Commit | Line | Data |
9f98c4b2 |
1 | package # hide from the pauses |
2 | DBIx::Class::ResultSource::RowParser::Util; |
3 | |
4 | use strict; |
5 | use warnings; |
6 | |
7 | use B 'perlstring'; |
8 | |
9 | use base 'Exporter'; |
10 | our @EXPORT_OK = qw( |
11 | assemble_simple_parser |
12 | assemble_collapsing_parser |
13 | ); |
14 | |
15 | sub assemble_simple_parser { |
16 | #my ($args) = @_; |
17 | |
18 | # the non-collapsing assembler is easy |
19 | # FIXME SUBOPTIMAL there could be a yet faster way to do things here, but |
20 | # need to try an actual implementation and benchmark it: |
21 | # |
22 | # <timbunce_> First setup the nested data structure you want for each row |
23 | # Then call bind_col() to alias the row fields into the right place in |
24 | # the data structure, then to fetch the data do: |
25 | # push @rows, dclone($row_data_struct) while ($sth->fetchrow); |
26 | # |
27 | my $parser_src = sprintf('$_ = %s for @{$_[0]}', __visit_infmap_simple($_[0]) ); |
28 | |
29 | # change the quoted placeholders to unquoted alias-references |
30 | $parser_src =~ s/ \' \xFF__VALPOS__(\d+)__\xFF \' /"\$_->[$1]"/gex; |
31 | |
32 | return $parser_src; |
33 | } |
34 | |
35 | # the simple non-collapsing nested structure recursor |
36 | sub __visit_infmap_simple { |
37 | my $args = shift; |
38 | |
39 | my $my_cols = {}; |
40 | my $rel_cols; |
41 | for (keys %{$args->{val_index}}) { |
42 | if ($_ =~ /^ ([^\.]+) \. (.+) /x) { |
43 | $rel_cols->{$1}{$2} = $args->{val_index}{$_}; |
44 | } |
45 | else { |
46 | $my_cols->{$_} = $args->{val_index}{$_}; |
47 | } |
48 | } |
49 | |
50 | my @relperl; |
51 | for my $rel (sort keys %$rel_cols) { |
52 | |
53 | # DISABLEPRUNE |
54 | #my $optional = $args->{is_optional}; |
55 | #$optional ||= ($args->{rsrc}->relationship_info($rel)->{attrs}{join_type} || '') =~ /^left/i; |
56 | |
57 | push @relperl, join ' => ', perlstring($rel), __visit_infmap_simple({ %$args, |
58 | val_index => $rel_cols->{$rel}, |
59 | # DISABLEPRUNE |
60 | #non_top => 1, |
61 | #is_optional => $optional, |
62 | }); |
63 | |
64 | # FIXME SUBOPTIMAL DISABLEPRUNE - disabled to satisfy t/resultset/inflate_result_api.t |
65 | #if ($optional and my @branch_null_checks = map |
66 | # { "(! defined '\xFF__VALPOS__${_}__\xFF')" } |
67 | # sort { $a <=> $b } values %{$rel_cols->{$rel}} |
68 | #) { |
69 | # $relperl[-1] = sprintf ( '(%s) ? ( %s => [] ) : ( %s )', |
70 | # join (' && ', @branch_null_checks ), |
71 | # perlstring($rel), |
72 | # $relperl[-1], |
73 | # ); |
74 | #} |
75 | } |
76 | |
77 | my $me_struct = keys %$my_cols |
78 | ? __visit_dump({ map { $_ => "\xFF__VALPOS__$my_cols->{$_}__\xFF" } (keys %$my_cols) }) |
79 | : 'undef' |
80 | ; |
81 | |
82 | return sprintf '[%s]', join (',', |
83 | $me_struct, |
84 | @relperl ? sprintf ('{ %s }', join (',', @relperl)) : (), |
85 | ); |
86 | } |
87 | |
88 | sub assemble_collapsing_parser { |
89 | my $args = shift; |
90 | |
91 | my ($top_node_key, $top_node_key_assembler); |
92 | |
93 | if (scalar @{$args->{collapse_map}{-identifying_columns}}) { |
94 | $top_node_key = join ('', map |
95 | { "{'\xFF__IDVALPOS__${_}__\xFF'}" } |
96 | @{$args->{collapse_map}{-identifying_columns}} |
97 | ); |
98 | } |
99 | elsif( my @variants = @{$args->{collapse_map}{-identifying_columns_variants}} ) { |
100 | |
101 | my @path_parts = map { sprintf |
102 | "( ( defined '\xFF__VALPOS__%d__\xFF' ) && (join qq(\xFF), '', %s, '') )", |
103 | $_->[0], # checking just first is enough - one defined, all defined |
