1 package #hide from PAUSE
2 DBIx::Class::Storage::DBIHacks;
5 # This module contains code that should never have seen the light of day,
6 # does not belong in the Storage, or is otherwise unfit for public
7 # display. The arrival of SQLA2 should immediately obsolete 90% of this
13 use base 'DBIx::Class::Storage';
16 use List::Util 'first';
17 use Scalar::Util 'blessed';
18 use Sub::Name 'subname';
19 use Data::Query::Constants;
20 use Data::Query::ExprHelpers;
24 # This code will remove non-selecting/non-restricting joins from
25 # {from} specs, aiding the RDBMS query optimizer
27 sub _prune_unused_joins {
29 my ($from, $select, $where, $attrs) = @_;
31 return $from unless $self->_use_join_optimizer;
33 if (ref $from ne 'ARRAY' || ref $from->[0] ne 'HASH' || ref $from->[1] ne 'ARRAY') {
34 return $from; # only standard {from} specs are supported
37 my $aliastypes = $self->_resolve_aliastypes_from_select_args(@_);
39 my $orig_joins = delete $aliastypes->{joining};
40 my $orig_multiplying = $aliastypes->{multiplying};
42 # a grouped set will not be affected by amount of rows. Thus any
43 # {multiplying} joins can go
44 delete $aliastypes->{multiplying}
45 if $attrs->{_force_prune_multiplying_joins} or $attrs->{group_by};
47 my @newfrom = $from->[0]; # FROM head is always present
51 for (values %$aliastypes) {
52 # add all requested aliases
53 $need_joins{$_} = 1 for keys %$_;
55 # add all their parents (as per joinpath which is an AoH { table => alias })
56 $need_joins{$_} = 1 for map { values %$_ } map { @{$_->{-parents}} } values %$_;
59 for my $j (@{$from}[1..$#$from]) {
60 push @newfrom, $j if (
61 (! defined $j->[0]{-alias}) # legacy crap
63 $need_joins{$j->[0]{-alias}}
68 multiplying => { map { $need_joins{$_} ? ($_ => $orig_multiplying->{$_}) : () } keys %$orig_multiplying },
70 joining => { map { $_ => $orig_joins->{$_} } keys %need_joins },
75 # This is the code producing joined subqueries like:
76 # SELECT me.*, other.* FROM ( SELECT me.* FROM ... ) JOIN other ON ...
78 sub _adjust_select_args_for_complex_prefetch {
79 my ($self, $from, $select, $where, $attrs) = @_;
81 $self->throw_exception ('Complex prefetches are not supported on resultsets with a custom from attribute')
82 if (ref $from ne 'ARRAY' || ref $from->[0] ne 'HASH' || ref $from->[1] ne 'ARRAY');
84 my $root_alias = $attrs->{alias};
86 # generate inner/outer attribute lists, remove stuff that doesn't apply
87 my $outer_attrs = { %$attrs };
88 delete @{$outer_attrs}{qw(where bind rows offset group_by _grouped_by_distinct having)};
90 my $inner_attrs = { %$attrs };
91 delete @{$inner_attrs}{qw(from for collapse select as _related_results_construction)};
93 # there is no point of ordering the insides if there is no limit
94 delete $inner_attrs->{order_by} if (
95 delete $inner_attrs->{_order_is_artificial}
97 ! $inner_attrs->{rows}
100 # generate the inner/outer select lists
101 # for inside we consider only stuff *not* brought in by the prefetch
102 # on the outside we substitute any function for its alias
103 my $outer_select = [ @$select ];
106 my ($root_node, $root_node_offset);
108 for my $i (0 .. $#$from) {
109 my $node = $from->[$i];
110 my $h = (ref $node eq 'HASH') ? $node
