; + my $hm = Hash::Merge->new; + + $hm->specify_behavior({ + SCALAR => { + SCALAR => sub { + my ($defl, $defr) = map { defined $_ } (@_[0,1]); + + if ($defl xor $defr) { + return [ $defl ? $_[0] : $_[1] ]; + } + elsif (! $defl) { + return []; + } + elsif (__HM_DEDUP and $_[0] eq $_[1]) { + return [ $_[0] ]; + } + else { + return [$_[0], $_[1]]; + } + }, + ARRAY => sub { + return $_[1] if !defined $_[0]; + return $_[1] if __HM_DEDUP and List::Util::first { $_ eq $_[0] } @{$_[1]}; + return [$_[0], @{$_[1]}] + }, + HASH => sub { + return [] if !defined $_[0] and !keys %{$_[1]}; + return [ $_[1] ] if !defined $_[0]; + return [ $_[0] ] if !keys %{$_[1]}; + return [$_[0], $_[1]] + }, + }, + ARRAY => { + SCALAR => sub { + return $_[0] if !defined $_[1]; + return $_[0] if __HM_DEDUP and List::Util::first { $_ eq $_[1] } @{$_[0]}; + return [@{$_[0]}, $_[1]] + }, + ARRAY => sub { + my @ret = @{$_[0]} or return $_[1]; + return [ @ret, @{$_[1]} ] unless __HM_DEDUP; + my %idx = map { $_ => 1 } @ret; + push @ret, grep { ! defined $idx{$_} } (@{$_[1]}); + \@ret; + }, + HASH => sub { + return [ $_[1] ] if ! @{$_[0]}; + return $_[0] if !keys %{$_[1]}; + return $_[0] if __HM_DEDUP and List::Util::first { $_ eq $_[1] } @{$_[0]}; + return [ @{$_[0]}, $_[1] ]; + }, + }, + HASH => { + SCALAR => sub { + return [] if !keys %{$_[0]} and !defined $_[1]; + return [ $_[0] ] if !defined $_[1]; + return [ $_[1] ] if !keys %{$_[0]}; + return [$_[0], $_[1]] + }, + ARRAY => sub { + return [] if !keys %{$_[0]} and !@{$_[1]}; + return [ $_[0] ] if !@{$_[1]}; + return $_[1] if !keys %{$_[0]}; + return $_[1] if __HM_DEDUP and List::Util::first { $_ eq $_[0] } @{$_[1]}; + return [ $_[0], @{$_[1]} ]; + }, + HASH => sub { + return [] if !keys %{$_[0]} and !keys %{$_[1]}; + return [ $_[0] ] if !keys %{$_[1]}; + return [ $_[1] ] if !keys %{$_[0]}; + return [ $_[0] ] if $_[0] eq $_[1]; + return [ $_[0], $_[1] ]; + }, + } + } => 'DBIC_RS_ATTR_MERGER'); + $hm; + }; + + return $hm->merge ($_[1], $_[2]); + } +} + +sub STORABLE_freeze { + my ($self, $cloning) = @_; + my $to_serialize = { %$self }; + + # A cursor in progress can't be serialized (and would make little sense anyway) + # the parser can be regenerated (and can't be serialized) + delete @{$to_serialize}{qw/cursor _row_parser _result_inflator/}; + + # nor is it sensical to store a not-yet-fired-count pager + if ($to_serialize->{pager} and ref $to_serialize->{pager}{total_entries} eq 'CODE') { + delete $to_serialize->{pager}; + } + + Storable::nfreeze($to_serialize); } +# need this hook for symmetry +sub STORABLE_thaw { + my ($self, $cloning, $serialized) = @_; + + %$self = %{ Storable::thaw($serialized) }; + + $self; +} + + =head2 throw_exception See L for details. @@ -3324,14 +3909,18 @@ See L for details. sub throw_exception { my $self=shift; - if (ref $self && $self->_source_handle->schema) { - $self->_source_handle->schema->throw_exception(@_) + if (ref $self and my $rsrc = $self->result_source) { + $rsrc->throw_exception(@_) } else { DBIx::Class::Exception->throw(@_); } } +1; + +__END__ + # XXX: FIXME: Attributes docs need clearing up =head1 ATTRIBUTES @@ -3341,6 +3930,10 @@ searching for data. They can be passed to any method which takes an C<\%attrs> argument. See L, L, L, L. +Default attributes can be set on the result class using +L. (Please read +the CAVEATS on that feature before using it!) + These are in no particular order: =head2 order_by @@ -3377,7 +3970,7 @@ syntax as outlined above. =over 4 -=item Value: \@columns +=item Value: \@columns | \%columns | $column =back @@ -3387,7 +3980,30 @@ case the key is the C value, and the value is used as the C from that, then auto-populates C from C and L. + + columns => [ 'foo', { bar => 'baz' } ] + +is the same as + + select => [qw/foo baz/], + as => [qw/foo bar/] + +Like elsewhere, literal SQL or literal values can be included by using a +scalar reference or a literal bind value, and these values will be available +in the result with C (see also +L): + + # equivalent SQL: