Commit | Line | Data |
9f98c4b2 |
1 | package # hide from the pauses |
2 | DBIx::Class::ResultSource::RowParser::Util; |
3 | |
4 | use strict; |
5 | use warnings; |
6 | |
ce556881 |
7 | use List::Util 'first'; |
9f98c4b2 |
8 | use B 'perlstring'; |
9 | |
10 | use base 'Exporter'; |
11 | our @EXPORT_OK = qw( |
12 | assemble_simple_parser |
13 | assemble_collapsing_parser |
14 | ); |
15 | |
16 | sub assemble_simple_parser { |
17 | #my ($args) = @_; |
18 | |
19 | # the non-collapsing assembler is easy |
20 | # FIXME SUBOPTIMAL there could be a yet faster way to do things here, but |
21 | # need to try an actual implementation and benchmark it: |
22 | # |
23 | # <timbunce_> First setup the nested data structure you want for each row |
24 | # Then call bind_col() to alias the row fields into the right place in |
25 | # the data structure, then to fetch the data do: |
26 | # push @rows, dclone($row_data_struct) while ($sth->fetchrow); |
27 | # |
28 | my $parser_src = sprintf('$_ = %s for @{$_[0]}', __visit_infmap_simple($_[0]) ); |
29 | |
30 | # change the quoted placeholders to unquoted alias-references |
31 | $parser_src =~ s/ \' \xFF__VALPOS__(\d+)__\xFF \' /"\$_->[$1]"/gex; |
32 | |
33 | return $parser_src; |
34 | } |
35 | |
36 | # the simple non-collapsing nested structure recursor |
37 | sub __visit_infmap_simple { |
38 | my $args = shift; |
39 | |
40 | my $my_cols = {}; |
41 | my $rel_cols; |
42 | for (keys %{$args->{val_index}}) { |
43 | if ($_ =~ /^ ([^\.]+) \. (.+) /x) { |
44 | $rel_cols->{$1}{$2} = $args->{val_index}{$_}; |
45 | } |
46 | else { |
47 | $my_cols->{$_} = $args->{val_index}{$_}; |
48 | } |
49 | } |
50 | |
51 | my @relperl; |
52 | for my $rel (sort keys %$rel_cols) { |
53 | |
9f98c4b2 |
54 | push @relperl, join ' => ', perlstring($rel), __visit_infmap_simple({ %$args, |
55 | val_index => $rel_cols->{$rel}, |
9f98c4b2 |
56 | }); |
57 | |
ce556881 |
58 | if ($args->{prune_null_branches} and keys %$my_cols) { |
59 | |
60 | my @branch_null_checks = map |
61 | { "( ! defined '\xFF__VALPOS__${_}__\xFF' )" } |
62 | sort { $a <=> $b } values %{$rel_cols->{$rel}} |
63 | ; |
64 | |
65 | $relperl[-1] = sprintf ( '(%s) ? ( %s => %s ) : ( %s )', |
66 | join (' && ', @branch_null_checks ), |
67 | perlstring($rel), |
68 | $args->{hri_style} ? 'undef' : '[]', |
69 | $relperl[-1], |
70 | ); |
71 | } |
9f98c4b2 |
72 | } |
73 | |
ce556881 |
74 | my $me_struct; |
75 | $me_struct = __visit_dump({ map { $_ => "\xFF__VALPOS__$my_cols->{$_}__\xFF" } (keys %$my_cols) }) |
76 | if keys %$my_cols; |
9f98c4b2 |
77 | |
ce556881 |
78 | if ($args->{hri_style}) { |
79 | $me_struct =~ s/^ \s* \{ | \} \s* $//gx |
80 | if $me_struct; |
81 | |
82 | return sprintf '{ %s }', join (', ', $me_struct||(), @relperl); |
83 | } |
84 | else { |
85 | return sprintf '[%s]', join (',', |
86 | $me_struct || 'undef', |