111 : (ref $node eq 'ARRAY' and ref $node->[0] eq 'HASH') ? $node->[0]
115 if ( ($h->{-alias}||'') eq $root_alias and $h->{-rsrc} ) {
117 $root_node_offset = $i;
122 $self->throw_exception ('Complex prefetches are not supported on resultsets with a custom from attribute')
125 # use the heavy duty resolver to take care of aliased/nonaliased naming
126 my $colinfo = $self->_resolve_column_info($from);
127 my $selected_root_columns;
129 for my $i (0 .. $#$outer_select) {
130 my $sel = $outer_select->[$i];
133 $colinfo->{$sel} and $colinfo->{$sel}{-source_alias} ne $root_alias
136 if (ref $sel eq 'HASH' ) {
137 $sel->{-as} ||= $attrs->{as}[$i];
138 $outer_select->[$i] = join ('.', $root_alias, ($sel->{-as} || "inner_column_$i") );
140 elsif (! ref $sel and my $ci = $colinfo->{$sel}) {
141 $selected_root_columns->{$ci->{-colname}} = 1;
144 push @$inner_select, $sel;
146 push @{$inner_attrs->{as}}, $attrs->{as}[$i];
149 # We will need to fetch all native columns in the inner subquery, which may
150 # be a part of an *outer* join condition, or an order_by (which needs to be
152 # We can not just fetch everything because a potential has_many restricting
153 # join collapse *will not work* on heavy data types.
154 my $connecting_aliastypes = $self->_resolve_aliastypes_from_select_args(
161 for (sort map { keys %{$_->{-seen_columns}||{}} } map { values %$_ } values %$connecting_aliastypes) {
162 my $ci = $colinfo->{$_} or next;
164 $ci->{-source_alias} eq $root_alias
166 ! $selected_root_columns->{$ci->{-colname}}++
168 # adding it to both to keep limits not supporting dark selectors happy
169 push @$inner_select, $ci->{-fq_colname};
170 push @{$inner_attrs->{as}}, $ci->{-fq_colname};
174 # construct the inner $from and lock it in a subquery
175 # we need to prune first, because this will determine if we need a group_by below
176 # throw away all non-selecting, non-restricting multijoins
177 # (since we def. do not care about multiplication those inside the subquery)
178 my $inner_subq = do {
180 # must use it here regardless of user requests
181 local $self->{_use_join_optimizer} = 1;
183 # throw away multijoins since we def. do not care about those inside the subquery
184 my ($inner_from, $inner_aliastypes) = $self->_prune_unused_joins ($from, $inner_select, $where, {
185 %$inner_attrs, _force_prune_multiplying_joins => 1
188 # uh-oh a multiplier (which is not us) left in, this is a problem
190 $inner_aliastypes->{multiplying}
192 # if there are user-supplied groups - assume user knows wtf they are up to
193 ( ! $inner_aliastypes->{grouping} or $inner_attrs->{_grouped_by_distinct} )
195 my @multipliers = grep { $_ ne $root_alias } keys %{$inner_aliastypes->{multiplying}}
198 # if none of the multipliers came from an order_by (guaranteed to have been combined
199 # with a limit) - easy - just slap a group_by to simulate a collapse and be on our way
201 ! $inner_aliastypes->{ordering}
203 ! first { $inner_aliastypes->{ordering}{$_} } @multipliers
206 my $unprocessed_order_chunks;
207 ($inner_attrs->{group_by}, $unprocessed_order_chunks) = $self->_group_over_selection ({
210 select => $inner_select,
213 $self->throw_exception (
214 'A required group_by clause could not be constructed automatically due to a complex '