SELECT 1, 'a string', IF(x,1,2) ... + columns => [ + { + foo => \1, + bar => \q{'a string'}, + baz => \[ '?', 'IF(x,1,2)' ], + } + ] =head2 +columns @@ -3397,10 +4013,10 @@ earlier versions of DBIC.) =back -Indicates additional columns to be selected from storage. Works the same -as L but adds columns to the selection. (You may also use the -C attribute, as in earlier versions of DBIC). For -example:- +Indicates additional columns to be selected from storage. Works the same as +L but adds columns to the selection. (You may also use the +C attribute, as in earlier versions of DBIC, but this is +deprecated). For example:- $schema->resultset('CD')->search(undef, { '+columns' => ['artist.name'], @@ -3412,6 +4028,10 @@ passed to object inflation. Note that the 'artist' is the name of the column (or relationship) accessor, and 'name' is the name of the column accessor in the related table. +B You need to explicitly quote '+columns' when defining the attribute. +Not doing so causes Perl to incorrectly interpret +columns as a bareword with a +unary plus operator before it. + =head2 include_columns =over 4 @@ -3438,31 +4058,31 @@ names: select => [ 'name', { count => 'employeeid' }, - { sum => 'salary' } + { max => { length => 'name' }, -as => 'longest_name' } ] }); -When you use function/stored procedure names and do not supply an C -attribute, the column names returned are storage-dependent. E.g. MySQL would -return a column named C in the above example. + # Equivalent SQL + SELECT name, COUNT( employeeid ), MAX( LENGTH( name ) ) AS longest_name FROM employee + +B You will almost always need a corresponding L attribute when you +use L, to instruct DBIx::Class how to store the result of the column. +Also note that the L attribute has nothing to do with the SQL-side 'AS' +identifier aliasing. You can however alias a function, so you can use it in +e.g. an C clause. This is done via the C<-as> B but adds columns to the selection. - -=back - -=head2 +as - -=over 4 - -Indicates additional column names for those added via L. See L. +L but adds columns to the default selection, instead of specifying +an explicit list. =back @@ -3474,25 +4094,26 @@ Indicates additional column names for those added via L. See L. =back -Indicates column names for object inflation. That is, C -indicates the name that the column can be accessed as via the -C method (or via the object accessor, B). It has nothing to do with the SQL code C, -usually when C for details. $rs = $schema->resultset('Employee')->search(undef, { select => [ 'name', - { count => 'employeeid' } + { count => 'employeeid' }, + { max => { length => 'name' }, -as => 'longest_name' } ], - as => ['name', 'employee_count'], + as => [qw/ + name + employee_count + max_name_length + /], }); - my $employee = $rs->first(); # get the first Employee - If the object against which the search is performed already has an accessor matching a column name specified in C, the value can be retrieved using the accessor as normal: @@ -3507,15 +4128,13 @@ use C instead: You can create your own accessors if required - see L for details. -Please note: This will NOT insert an C into the SQL -statement produced, it is used for internal access only. Thus -attempting to use the accessor in an C clause or similar -will fail miserably. +=head2 +as + +=over 4 -To get around this limitation, you can supply literal SQL to your -C -attributes will be ignored. - -B: Prefetch does a lot of deep magic. As such, it may not behave -exactly as you might expect. +=head2 prefetch =over 4 -=item * +=item Value: ($rel_name | \@rel_names | \%rel_names) -Prefetch uses the L to populate the prefetched relationships. This -may or may not be what you want. +=back -=item * +This attribute is a shorthand for specifying a L spec, adding all +columns from the joined related sources as L and setting +L to a true value. For example, the following two queries are +equivalent: -If you specify a condition on a prefetched relationship, ONLY those -rows that match the prefetched condition will be fetched into that relationship. -This means that adding prefetch to a search() B what is returned by -traversing a relationship. So, if you have C<< Artist->has_many(CDs) >> and you do + my $rs = $schema->resultset('Artist')->search({}, { + prefetch => { cds => ['genre', 'tracks' ] }, + }); - my $artist_rs = $schema->resultset('Artist')->search({ - 'cds.year' => 2008, - }, { - join => 'cds', +and + + my $rs = $schema->resultset('Artist')->search({}, { + join => { cds => ['genre', 'tracks' ] }, + collapse => 1, + '+columns' => [ + (map + { +{ "cds.