87 | @relperl ? sprintf ('{ %s }', join (',', @relperl)) : (), |
88 | ); |
89 | } |
9f98c4b2 |
90 | } |
91 | |
92 | sub assemble_collapsing_parser { |
93 | my $args = shift; |
94 | |
95 | my ($top_node_key, $top_node_key_assembler); |
96 | |
97 | if (scalar @{$args->{collapse_map}{-identifying_columns}}) { |
98 | $top_node_key = join ('', map |
99 | { "{'\xFF__IDVALPOS__${_}__\xFF'}" } |
100 | @{$args->{collapse_map}{-identifying_columns}} |
101 | ); |
102 | } |
103 | elsif( my @variants = @{$args->{collapse_map}{-identifying_columns_variants}} ) { |
104 | |
105 | my @path_parts = map { sprintf |
106 | "( ( defined '\xFF__VALPOS__%d__\xFF' ) && (join qq(\xFF), '', %s, '') )", |
ce556881 |
107 | $_->[0], # checking just first is enough - one ID defined, all defined |
9f98c4b2 |
108 | ( join ', ', map { "'\xFF__VALPOS__${_}__\xFF'" } @$_ ), |
109 | } @variants; |
110 | |
111 | my $virtual_column_idx = (scalar keys %{$args->{val_index}} ) + 1; |
112 | |
113 | $top_node_key_assembler = sprintf '$cur_row_ids{%d} = (%s);', |
114 | $virtual_column_idx, |
115 | "\n" . join( "\n or\n", @path_parts, qq{"\0\$rows_pos\0"} ); |
116 | |
117 | $top_node_key = sprintf '{$cur_row_ids{%d}}', $virtual_column_idx; |
118 | |
119 | $args->{collapse_map} = { |
120 | %{$args->{collapse_map}}, |
121 | -custom_node_key => $top_node_key, |
122 | }; |
123 | |
124 | } |
125 | else { |
126 | die('Unexpected collapse map contents'); |
127 | } |
128 | |
129 | my ($data_assemblers, $stats) = __visit_infmap_collapse ($args); |
130 | |
131 | my $list_of_idcols = join(', ', sort { $a <=> $b } keys %{ $stats->{idcols_seen} } ); |
132 | |
ce556881 |
133 | my $parser_src = sprintf (<<'EOS', $list_of_idcols, $top_node_key, $top_node_key_assembler||'', join( "\n", @{$data_assemblers||[]} ) ); |
9f98c4b2 |
134 | ### BEGIN LITERAL STRING EVAL |
135 | my ($rows_pos, $result_pos, $cur_row_data, %%cur_row_ids, @collapse_idx, $is_new_res) = (0,0); |
9f98c4b2 |
136 | # this loop is a bit arcane - the rationale is that the passed in |
137 | # $_[0] will either have only one row (->next) or will have all |
138 | # rows already pulled in (->all and/or unordered). Given that the |
139 | # result can be rather large - we reuse the same already allocated |
140 | # array, since the collapsed prefetch is smaller by definition. |
141 | # At the end we cut the leftovers away and move on. |
142 | while ($cur_row_data = |
143 | ( ( $rows_pos >= 0 and $_[0][$rows_pos++] ) or do { $rows_pos = -1; undef } ) |
144 | || |
145 | ($_[1] and $_[1]->()) |
146 | ) { |
9f98c4b2 |
147 | # due to left joins some of the ids may be NULL/undef, and |
148 | # won't play well when used as hash lookups |
149 | # we also need to differentiate NULLs on per-row/per-col basis |
ce556881 |
150 | # (otherwise folding of optional 1:1s will be greatly confused |
9f98c4b2 |
151 | $cur_row_ids{$_} = defined $cur_row_data->[$_] ? $cur_row_data->[$_] : "\0NULL\xFF$rows_pos\xFF$_\0" |
