| @@ -50,9 +50,30 @@ | ||
| 50 | 50 | */ |
| 51 | 51 | private const BUILDER_META_KEYS = [ |
| 52 | 52 | '_breakdance_data', // Oxygen 6+ / Breakdance |
| 53 | 53 | '_oxygen_data', // Oxygen (earlier releases) |
| 54 | - 'ct_builder_shortcodes', // Oxygen classic | |
| 54 | + // Oxygen classic. 4.x writes the tree as JSON to `ct_builder_json` | |
| 55 | + // while still keeping `ct_builder_shortcodes`. A post carrying only | |
| 56 | + // the JSON key used to match no key at all and fall through to an | |
| 57 | + // empty `post_content`, which reads as a one-word page (#776). | |
| 58 | + // | |
| 59 | + // Oxygen 4.8.3 then renamed every `ct_*` post meta key to `_ct_*` | |
| 60 | + // (`oxygen_vsb_update_4_8_3()` runs `oxy_prefix_meta_keys()` on | |
| 61 | + // upgrade, and `oxy_get_post_meta()` only reads the prefixed name | |
| 62 | + // from then on). A current Oxygen classic site therefore has only the | |
| 63 | + // underscored keys, which nothing here listed, so every one of its | |
| 64 | + // pages resolved as empty. The prefixed keys come first because they | |
| 65 | + // are what Oxygen itself reads; the bare ones cover a site that has | |
| 66 | + // not run the migration (or reverted it with `?unprefix_meta`). | |
| 67 | + // | |
| 68 | + // These four are not read in this loop: from_oxygen_classic() pairs | |
| 69 | + // each JSON key with its shortcode sibling so the two forms can be | |
| 70 | + // compared. They are listed here because this list is also what the | |
| 71 | + // word-count index watches and what FAQ detection scans. | |
| 72 | + '_ct_builder_json', // Oxygen classic 4.8.3+ (JSON tree) | |
| 73 | + 'ct_builder_json', // Oxygen classic 4.0-4.8.2 (JSON tree) | |
| 74 | + '_ct_builder_shortcodes', // Oxygen classic 4.8.3+ (shortcode tree) | |
| 75 | + 'ct_builder_shortcodes', // Oxygen classic < 4.8.3 (shortcode tree) | |
| 55 | 76 | '_elementor_data', // Elementor |
| 56 | 77 | // Beaver Builder. Published layout first: `_fl_builder_draft` holds |
| 57 | 78 | // unsaved changes and would score content the visitor cannot see. |
| 58 | 79 | // Both are arrays of stdClass nodes, which is why the walker below |
| @@ -142,11 +163,38 @@ | ||
| 142 | 163 | private const CONTENT_KEYS = [ |
| 143 | 164 | 'text', 'title', 'subtitle', 'heading', 'subheading', 'content', |
| 144 | 165 | 'description', 'caption', 'excerpt', 'label', 'value', 'html', |
| 145 | 166 | 'editor', 'quote', 'answer', 'question', 'body', 'button_text', |
| 167 | + // Oxygen classic keeps an element's copy in `options.ct_content` | |
| 168 | + // (headline, text block, rich text, link and button labels). It is the | |
| 169 | + // field Oxygen's own serializer moves between the tags when it writes | |
| 170 | + // shortcodes (`parse_components_tree()`), and the one Relevanssi and | |
| 171 | + // Oxygen's WPML integration read. Missing from this list, the walker | |
| 172 | + // kept only copy that happened to contain markup: a page of plain | |
| 173 | + // headings and paragraphs lost almost all of its words. | |
| 174 | + 'ct_content', | |
| 175 | + // Oxygen's composite elements keep their copy under `options.original` | |
| 176 | + // instead, one key per field. Taken from the list Oxygen itself treats | |
| 177 | + // as text when it serializes (`$options_to_encode`); the numeric price | |
| 178 | + // fields and the progress bar's right-hand percentage are left out. | |
| 179 | + 'testimonial_text', 'testimonial_author', 'testimonial_author_info', | |
| 180 | + 'icon_box_heading', 'icon_box_text', | |
| 181 | + 'pricing_box_package_title', 'pricing_box_package_subtitle', 'pricing_box_content', | |
| 182 | + 'progress_bar_left_text', | |
| 146 | 183 | ]; |
| 147 | 184 | |
| 148 | 185 | /** |
| 186 | + * Oxygen classic's storage generations, as JSON key => shortcode key. | |