$_" => "cds.$_" } } + $schema->source('Artist')->related_source('cds')->columns + ), + (map + { +{ "cds.genre.$_" => "genre.$_" } } + $schema->source('Artist')->related_source('cds')->related_source('genre')->columns + ), + (map + { +{ "cds.tracks.$_" => "tracks.$_" } } + $schema->source('Artist')->related_source('cds')->related_source('tracks')->columns + ), + ], }); - my $count = $artist_rs->first->cds->count; +Both producing the following SQL: + + SELECT me.artistid, me.name, me.rank, me.charfield, + cds.cdid, cds.artist, cds.title, cds.year, cds.genreid, cds.single_track, + genre.genreid, genre.name, + tracks.trackid, tracks.cd, tracks.position, tracks.title, tracks.last_updated_on, tracks.last_updated_at + FROM artist me + LEFT JOIN cd cds + ON cds.artist = me.artistid + LEFT JOIN genre genre + ON genre.genreid = cds.genreid + LEFT JOIN track tracks + ON tracks.cd = cds.cdid + ORDER BY me.artistid + +While L implies a L, it is ok to mix the two together, as +the arguments are properly merged and generally do the right thing. For +example, you may want to do the following: + + my $artists_and_cds_without_genre = $schema->resultset('Artist')->search( + { 'genre.genreid' => undef }, + { + join => { cds => 'genre' }, + prefetch => 'cds', + } + ); - my $artist_rs_prefetch = $artist_rs->search( {}, { prefetch => 'cds' } ); +Which generates the following SQL: - my $prefetch_count = $artist_rs_prefetch->first->cds->count; + SELECT me.artistid, me.name, me.rank, me.charfield, + cds.cdid, cds.artist, cds.title, cds.year, cds.genreid, cds.single_track + FROM artist me + LEFT JOIN cd cds + ON cds.artist = me.artistid + LEFT JOIN genre genre + ON genre.genreid = cds.genreid + WHERE genre.genreid IS NULL + ORDER BY me.artistid - cmp_ok( $count, '==', $prefetch_count, "Counts should be the same" ); +For a more in-depth discussion, see L. -that cmp_ok() may or may not pass depending on the datasets involved. This -behavior may or may not survive the 0.09 transition. +=head2 alias + +=over 4 + +=item Value: $source_alias =back +Sets the source alias for the query. Normally, this defaults to C, but +nested search queries (sub-SELECTs) might need specific aliases set to +reference inner queries. For example: + + my $q = $rs + ->related_resultset('CDs') + ->related_resultset('Tracks') + ->search({ + 'track.id' => { -ident => 'none_search.id' }, + }) + ->as_query; + + my $ids = $self->search({ + -not_exists => $q, + }, { + alias => 'none_search', + group_by => 'none_search.id', + })->get_column('id')->as_query; + + $self->search({ id => { -in => $ids } }) + +This attribute is directly tied to L. + =head2 page =over 4 @@ -3689,7 +4381,7 @@ Makes the resultset paged and specifies the page to retrieve. Effectively identical to creating a non-pages resultset and then calling ->page($page) on it. -If L attribute is not specified it defaults to 10 rows per page. +If L attribute is not specified it defaults to 10 rows per page. When you have a paged resultset, L will only return the number of rows in the page. To get the total, use the L and call @@ -3717,6 +4409,24 @@ rows per page if the page attribute or method is used. Specifies the (zero-based) row number for the first row to be returned, or the of the first row of the first page if paging is used. +=head2 software_limit + +=over 4 + +=item Value: (0 | 1) + +=back + +When combined with L and/or L the generated SQL will not +include any