152 | for (%1$s); |
153 | |
154 | # maybe(!) cache the top node id calculation |
155 | %3$s |
156 | |
157 | $is_new_res = ! $collapse_idx[0]%2$s and ( |
158 | $_[1] and $result_pos and (unshift @{$_[2]}, $cur_row_data) and last |
159 | ); |
160 | |
161 | # the rel assemblers |
162 | %4$s |
163 | |
164 | $_[0][$result_pos++] = $collapse_idx[0]%2$s |
165 | if $is_new_res; |
166 | } |
167 | |
168 | splice @{$_[0]}, $result_pos; # truncate the passed in array for cases of collapsing ->all() |
169 | ### END LITERAL STRING EVAL |
170 | EOS |
171 | |
172 | # !!! note - different var than the one above |
173 | # change the quoted placeholders to unquoted alias-references |
174 | $parser_src =~ s/ \' \xFF__VALPOS__(\d+)__\xFF \' /"\$cur_row_data->[$1]"/gex; |
175 | $parser_src =~ s/ \' \xFF__IDVALPOS__(\d+)__\xFF \' /"\$cur_row_ids{$1}"/gex; |
176 | |
177 | $parser_src; |
178 | } |
179 | |
180 | |
181 | # the collapsing nested structure recursor |
182 | sub __visit_infmap_collapse { |
183 | my $args = {%{ shift() }}; |
184 | |
185 | my $cur_node_idx = ${ $args->{-node_idx_counter} ||= \do { my $x = 0} }++; |
186 | |
ce556881 |
187 | my ($my_cols, $rel_cols) = {}; |
9f98c4b2 |
188 | for ( keys %{$args->{val_index}} ) { |
189 | if ($_ =~ /^ ([^\.]+) \. (.+) /x) { |
190 | $rel_cols->{$1}{$2} = $args->{val_index}{$_}; |
191 | } |
192 | else { |
193 | $my_cols->{$_} = $args->{val_index}{$_}; |
194 | } |
195 | } |
196 | |
ce556881 |
197 | |
9f98c4b2 |
198 | my $node_key = $args->{collapse_map}->{-custom_node_key} || join ('', map |
199 | { "{'\xFF__IDVALPOS__${_}__\xFF'}" } |
200 | @{$args->{collapse_map}->{-identifying_columns}} |
201 | ); |
202 | |
ce556881 |
203 | my $me_struct; |
204 | |
205 | if ($args->{hri_style}) { |
206 | delete $my_cols->{$_} for grep { $rel_cols->{$_} } keys %$my_cols; |
207 | } |
9f98c4b2 |
208 | |
ce556881 |
209 | if (keys %$my_cols) { |
210 | $me_struct = __visit_dump({ map { $_ => "\xFF__VALPOS__$my_cols->{$_}__\xFF" } (keys %$my_cols) }); |
211 | $me_struct = "[ $me_struct ]" unless $args->{hri_style}; |
212 | } |
213 | |
214 | my $node_idx_slot = sprintf '$collapse_idx[%d]%s', $cur_node_idx, $node_key; |
9f98c4b2 |
215 | |
216 | my @src; |
ce556881 |
217 | |
9f98c4b2 |
218 | if ($cur_node_idx == 0) { |
219 | push @src, sprintf( '%s ||= %s;', |
220 | $node_idx_slot, |
221 | $me_struct, |
222 | ) if $me_struct; |
223 | } |
9f98c4b2 |
224 | else { |
ce556881 |
225 | my $parent_attach_slot = sprintf( '$collapse_idx[%d]%s%s{%s}', |
226 | @{$args}{qw/-parent_node_idx -parent_node_key/}, |
227 | $args->{hri_style} ? '' : '[1]', |
228 | perlstring($args->{-node_relname}), |
9f98c4b2 |
229 | ); |
ce556881 |
230 | |
231 | if ($args->{collapse_map}->{-is_single}) { |
232 | push @src, sprintf ( '%s ||= %s%s;', |
233 | $parent_attach_slot, |
234 | $node_idx_slot, |