| 187 | + * | |
| 188 | + * @since 2.10.0 | |
| 189 | + * @var array<string,string> | |
| 190 | + */ | |
| 191 | + private const OXYGEN_CLASSIC_KEYS = [ | |
| 192 | + '_ct_builder_json' => '_ct_builder_shortcodes', | |
| 193 | + 'ct_builder_json' => 'ct_builder_shortcodes', | |
| 194 | + ]; | |
| 195 | + | |
| 196 | + /** | |
| 149 | 197 | * JSON keys whose values hold a link destination. |
| 150 | 198 | * |
| 151 | 199 | * Builders store a link's destination in a structured field separate from |
| 152 | 200 | * its label, either as a bare URL string or as a `{ url: … }` object. |
| @@ -255,8 +303,22 @@ | ||
| 255 | 303 | */ |
| 256 | 304 | private const ALT_KEYS = ['alt', 'alt_text', 'image_alt', 'title']; |
| 257 | 305 | |
| 258 | 306 | /** |
| 307 | + * The global post and every global `setup_postdata()` writes. | |
| 308 | + * | |
| 309 | + * Rendering points them at the post being analyzed, then puts each one | |
| 310 | + * back exactly as it was, unset included (#860). | |
| 311 | + * | |
| 312 | + * @since 2.12.0 | |
| 313 | + * @var string[] | |
| 314 | + */ | |
| 315 | + private const POSTDATA_GLOBALS = [ | |
| 316 | + 'post', 'id', 'authordata', 'currentday', 'currentmonth', | |
| 317 | + 'page', 'pages', 'multipage', 'more', 'numpages', | |
| 318 | + ]; | |
| 319 | + | |
| 320 | + /** | |
| 259 | 321 | * Resolve the content worth analyzing for a post. |
| 260 | 322 | * |
| 261 | 323 | * @param \WP_Post $post Post being analyzed. |
| 262 | 324 | * @return string HTML/text to analyze. |
| @@ -506,9 +568,9 @@ | ||
| 506 | 568 | * @param \WP_Post $post Post the markup belongs to. |
| 507 | 569 | * @return string Content to analyze. |
| 508 | 570 | */ |
| 509 | 571 | public static function resolve_markup(string $raw, \WP_Post $post): string { |
| 510 | - $content = self::render_post_content($raw); | |
| 572 | + $content = self::render_post_content($raw, $post); | |
| 511 | 573 | |
| 512 | 574 | // Block markup that renders to nothing usually means the builder that |
| 513 | 575 | // owns those blocks did not register them in this context — Divi 5 |
| 514 | 576 | // loads its module library lazily per-request, so in CLI, REST, admin |
| @@ -556,19 +618,45 @@ | ||
| 556 | 618 | * |
| 557 | 619 | * Best-effort: a third-party block that fatals must not take the whole |
| 558 | 620 | * score down with it. |
| 559 | 621 | * |
| 560 | - * @param string $raw Raw post content. | |
| 622 | + * Runs as the post's own context, as it would on the front end. Admin and | |
| 623 | + * REST requests have no current post, so a shortcode reading | |
| 624 | + * `get_the_ID()` got nothing, and one looping a related-posts query left | |
| 625 | + * the global post on the last of them: its `wp_reset_postdata()` goes back | |
| 626 | + * to the main query's post, and there is none. On the Classic Editor this | |
| 627 | + * runs after the form prints its hidden `post_ID` and before the title and | |
| 628 | + * editor, which then showed the related post, and Update saved it over the | |
| 629 | + * original (#860). | |
| 630 | + * | |
| 631 | + * @param string $raw Raw post content. | |
| 632 | + * @param \WP_Post $post Post the content belongs to. | |
| 561 | 633 | * @return string Rendered content. |
| 562 | 634 | */ |
| 563 | - private static function render_post_content(string $raw): string { | |
| 635 | + private static function render_post_content(string $raw, \WP_Post $post): string { | |
| 564 | 636 | if ('' === trim($raw)) { |
| 565 | 637 | return ''; |
| 566 | 638 | } |
| 567 | 639 | |
| 568 | - $content = $raw; | |
| 640 | + $content = $raw; | |
| 641 | + $previous = self::snapshot_post_globals(); | |
| 569 | 642 | |
| 570 | 643 | try { |
| 644 | + // phpcs:ignore WordPress.WP.GlobalVariablesOverride.Prohibited -- Made current for the render, restored in finally. | |
| 645 | + $GLOBALS['post'] = $post; | |