104 | ( join ', ', map { "'\xFF__VALPOS__${_}__\xFF'" } @$_ ), |
105 | } @variants; |
106 | |
107 | my $virtual_column_idx = (scalar keys %{$args->{val_index}} ) + 1; |
108 | |
109 | $top_node_key_assembler = sprintf '$cur_row_ids{%d} = (%s);', |
110 | $virtual_column_idx, |
111 | "\n" . join( "\n or\n", @path_parts, qq{"\0\$rows_pos\0"} ); |
112 | |
113 | $top_node_key = sprintf '{$cur_row_ids{%d}}', $virtual_column_idx; |
114 | |
115 | $args->{collapse_map} = { |
116 | %{$args->{collapse_map}}, |
117 | -custom_node_key => $top_node_key, |
118 | }; |
119 | |
120 | } |
121 | else { |
122 | die('Unexpected collapse map contents'); |
123 | } |
124 | |
125 | my ($data_assemblers, $stats) = __visit_infmap_collapse ($args); |
126 | |
127 | my $list_of_idcols = join(', ', sort { $a <=> $b } keys %{ $stats->{idcols_seen} } ); |
128 | |
129 | my $parser_src = sprintf (<<'EOS', $list_of_idcols, $top_node_key, $top_node_key_assembler||'', $data_assemblers); |
130 | ### BEGIN LITERAL STRING EVAL |
131 | my ($rows_pos, $result_pos, $cur_row_data, %%cur_row_ids, @collapse_idx, $is_new_res) = (0,0); |
132 | |
133 | # this loop is a bit arcane - the rationale is that the passed in |
134 | # $_[0] will either have only one row (->next) or will have all |
135 | # rows already pulled in (->all and/or unordered). Given that the |
136 | # result can be rather large - we reuse the same already allocated |
137 | # array, since the collapsed prefetch is smaller by definition. |
138 | # At the end we cut the leftovers away and move on. |
139 | while ($cur_row_data = |
140 | ( ( $rows_pos >= 0 and $_[0][$rows_pos++] ) or do { $rows_pos = -1; undef } ) |
141 | || |
142 | ($_[1] and $_[1]->()) |
143 | ) { |
144 | |
145 | # due to left joins some of the ids may be NULL/undef, and |
146 | # won't play well when used as hash lookups |
147 | # we also need to differentiate NULLs on per-row/per-col basis |
148 | #(otherwise folding of optional 1:1s will be greatly confused |
149 | $cur_row_ids{$_} = defined $cur_row_data->[$_] ? $cur_row_data->[$_] : "\0NULL\xFF$rows_pos\xFF$_\0" |
150 | for (%1$s); |
151 | |
152 | # maybe(!) cache the top node id calculation |
153 | %3$s |
154 | |
155 | $is_new_res = ! $collapse_idx[0]%2$s and ( |
156 | $_[1] and $result_pos and (unshift @{$_[2]}, $cur_row_data) and last |
157 | ); |
158 | |
159 | # the rel assemblers |
160 | %4$s |
161 | |
162 | $_[0][$result_pos++] = $collapse_idx[0]%2$s |
163 | if $is_new_res; |
164 | } |
165 | |
166 | splice @{$_[0]}, $result_pos; # truncate the passed in array for cases of collapsing ->all() |
167 | ### END LITERAL STRING EVAL |
168 | EOS |
169 | |
170 | # !!! note - different var than the one above |
171 | # change the quoted placeholders to unquoted alias-references |
172 | $parser_src =~ s/ \' \xFF__VALPOS__(\d+)__\xFF \' /"\$cur_row_data->[$1]"/gex; |
173 | $parser_src =~ s/ \' \xFF__IDVALPOS__(\d+)__\xFF \' /"\$cur_row_ids{$1}"/gex; |
174 | |
175 | $parser_src; |
176 | } |
177 | |
178 | |
179 | # the collapsing nested structure recursor |
180 | sub __visit_infmap_collapse { |
181 | my $args = {%{ shift() }}; |
182 | |
183 | my $cur_node_idx = ${ $args->{-node_idx_counter} ||= \do { my $x = 0} }++; |
184 | |
185 | my ($my_cols, $rel_cols); |
186 | for ( keys %{$args->{val_index}} ) { |
187 | if ($_ =~ /^ ([^\.]+) \. (.+) /x) { |
188 | $rel_cols->{$1}{$2} = $args->{val_index}{$_}; |
189 | } |
190 | else { |
191 | $my_cols->{$_} = $args->{val_index}{$_}; |
192 | } |
193 | } |
194 | |