215 . 'order_by criteria. Either order_by columns only (no functions) or construct a suitable '
217 ) if $unprocessed_order_chunks;
220 # We need to order by external columns and group at the same time
221 # so we can calculate the proper limit
222 # This doesn't really make sense in SQL, however from DBICs point
223 # of view is rather valid (order the leftmost objects by whatever
224 # criteria and get the offset/rows many). There is a way around
225 # this however in SQL - we simply tae the direction of each piece
226 # of the foreign order and convert them to MIN(X) for ASC or MAX(X)
227 # for DESC, and group_by the root columns. The end result should be
228 # exactly what we expect
230 # supplement the main selection with pks if not already there,
231 # as they will have to be a part of the group_by to collapse
233 my $cur_sel = { map { $_ => 1 } @$inner_select };
235 my @pks = map { "$root_alias.$_" } $root_node->{-rsrc}->primary_columns
236 or $self->throw_exception( sprintf
237 'Unable to perform complex limited prefetch off %s without declared primary key',
238 $root_node->{-rsrc}->source_name,
241 push @$inner_select, $col
242 unless $cur_sel->{$col}++;
245 # wrap any part of the order_by that "responds" to an ordering alias
247 # FIXME - this code is a joke, will need to be completely rewritten in
248 # the DQ branch. But I need to push a POC here, otherwise the
249 # pesky tests won't pass
250 my $sql_maker = $self->sql_maker;
251 my ($lquote, $rquote, $sep) = map { quotemeta $_ } ($sql_maker->_quote_chars, $sql_maker->name_sep);
252 my $own_re = qr/ $lquote \Q$root_alias\E $rquote $sep | \b \Q$root_alias\E $sep /x;
253 my $inner_columns_info = $self->_resolve_column_info($inner_from);
255 my $order_dq = $sql_maker->converter
256 ->_order_by_to_dq($attrs->{order_by});
260 # loop through and replace stuff that is not "ours" with a min/max func
261 # everything is a literal at this point, since we are likely properly
263 while (is_Order($order_dq)) {
264 my ($chunk, @args) = $sql_maker->_render_dq(my $by = $order_dq->{by});
266 my $is_desc = $order_dq->{reverse};
268 push @new_order, $is_desc ? { -desc => \$by } : \$by;
270 $order_dq = $order_dq->{from};
273 next if $chunk =~ $own_re;
275 # maybe our own unqualified column
277 $lquote and $sep and $chunk =~ /^ $lquote ([^$sep]+) $rquote $/x
283 $inner_columns_info->{$ord_bit}
285 $inner_columns_info->{$ord_bit}{-source_alias} eq $root_alias
288 ($new_order[-1]) = map {
289 ($is_desc ? { -desc => $_ } : $_)
293 ($is_desc ? 'MAX' : 'MIN'),
300 $inner_attrs->{order_by} = \@new_order;
302 # do not care about leftovers here - it will be all the functions
304 ($inner_attrs->{group_by}) = $self->_group_over_selection ({
307 select => $inner_select,
312 # we already optimized $inner_from above
313 # and already local()ized
314 $self->{_use_join_optimizer} = 0;
316 # generate the subquery
317 $self->_select_args_to_query (
325 # Generate the outer from - this is relatively easy (really just replace
326 # the join slot with the subquery), with a major caveat - we can not
327 # join anything that is non-selecting (not part of the prefetch), but at
328 # the same time is a multi-type relationship, as it will explode the result.