| 646 | + | |
| 647 | + // Fires `the_post`, which these paths never fired before: admin, | |
| 648 | + // REST and cron analysis had no current post at all. That is the | |
| 649 | + // same signal the front-end loop sends and it is what makes | |
| 650 | + // `get_the_ID()` work inside a shortcode, but it is a new call on a | |
| 651 | + // path that runs in bulk — the word-count index resolves every post | |
| 652 | + // it visits — so a theme that counts views on `the_post` will count | |
| 653 | + // them during indexing. Accepted deliberately: without it a | |
| 654 | + // shortcode cannot resolve its own post, which is the bug (#860). | |
| 655 | + if (function_exists('setup_postdata')) { | |
| 656 | + setup_postdata($post); | |
| 657 | + } | |
| 658 | + | |
| 571 | 659 | if (function_exists('has_blocks') && function_exists('do_blocks') && has_blocks($raw)) { |
| 572 | 660 | $content = do_blocks($raw); |
| 573 | 661 | } |
| 574 | 662 | |
| @@ -577,8 +665,10 @@ | ||
| 577 | 665 | $content = do_shortcode($content); |
| 578 | 666 | } |
| 579 | 667 | } catch (\Throwable $e) { |
| 580 | 668 | return $raw; |
| 669 | + } finally { | |
| 670 | + self::restore_post_globals($previous); | |
| 581 | 671 | } |
| 582 | 672 | |
| 583 | 673 | return self::is_blank($content) ? $raw : $content; |
| 584 | 674 | } |
| @@ -583,8 +673,49 @@ | ||
| 583 | 673 | return self::is_blank($content) ? $raw : $content; |
| 584 | 674 | } |
| 585 | 675 | |
| 586 | 676 | /** |
| 677 | + * The post globals as they are now; a global that is unset has no key. | |
| 678 | + * | |
| 679 | + * @since 2.12.0 | |
| 680 | + * | |
| 681 | + * @return array<string,mixed> | |
| 682 | + */ | |
| 683 | + private static function snapshot_post_globals(): array { | |
| 684 | + $snapshot = []; | |
| 685 | + | |
| 686 | + foreach (self::POSTDATA_GLOBALS as $name) { | |
| 687 | + if (array_key_exists($name, $GLOBALS)) { | |
| 688 | + $snapshot[$name] = $GLOBALS[$name]; | |
| 689 | + } | |
| 690 | + } | |
| 691 | + | |
| 692 | + return $snapshot; | |
| 693 | + } | |
| 694 | + | |
| 695 | + /** | |
| 696 | + * Put the post globals back as snapshot_post_globals() found them. | |
| 697 | + * | |
| 698 | + * Assigned directly rather than through `setup_postdata()`: there may have | |
| 699 | + * been no post to set up, and re-running it would fire `the_post` again. | |
| 700 | + * | |
| 701 | + * @since 2.12.0 | |
| 702 | + * | |
| 703 | + * @param array<string,mixed> $snapshot From snapshot_post_globals(). | |
| 704 | + * @return void | |
| 705 | + */ | |
| 706 | + private static function restore_post_globals(array $snapshot): void { | |
| 707 | + foreach (self::POSTDATA_GLOBALS as $name) { | |
| 708 | + if (array_key_exists($name, $snapshot)) { | |
| 709 | + // phpcs:ignore WordPress.NamingConventions.PrefixAllGlobals.NonPrefixedVariableFound -- Core's own globals, put back as they were. | |
| 710 | + $GLOBALS[$name] = $snapshot[$name]; | |
| 711 | + } else { | |
| 712 | + unset($GLOBALS[$name]); | |
| 713 | + } | |
| 714 | + } | |
| 715 | + } | |
| 716 | + | |
| 717 | + /** | |
| 587 | 718 | * Extract text from the attributes of parsed blocks. |
| 588 | 719 | * |
| 589 | 720 | * @param string $raw Raw post content containing block markup. |
| 590 | 721 | * @return string Collected text, or '' when nothing was found. |
| @@ -972,8 +1103,21 @@ | ||
| 972 | 1103 | return $bricks; |
| 973 | 1104 | } |
| 974 | 1105 | |
| 975 | 1106 | foreach (self::BUILDER_META_KEYS as $key) { |
| 1107 | + if (self::is_oxygen_classic_key($key)) { | |
| 1108 | + // Resolved as a pair, once, at the first of its keys. | |
| 1109 | + if ('_ct_builder_json' !== $key) { | |
| 1110 | + continue; | |
| 1111 | + } | |
| 1112 | + | |
| 1113 | + $oxygen = self::from_oxygen_classic($post_id); | |
| 1114 | + if (!self::is_blank($oxygen)) { | |