195 | my $node_key = $args->{collapse_map}->{-custom_node_key} || join ('', map |
196 | { "{'\xFF__IDVALPOS__${_}__\xFF'}" } |
197 | @{$args->{collapse_map}->{-identifying_columns}} |
198 | ); |
199 | |
200 | my $me_struct = $my_cols |
201 | ? __visit_dump([{ map { $_ => "\xFF__VALPOS__$my_cols->{$_}__\xFF" } (keys %$my_cols) }]) |
202 | : undef |
203 | ; |
204 | my $node_idx_slot = sprintf '$collapse_idx[%d]%s', $cur_node_idx, $node_key; |
205 | |
206 | my $parent_attach_slot = sprintf( '$collapse_idx[%d]%s[1]{%s}', |
207 | @{$args}{qw/-parent_node_idx -parent_node_key/}, |
208 | perlstring($args->{-node_relname}), |
209 | ) if $args->{-node_relname}; |
210 | |
211 | my @src; |
212 | if ($cur_node_idx == 0) { |
213 | push @src, sprintf( '%s ||= %s;', |
214 | $node_idx_slot, |
215 | $me_struct, |
216 | ) if $me_struct; |
217 | } |
218 | elsif ($args->{collapse_map}->{-is_single}) { |
219 | push @src, sprintf ( '%s ||= %s%s;', |
220 | $parent_attach_slot, |
221 | $node_idx_slot, |
222 | $me_struct ? " ||= $me_struct" : '', |
223 | ); |
224 | } |
225 | else { |
226 | push @src, sprintf('push @{%s}, %s%s unless %s;', |
227 | $parent_attach_slot, |
228 | $node_idx_slot, |
229 | $me_struct ? " ||= $me_struct" : '', |
230 | $node_idx_slot, |
231 | ); |
232 | } |
233 | |
234 | # DISABLEPRUNE |
235 | #my $known_defined = { %{ $parent_info->{known_defined} || {} } }; |
236 | #$known_defined->{$_}++ for @{$args->{collapse_map}->{-identifying_columns}}; |
237 | my $stats; |
238 | for my $rel (sort keys %$rel_cols) { |
239 | |
240 | # push @src, sprintf( |
241 | # '%s[1]{%s} ||= [];', $node_idx_slot, perlstring($rel) |
242 | # ) unless $args->{collapse_map}->{$rel}{-is_single}; |
243 | |
244 | ($src[$#src + 1], $stats->{$rel}) = __visit_infmap_collapse({ %$args, |
245 | val_index => $rel_cols->{$rel}, |
246 | collapse_map => $args->{collapse_map}{$rel}, |
247 | -parent_node_idx => $cur_node_idx, |
248 | -parent_node_key => $node_key, |
249 | -node_relname => $rel, |
250 | }); |
251 | |
252 | # FIXME SUBOPTIMAL DISABLEPRUNE - disabled to satisfy t/resultset/inflate_result_api.t |
253 | #if ($args->{collapse_map}->{$rel}{-is_optional} and my @null_checks = map |
254 | # { "(! defined '\xFF__IDVALPOS__${_}__\xFF')" } |
255 | # sort { $a <=> $b } grep |
256 | # { ! $known_defined->{$_} } |
257 | # @{$args->{collapse_map}->{$rel}{-identifying_columns}} |
258 | #) { |
259 | # $src[-1] = sprintf( '(%s) or %s', |
260 | # join (' || ', @null_checks ), |
261 | # $src[-1], |
262 | # ); |
263 | #} |
264 | } |
265 | |
266 | return ( |
267 | join("\n", @src), |
268 | { |
269 | idcols_seen => { |
270 | ( map { %{ $_->{idcols_seen} } } values %$stats ), |
271 | ( map { $_ => 1 } @{$args->{collapse_map}->{-identifying_columns}} ), |
272 | } |
273 | } |
274 | ); |
275 | } |
276 | |
277 | # keep our own DD object around so we don't have to fitz with quoting |
278 | my $dumper_obj; |
279 | sub __visit_dump { |
280 | # we actually will be producing functional perl code here, |
281 | # thus no second-guessing of what these globals might have |
282 | # been set to. DO NOT CHANGE! |
283 | ($dumper_obj ||= do { |
284 | require Data::Dumper; |
285 | Data::Dumper->new([]) |
286 | ->Useperl (0) |
287 | ->Purity (1) |
288 | ->Pad ('') |
289 | ->Useqq (0) |
290 | ->Terse (1) |
291 | ->Quotekeys (1) |
292 | ->Deepcopy (0) |
293 | ->Deparse (0) |
294 | ->Maxdepth (0) |
295 | ->Indent (0) # faster but harder to read, perhaps leave at 1 ? |
296 | })->Values ([$_[0]])->Dump; |
297 | } |
298 | |
299 | 1; |