330 # There are two possibilities here
331 # - either the join is non-restricting, in which case we simply throw it away
332 # - it is part of the restrictions, in which case we need to collapse the outer
333 # result by tackling yet another group_by to the outside of the query
335 # work on a shallow copy
340 # we may not be the head
341 if ($root_node_offset) {
342 # first generate the outer_from, up and including the substitution point
343 @outer_from = splice @$from, 0, $root_node_offset;
347 -alias => $root_alias,
348 -rsrc => $root_node->{-rsrc},
349 $root_alias => $inner_subq,
351 @{$from->[0]}[1 .. $#{$from->[0]}],
356 -alias => $root_alias,
357 -rsrc => $root_node->{-rsrc},
358 $root_alias => $inner_subq,
362 shift @$from; # what we just replaced above
364 # scan the *remaining* from spec against different attributes, and see which joins are needed
366 my $outer_aliastypes = $outer_attrs->{_aliastypes} =
367 $self->_resolve_aliastypes_from_select_args( $from, $outer_select, $where, $outer_attrs );
370 my ($outer_select_chain, @outer_nonselecting_chains) = map { +{
371 map { $_ => 1 } map { values %$_} map { @{$_->{-parents}} } values %{ $outer_aliastypes->{$_} || {} }
372 } } qw/selecting restricting grouping ordering/;
374 # see what's left - throw away if not selecting/restricting
375 # also throw in a group_by if a non-selecting multiplier,
376 # to guard against cross-join explosions
377 my $need_outer_group_by;
378 while (my $j = shift @$from) {
379 my $alias = $j->[0]{-alias};
382 $outer_select_chain->{$alias}
386 elsif (first { $_->{$alias} } @outer_nonselecting_chains ) {
387 push @outer_from, $j;
388 $need_outer_group_by ||= $outer_aliastypes->{multiplying}{$alias} ? 1 : 0;
392 if ( $need_outer_group_by and $attrs->{_grouped_by_distinct} ) {
393 my $unprocessed_order_chunks;
394 ($outer_attrs->{group_by}, $unprocessed_order_chunks) = $self->_group_over_selection ({
396 from => \@outer_from,
397 select => $outer_select,
400 $self->throw_exception (
401 'A required group_by clause could not be constructed automatically due to a complex '
402 . 'order_by criteria. Either order_by columns only (no functions) or construct a suitable '
404 ) if $unprocessed_order_chunks;
408 # This is totally horrific - the $where ends up in both the inner and outer query
409 # Unfortunately not much can be done until SQLA2 introspection arrives, and even
410 # then if where conditions apply to the *right* side of the prefetch, you may have
411 # to both filter the inner select (e.g. to apply a limit) and then have to re-filter
412 # the outer select to exclude joins you didn't want in the first place
414 # OTOH it can be seen as a plus: <ash> (notes that this query would make a DBA cry ;)
415 return (\@outer_from, $outer_select, $where, $outer_attrs);
419 # I KNOW THIS SUCKS! GET SQLA2 OUT THE DOOR SO THIS CAN DIE!
421 # Due to a lack of SQLA2 we fall back to crude scans of all the
422 # select/where/order/group attributes, in order to determine what
423 # aliases are needed to fulfill the query. This information is used
424 # throughout the code to prune unnecessary JOINs from the queries
425 # in an attempt to reduce the execution time.
426 # Although the method is pretty horrific, the worst thing that can
427 # happen is for it to fail due to some scalar SQL, which in turn will
428 # result in a vocal exception.
429 sub _resolve_aliastypes_from_select_args {
430 my ( $self, $from, $select, $where, $attrs ) = @_;
432 $self->throw_exception ('Unable to analyze custom {from}')
433 if ref $from ne 'ARRAY';
435 # what we will return
438 # see what aliases are there to work with
442 $j = $j->[0] if ref $j eq 'ARRAY';
443 my $al = $j->{-alias}
446 $alias_list->{$al} = $j;
447 $aliases_by_type->{multiplying}{$al} ||= { -parents => $j->{-join_path}||[] } if (
448 # not array == {from} head == can't be multiplying
449 ( ref($_) eq 'ARRAY' and ! $j->{-is_single} )
451 # a parent of ours is already a multiplier
452 ( grep { $aliases_by_type->{multiplying}{$_} } @{ $j->{-join_path}||[] } )
456 # get a column to source/alias map (including unambiguous unqualified ones)
457 my $colinfo = $self->_resolve_column_info ($from);
459 # set up a botched SQLA
460 my $sql_maker = $self->sql_maker;
462 # these are throw away results, do not pollute the bind stack
463 local $sql_maker->{select_bind};
464 local $sql_maker->{where_bind};
465 local $sql_maker->{group_bind};
466 local $sql_maker->{having_bind};
467 local $sql_maker->{from_bind};
469 # we can't scan properly without any quoting (\b doesn't cut it
470 # everywhere), so unless there is proper quoting set - use our
471 # own weird impossible character.