limit dialect stanzas. Instead the entire result will be selected +as if no limits were specified, and DBIC will perform the limit locally, by +artificially advancing and finishing the resulting L. + +This is the recommended way of performing resultset limiting when no sane RDBMS +implementation is available (e.g. +L using the +L hack) + =head2 group_by =over 4 @@ -3741,7 +4451,11 @@ HAVING is a select statement attribute that is applied between GROUP BY and ORDER BY. It is applied to the after the grouping calculations have been done. - having => { 'count(employee)' => { '>=', 100 } } + having => { 'count_employee' => { '>=', 100 } } + +or with an in-place function in which case literal SQL is required: + + having => \[ 'count(employee) >= ?', [ count => 100 ] ] =head2 distinct @@ -3751,8 +4465,17 @@ done. =back -Set to 1 to group by all columns. If the resultset already has a group_by -attribute, this setting is ignored and an appropriate warning is issued. +Set to 1 to automatically generate a L clause based on the selection +(including intelligent handling of L contents). Note that the group +criteria calculation takes place over the B selection. This includes +any L, L or L additions in subsequent +L calls, and standalone columns selected via +L (L). A notable exception are the +extra selections specified via L - such selections are explicitly +excluded from group criteria calculations. + +If the final ResultSet also explicitly defines a L attribute, this +setting is ignored and an appropriate warning is issued. =head2 where @@ -3761,11 +4484,13 @@ attribute, this setting is ignored and an appropriate warning is issued. Adds to the WHERE clause. # only return rows WHERE deleted IS NULL for all searches - __PACKAGE__->resultset_attributes({ where => { deleted => undef } }); ) + __PACKAGE__->resultset_attributes({ where => { deleted => undef } }); Can be overridden by passing C<< { where => undef } >> as an attribute to a resultset. +For more complicated where clauses see L. + =back =head2 cache @@ -3790,13 +4515,193 @@ L. =over 4 -=item Value: ( 'update' | 'shared' ) +=item Value: ( 'update' | 'shared' | \$scalar ) =back Set to 'update' for a SELECT ... FOR UPDATE or 'shared' for a SELECT -... FOR SHARED. +... FOR SHARED. If \$scalar is passed, this is taken directly and embedded in the +query. -=cut +=head1 PREFETCHING + +DBIx::Class supports arbitrary related data prefetching from multiple related +sources. Any combination of relationship types and column sets are supported. +If L is requested, there is an additional requirement of +selecting enough data to make every individual object uniquely identifiable. + +Here are some more involved examples, based on the following relationship map: + + # Assuming: + My::Schema::CD->belongs_to( artist => 'My::Schema::Artist' ); + My::Schema::CD->might_have( liner_note => 'My::Schema::LinerNotes' ); + My::Schema::CD->has_many( tracks => 'My::Schema::Track' ); + + My::Schema::Artist->belongs_to( record_label => 'My::Schema::RecordLabel' ); + + My::Schema::Track->has_many( guests => 'My::Schema::Guest' ); + + + + my $rs = $schema->resultset('Tag')->search( + undef, + { + prefetch => { + cd => 'artist' + } + } + ); + +The initial search results in SQL like the following: + + SELECT tag.*, cd.*, artist.* FROM tag + JOIN cd ON tag.cd = cd.cdid + JOIN artist ON cd.artist = artist.artistid + +L has no need to go back to the database when we access the +C or C relationships, which saves us two SQL statements in this +case. + +Simple prefetches will be joined automatically, so there is no need +for a C attribute in the above search. + +The L attribute can be used with any of the relationship types +and multiple prefetches can be specified together. Below is a more complex +example that prefetches a CD's artist, its liner notes (if present), +the cover image, the tracks on that CD, and the guests on those +tracks. + + my $rs = $schema->resultset('CD')->search( + undef, + { + prefetch => [ + { artist => 'record_label'}, # belongs_to => belongs_to + 'liner_note', # might_have + 'cover_image', # has_one + { tracks => 'guests' }, # has_many => has_many + ] + } + ); + +This will produce SQL like the following: + + SELECT cd.