235 | $me_struct ? " ||= $me_struct" : '', |
236 | ); |
237 | } |
238 | else { |
239 | push @src, sprintf('(! %s) and push @{%s}, %s%s;', |
240 | $node_idx_slot, |
241 | $parent_attach_slot, |
242 | $node_idx_slot, |
243 | $me_struct ? " = $me_struct" : '', |
244 | ); |
245 | } |
9f98c4b2 |
246 | } |
247 | |
ce556881 |
248 | my $known_present_ids = { map { $_ => 1 } @{$args->{collapse_map}{-identifying_columns}} }; |
249 | my ($stats, $rel_src); |
250 | |
9f98c4b2 |
251 | for my $rel (sort keys %$rel_cols) { |
252 | |
ce556881 |
253 | my $relinfo = $args->{collapse_map}{$rel}; |
254 | if ($args->{collapse_map}{-is_optional}) { |
255 | $relinfo = { %$relinfo, -is_optional => 1 }; |
256 | } |
9f98c4b2 |
257 | |
ce556881 |
258 | ($rel_src, $stats->{$rel}) = __visit_infmap_collapse({ %$args, |
9f98c4b2 |
259 | val_index => $rel_cols->{$rel}, |
ce556881 |
260 | collapse_map => $relinfo, |
9f98c4b2 |
261 | -parent_node_idx => $cur_node_idx, |
262 | -parent_node_key => $node_key, |
263 | -node_relname => $rel, |
264 | }); |
265 | |
ce556881 |
266 | my $rel_src_pos = $#src + 1; |
267 | push @src, @$rel_src; |
268 | |
269 | if ( |
270 | $args->{prune_null_branches} |
271 | and |
272 | $relinfo->{-is_optional} |
273 | and |
274 | defined ( my $first_distinct_child_idcol = first |
275 | { ! $known_present_ids->{$_} } |
276 | @{$relinfo->{-identifying_columns}} |
277 | ) |
278 | ) { |
279 | |
280 | $src[$rel_src_pos] = sprintf( '%s and %s', |
281 | "( defined '\xFF__VALPOS__${first_distinct_child_idcol}__\xFF' )", |
282 | $src[$rel_src_pos], |
283 | ); |
284 | |
285 | splice @src, $rel_src_pos + 1, 0, sprintf ( '%s%s{%s} ||= %s;', |
286 | $node_idx_slot, |
287 | $args->{hri_style} ? '' : '[1]', |
288 | perlstring($rel), |
289 | $args->{hri_style} && $relinfo->{-is_single} ? 'undef' : '[]', |
290 | ); |
291 | } |
9f98c4b2 |
292 | } |
293 | |
294 | return ( |
ce556881 |
295 | \@src, |
9f98c4b2 |
296 | { |
297 | idcols_seen => { |
298 | ( map { %{ $_->{idcols_seen} } } values %$stats ), |
299 | ( map { $_ => 1 } @{$args->{collapse_map}->{-identifying_columns}} ), |
300 | } |
301 | } |
302 | ); |
303 | } |
304 | |
305 | # keep our own DD object around so we don't have to fitz with quoting |
306 | my $dumper_obj; |
307 | sub __visit_dump { |
308 | # we actually will be producing functional perl code here, |
309 | # thus no second-guessing of what these globals might have |
310 | # been set to. DO NOT CHANGE! |
311 | ($dumper_obj ||= do { |
312 | require Data::Dumper; |
313 | Data::Dumper->new([]) |
314 | ->Useperl (0) |
315 | ->Purity (1) |
316 | ->Pad ('') |
317 | ->Useqq (0) |
318 | ->Terse (1) |
319 | ->Quotekeys (1) |
320 | ->Deepcopy (0) |
321 | ->Deparse (0) |
322 | ->Maxdepth (0) |
323 | ->Indent (0) # faster but harder to read, perhaps leave at 1 ? |
324 | })->Values ([$_[0]])->Dump; |
325 | } |
326 | |
327 | 1; |