472 # Also in the case of no quoting, we need to explicitly disable
473 # name_sep, otherwise sorry nasty legacy syntax like
474 # { 'count(foo.id)' => { '>' => 3 } } will stop working >:(
475 local $sql_maker->{quote_char} = $sql_maker->{quote_char};
476 local $sql_maker->{name_sep} = $sql_maker->{name_sep};
478 unless (defined $sql_maker->{quote_char} and length $sql_maker->{quote_char}) {
479 $sql_maker->{quote_char} = ["\x00", "\xFF"];
480 # if we don't unset it we screw up retarded but unfortunately working
481 # 'MAX(foo.bar)' => { '>', 3 }
482 $sql_maker->{name_sep} = '';
485 # local is not enough - need to ensure the inner objects get rebuilt
486 $sql_maker->clear_renderer;
487 $sql_maker->clear_converter;
489 my ($lquote, $rquote, $sep) = map { quotemeta $_ } ($sql_maker->_quote_chars, $sql_maker->name_sep);
491 # generate sql chunks
495 ? ($sql_maker->_recurse_where($where))[0]
499 ? ($sql_maker->_recurse_where($attrs->{having}))[0]
505 ? ($sql_maker->_render_sqla(group_by => $attrs->{group_by}))[0]
510 $sql_maker->_recurse_from (
511 ref $from->[0] eq 'ARRAY' ? $from->[0][0] : $from->[0],
512 @{$from}[1 .. $#$from],
517 ? ($sql_maker->_render_sqla(select_select => $select))[0]
521 map { $_->[0] } $self->_extract_order_criteria ($attrs->{order_by}, $sql_maker),
525 # local is not enough - need to ensure the inner objects get rebuilt
526 $sql_maker->clear_renderer;
527 $sql_maker->clear_converter;
529 # throw away empty chunks
530 $_ = [ map { $_ || () } @$_ ] for values %$to_scan;
532 # first see if we have any exact matches (qualified or unqualified)
533 for my $type (keys %$to_scan) {
534 for my $piece (@{$to_scan->{$type}}) {
535 if ($colinfo->{$piece} and my $alias = $colinfo->{$piece}{-source_alias}) {
536 $aliases_by_type->{$type}{$alias} ||= { -parents => $alias_list->{$alias}{-join_path}||[] };
537 $aliases_by_type->{$type}{$alias}{-seen_columns}{$colinfo->{$piece}{-fq_colname}} = $piece;
542 # now loop through all fully qualified columns and get the corresponding
543 # alias (should work even if they are in scalarrefs)
544 for my $alias (keys %$alias_list) {
546 $lquote $alias $rquote $sep (?: $lquote ([^$rquote]+) $rquote )?
548 \b $alias \. ([^\s\)\($rquote]+)?