*, artist.*, record_label.*, liner_note.*, cover_image.*, + tracks.*, guests.* + FROM cd me + JOIN artist artist + ON artist.artistid = me.artistid + JOIN record_label record_label + ON record_label.labelid = artist.labelid + LEFT JOIN track tracks + ON tracks.cdid = me.cdid + LEFT JOIN guest guests + ON guests.trackid = track.trackid + LEFT JOIN liner_notes liner_note + ON liner_note.cdid = me.cdid + JOIN cd_artwork cover_image + ON cover_image.cdid = me.cdid + ORDER BY tracks.cd + +Now the C, C, C, C, +C, and C of the CD will all be available through the +relationship accessors without the need for additional queries to the +database. + +=head3 CAVEATS + +Prefetch does a lot of deep magic. As such, it may not behave exactly +as you might expect. + +=over 4 + +=item * + +Prefetch uses the L to populate the prefetched relationships. This +may or may not be what you want. + +=item * + +If you specify a condition on a prefetched relationship, ONLY those +rows that match the prefetched condition will be fetched into that relationship. +This means that adding prefetch to a search() B what is returned by +traversing a relationship. So, if you have C<< Artist->has_many(CDs) >> and you do + + my $artist_rs = $schema->resultset('Artist')->search({ + 'cds.year' => 2008, + }, { + join => 'cds', + }); + + my $count = $artist_rs->first->cds->count; + + my $artist_rs_prefetch = $artist_rs->search( {}, { prefetch => 'cds' } ); + + my $prefetch_count = $artist_rs_prefetch->first->cds->count; + + cmp_ok( $count, '==', $prefetch_count, "Counts should be the same" ); + +That cmp_ok() may or may not pass depending on the datasets involved. In other +words the C condition would apply to the entire dataset, just like +it would in regular SQL. If you want to add a condition only to the "right side" +of a C - consider declaring and using a L + +=back + +=head1 DBIC BIND VALUES + +Because DBIC may need more information to bind values than just the column name +and value itself, it uses a special format for both passing and receiving bind +values. Each bind value should be composed of an arrayref of +C<< [ \%args => $val ] >>. The format of C<< \%args >> is currently: + +=over 4 + +=item dbd_attrs + +If present (in any form), this is what is being passed directly to bind_param. +Note that different DBD's expect different bind args. (e.g. DBD::SQLite takes +a single numerical type, while DBD::Pg takes a hashref if bind options.) + +If this is specified, all other bind options described below are ignored. + +=item sqlt_datatype + +If present, this is used to infer the actual bind attribute by passing to +C<< $resolved_storage->bind_attribute_by_data_type() >>. Defaults to the +"data_type" from the L. + +Note that the data type is somewhat freeform (hence the sqlt_ prefix); +currently drivers are expected to "Do the Right Thing" when given a common +datatype name. (Not ideal, but that's what we got at this point.) + +=item sqlt_size + +Currently used to correctly allocate buffers for bind_param_inout(). +Defaults to "size" from the L, +or to a sensible value based on the "data_type". + +=item dbic_colname + +Used to fill in missing sqlt_datatype and sqlt_size attributes (if they are +explicitly specified they are never overridden). Also used by some weird DBDs, +where the column name should be available at bind_param time (e.g. Oracle). + +=back + +For backwards compatibility and convenience, the following shortcuts are +supported: + + [ $name => $val ] === [ { dbic_colname => $name }, $val ] + [ \$dt => $val ] === [ { sqlt_datatype => $dt }, $val ] + [ undef, $val ] === [ {}, $val ] + $val === [ {}, $val ] + +=head1 AUTHOR AND CONTRIBUTORS + +See L and L in DBIx::Class + +=head1 LICENSE + +You may distribute this code under the same terms as Perl itself. -1;