551 for my $type (keys %$to_scan) {
552 for my $piece (@{$to_scan->{$type}}) {
553 if (my @matches = $piece =~ /$al_re/g) {
554 $aliases_by_type->{$type}{$alias} ||= { -parents => $alias_list->{$alias}{-join_path}||[] };
555 $aliases_by_type->{$type}{$alias}{-seen_columns}{"$alias.$_"} = "$alias.$_"
556 for grep { defined $_ } @matches;
562 # now loop through unqualified column names, and try to locate them within
564 for my $col (keys %$colinfo) {
565 next if $col =~ / \. /x; # if column is qualified it was caught by the above
567 my $col_re = qr/ $lquote ($col) $rquote /x;
569 for my $type (keys %$to_scan) {
570 for my $piece (@{$to_scan->{$type}}) {
571 if ( my @matches = $piece =~ /$col_re/g) {
572 my $alias = $colinfo->{$col}{-source_alias};
573 $aliases_by_type->{$type}{$alias} ||= { -parents => $alias_list->{$alias}{-join_path}||[] };
574 $aliases_by_type->{$type}{$alias}{-seen_columns}{"$alias.$_"} = $_
575 for grep { defined $_ } @matches;
581 # Add any non-left joins to the restriction list (such joins are indeed restrictions)
582 for my $j (values %$alias_list) {
583 my $alias = $j->{-alias} or next;
584 $aliases_by_type->{restricting}{$alias} ||= { -parents => $j->{-join_path}||[] } if (
585 (not $j->{-join_type})
587 ($j->{-join_type} !~ /^left (?: \s+ outer)? $/xi)
591 for (keys %$aliases_by_type) {
592 delete $aliases_by_type->{$_} unless keys %{$aliases_by_type->{$_}};
595 return $aliases_by_type;
598 # This is the engine behind { distinct => 1 }
599 sub _group_over_selection {
600 my ($self, $attrs) = @_;
602 my $colinfos = $self->_resolve_column_info ($attrs->{from});
604 my (@group_by, %group_index);
606 # the logic is: if it is a { func => val } we assume an aggregate,
607 # otherwise if \'...' or \[...] we assume the user knows what is
608 # going on thus group over it
609 for (@{$attrs->{select}}) {
610 if (! ref($_) or ref ($_) ne 'HASH' ) {
613 if ($colinfos->{$_} and $_ !~ /\./ ) {
614 # add a fully qualified version as well
615 $group_index{"$colinfos->{$_}{-source_alias}.$_"}++;
620 # add any order_by parts *from the main source* that are not already
621 # present in the group_by
622 # we need to be careful not to add any named functions/aggregates
623 # i.e. order_by => [ ... { count => 'foo' } ... ]
625 for ($self->_extract_order_criteria($attrs->{order_by})) {
626 # only consider real columns (for functions the user got to do an explicit group_by)
636 $colinfos->{$chunk}{-source_alias} ne $attrs->{alias}
642 $chunk = $colinfos->{$chunk}{-fq_colname};
643 push @group_by, $chunk unless $group_index{$chunk}++;
647 ? (\@group_by, (@leftovers ? \@leftovers : undef) )
652 sub _resolve_ident_sources {
653 my ($self, $ident) = @_;
655 my $alias2source = {};
657 # the reason this is so contrived is that $ident may be a {from}
658 # structure, specifying multiple tables to join
659 if ( blessed $ident && $ident->isa("DBIx::Class::ResultSource") ) {
660 # this is compat mode for insert/update/delete which do not deal with aliases
661 $alias2source->{me} = $ident;
663 elsif (ref $ident eq 'ARRAY') {
667 if (ref $_ eq 'HASH') {
670 if (ref $_ eq 'ARRAY' and ref $_->[0] eq 'HASH') {
674 $alias2source->{$tabinfo->{-alias}} = $tabinfo->{-rsrc}
675 if ($tabinfo->{-rsrc});
679 return $alias2source;
682 # Takes $ident, \@column_names
684 # returns { $column_name => \%column_info, ... }
685 # also note: this adds -result_source => $rsrc to the column info
687 # If no columns_names are supplied returns info about *all* columns
689 sub _resolve_column_info {
690 my ($self, $ident, $colnames) = @_;
691 my $alias2src = $self->_resolve_ident_sources($ident);
693 my (%seen_cols, @auto_colnames);
695 # compile a global list of column names, to be able to properly
696 # disambiguate unqualified column names (if at all possible)
697 for my $alias (keys %$alias2src) {
698 my $rsrc = $alias2src->{$alias};
699 for my $colname ($rsrc->columns) {
700 push @{$seen_cols{$colname}}, $alias;
701 push @auto_colnames, "$alias.$colname" unless $colnames;
707 grep { @{$seen_cols{$_}} == 1 } (keys %seen_cols),
710 my (%return, $colinfos);
711 foreach my $col (@$colnames) {
712 my ($source_alias, $colname) = $col =~ m/^ (?: ([^\.]+) \. )? (.+) $/x;
714 # if the column was seen exactly once - we know which rsrc it came from
715 $source_alias ||= $seen_cols{$colname}[0]
716 if ($seen_cols{$colname} and @{$seen_cols{$colname}} == 1);
718 next unless $source_alias;
720 my $rsrc = $alias2src->{$source_alias}
725 ( $colinfos->{$source_alias} ||= $rsrc->columns_info )->{$colname}
727 $self->throw_exception(
728 "No such column '$colname' on source " . $rsrc->source_name
731 -result_source => $rsrc,
732 -source_alias => $source_alias,
733 -fq_colname => $col eq $colname ? "$source_alias.$col" : $col,
734 -colname => $colname,
737 $return{"$source_alias.$colname"} = $return{$col} if $col eq $colname;
743 # The DBIC relationship chaining implementation is pretty simple - every
744 # new related_relationship is pushed onto the {from} stack, and the {select}
745 # window simply slides further in. This means that when we count somewhere
746 # in the middle, we got to make sure that everything in the join chain is an
747 # actual inner join, otherwise the count will come back with unpredictable
748 # results (a resultset may be generated with _some_ rows regardless of if
749 # the relation which the $rs currently selects has rows or not). E.g.
750 # $artist_rs->cds->count - normally generates:
751 # SELECT COUNT( * ) FROM artist me LEFT JOIN cd cds ON cds.artist = me.artistid
752 # which actually returns the number of artists * (number of cds || 1)
754 # So what we do here is crawl {from}, determine if the current alias is at
755 # the top of the stack, and if not - make sure the chain is inner-joined down
758 sub _inner_join_to_node {
759 my ($self, $from, $alias) = @_;
761 # subqueries and other oddness are naturally not supported
767 ref $from->[0] ne 'HASH'
771 $from->[0]{-alias} eq $alias # this last bit means $alias is the head of $from - nothing to do
774 # find the current $alias in the $from structure
777 for my $j (@{$from}[1 .. $#$from]) {
778 if ($j->[0]{-alias} eq $alias) {
779 $switch_branch = $j->[0]{-join_path};
784 # something else went quite wrong
785 return $from unless $switch_branch;
787 # So it looks like we will have to switch some stuff around.
788 # local() is useless here as we will be leaving the scope
789 # anyway, and deep cloning is just too fucking expensive
790 # So replace the first hashref in the node arrayref manually
791 my @new_from = ($from->[0]);
792 my $sw_idx = { map { (values %$_), 1 } @$switch_branch }; #there's one k/v per join-path
794 for my $j (@{$from}[1 .. $#$from]) {
795 my $jalias = $j->[0]{-alias};
797 if ($sw_idx->{$jalias}) {
798 my %attrs = %{$j->[0]};
799 delete $attrs{-join_type};
813 sub _extract_order_criteria {
814 my ($self, $order_by, $sql_maker, $ident_only) = @_;
816 $sql_maker ||= $self->sql_maker;
818 my $order_dq = $sql_maker->converter->_order_by_to_dq($order_by);
821 while (is_Order($order_dq)) {
822 push @by, $order_dq->{by};
823 $order_dq = $order_dq->{from};
826 delete local @{$sql_maker}{qw(quote_char renderer converter)};
828 return map { [ $sql_maker->_render_dq($_) ] } do {
831 scan_dq_nodes({ DQ_IDENTIFIER ,=> sub { push @by_ident, $_[0] } }, @by);
839 sub _order_by_is_stable {
840 my ($self, $ident, $order_by, $where) = @_;
842 my $colinfo = $self->_resolve_column_info($ident, [
843 (map { $_->[0] } $self->_extract_order_criteria($order_by, undef, 1)),
844 $where ? @{$self->_extract_fixed_condition_columns($where)} :(),
847 return undef unless keys %$colinfo;
850 $cols_per_src->{$_->{-source_alias}}{$_->{-colname}} = $_ for values %$colinfo;
852 for (values %$cols_per_src) {
853 my $src = (values %$_)[0]->{-result_source};
854 return 1 if $src->_identifying_column_set($_);
860 # this is almost identical to the above, except it accepts only
861 # a single rsrc, and will succeed only if the first portion of the order
863 # returns that portion as a colinfo hashref on success
864 sub _main_source_order_by_portion_is_stable {
865 my ($self, $main_rsrc, $order_by, $where) = @_;
867 die "Huh... I expect a blessed result_source..."
868 if ref($main_rsrc) eq 'ARRAY';
872 ( $self->_extract_order_criteria($order_by) )
874 return unless @ord_cols;
876 my $colinfos = $self->_resolve_column_info($main_rsrc);
878 for (0 .. $#ord_cols) {
880 ! $colinfos->{$ord_cols[$_]}
882 $colinfos->{$ord_cols[$_]}{-result_source} != $main_rsrc
889 # we just truncated it above
890 return unless @ord_cols;
892 my $order_portion_ci = { map {
893 $colinfos->{$_}{-colname} => $colinfos->{$_},
894 $colinfos->{$_}{-fq_colname} => $colinfos->{$_},
897 # since all we check here are the start of the order_by belonging to the
898 # top level $rsrc, a present identifying set will mean that the resultset
899 # is ordered by its leftmost table in a stable manner
901 # RV of _identifying_column_set contains unqualified names only
902 my $unqualified_idset = $main_rsrc->_identifying_column_set({
904 $self->_resolve_column_info(
905 $main_rsrc, $self->_extract_fixed_condition_columns($where)
912 my %unqualified_idcols_from_order = map {
913 $order_portion_ci->{$_} ? ( $_ => $order_portion_ci->{$_} ) : ()
914 } @$unqualified_idset;
916 # extra optimization - cut the order_by at the end of the identifying set
917 # (just in case the user was stupid and overlooked the obvious)
918 for my $i (0 .. $#ord_cols) {
919 my $col = $ord_cols[$i];
920 my $unqualified_colname = $order_portion_ci->{$col}{-colname};
921 $ret_info->{$col} = { %{$order_portion_ci->{$col}}, -idx_in_order_subset => $i };
922 delete $unqualified_idcols_from_order{$ret_info->{$col}{-colname}};
924 # we didn't reach the end of the identifying portion yet
925 return $ret_info unless keys %unqualified_idcols_from_order;
928 die 'How did we get here...';
931 # returns an arrayref of column names which *definitely* have some
932 # sort of non-nullable equality requested in the given condition
933 # specification. This is used to figure out if a resultset is
934 # constrained to a column which is part of a unique constraint,
935 # which in turn allows us to better predict how ordering will behave
938 # this is a rudimentary, incomplete, and error-prone extractor
939 # however this is OK - it is conservative, and if we can not find
940 # something that is in fact there - the stack will recover gracefully
941 # Also - DQ and the mst it rode in on will save us all RSN!!!
942 sub _extract_fixed_condition_columns {
943 my ($self, $where, $nested) = @_;
945 return unless ref $where eq 'HASH';
948 for my $lhs (keys %$where) {
949 if ($lhs =~ /^\-and$/i) {
950 push @cols, ref $where->{$lhs} eq 'ARRAY'
951 ? ( map { $self->_extract_fixed_condition_columns($_, 1) } @{$where->{$lhs}} )
952 : $self->_extract_fixed_condition_columns($where->{$lhs}, 1)
955 elsif ($lhs !~ /^\-/) {
956 my $val = $where->{$lhs};
958 push @cols, $lhs if (defined $val and (
961 (ref $val eq 'HASH' and keys %$val == 1 and defined $val->{'='})
965 return $nested ? @cols : \@cols;