| 1115 | + return $oxygen; | |
| 1116 | + } | |
| 1117 | + continue; | |
| 1118 | + } | |
| 1119 | + | |
| 976 | 1120 | $stored = get_post_meta($post_id, $key, true); |
| 977 | 1121 | |
| 978 | 1122 | if (is_string($stored) && '' !== trim($stored)) { |
| 979 | 1123 | $decoded = json_decode($stored, true); |
| @@ -986,20 +1130,8 @@ | ||
| 986 | 1130 | } |
| 987 | 1131 | continue; |
| 988 | 1132 | } |
| 989 | 1133 | |
| 990 | - // Shortcode tree (Oxygen classic). | |
| 991 | - if (strpos($stored, '[') !== false && function_exists('do_shortcode')) { | |
| 992 | - try { | |
| 993 | - $rendered = do_shortcode($stored); | |
| 994 | - } catch (\Throwable $e) { | |
| 995 | - $rendered = $stored; | |
| 996 | - } | |
| 997 | - if (!self::is_blank($rendered)) { | |
| 998 | - return $rendered; | |
| 999 | - } | |
| 1000 | - } | |
| 1001 | - | |
| 1002 | 1134 | continue; |
| 1003 | 1135 | } |
| 1004 | 1136 | |
| 1005 | 1137 | // Some builders store an already-decoded tree — an array for most, |
| @@ -1013,8 +1145,258 @@ | ||
| 1013 | 1145 | } |
| 1014 | 1146 | } |
| 1015 | 1147 | |
| 1016 | 1148 | return ''; |
| 1149 | + } | |
| 1150 | + | |
| 1151 | + /** | |
| 1152 | + * Whether a meta key is one of Oxygen classic's storage keys. | |
| 1153 | + * | |
| 1154 | + * @since 2.10.0 | |
| 1155 | + * | |
| 1156 | + * @param string $key Meta key. | |
| 1157 | + * @return bool | |
| 1158 | + */ | |
| 1159 | + private static function is_oxygen_classic_key(string $key): bool { | |
| 1160 | + return isset(self::OXYGEN_CLASSIC_KEYS[$key]) || in_array($key, self::OXYGEN_CLASSIC_KEYS, true); | |
| 1161 | + } | |
| 1162 | + | |
| 1163 | + /** | |
| 1164 | + * Text of an Oxygen classic page, from whichever stored form holds more. | |
| 1165 | + * | |
| 1166 | + * Oxygen 4.x keeps the same tree twice: as JSON, and as the shortcodes it | |
| 1167 | + * used before 4.0. The JSON is preferred because it carries copy the | |
| 1168 | + * shortcode form hides (a composite element's text is base64-encoded | |
| 1169 | + * inside `ct_options`, which is configuration and stripped). It is not | |
| 1170 | + * trusted blindly, though. Reading `ct_builder_json` first once meant a | |
| 1171 | + * key missing from CONTENT_KEYS silently threw the page away while the | |
| 1172 | + * shortcode copy sat unread next to it, because a non-empty JSON result | |
| 1173 | + * stopped the search. Comparing the two means the next such gap costs | |
| 1174 | + * nothing: the richer form wins. | |
| 1175 | + * | |
| 1176 | + * A generation is only read as a pair. The prefixed keys are what Oxygen | |
| 1177 | + * 4.8.3+ reads, so an unprefixed leftover next to them is stale. | |
| 1178 | + * | |
| 1179 | + * @since 2.10.0 | |
| 1180 | + * | |
| 1181 | + * @param int $post_id Post ID. | |
| 1182 | + * @return string Extracted text, or '' when Oxygen classic stored nothing. | |
| 1183 | + */ | |
| 1184 | + private static function from_oxygen_classic(int $post_id): string { | |
| 1185 | + foreach (self::OXYGEN_CLASSIC_KEYS as $json_key => $shortcode_key) { | |
| 1186 | + $json = get_post_meta($post_id, $json_key, true); | |
| 1187 | + $shortcodes = get_post_meta($post_id, $shortcode_key, true); | |
| 1188 | + | |
| 1189 | + $from_json = ''; | |
| 1190 | + if (is_string($json) && '' !== trim($json)) { | |
| 1191 | + $decoded = json_decode($json, true); | |
| 1192 | + if (is_array($decoded)) { | |
| 1193 | + // `[oxygen data="..."]` is a dynamic-data placeholder | |
| 1194 | + // Oxygen fills at render time. The shortcode path drops it | |
| 1195 | + // with every other tag, so it goes here too or the two | |
| 1196 | + // forms would disagree on the same page. | |
| 1197 | + $from_json = (string) preg_replace( | |
| 1198 | + '/\[oxygen\b[^\]]*\]/i', | |
| 1199 | + ' ', | |
| 1200 | + self::text_from_tree($decoded) | |
| 1201 | + ); | |
| 1202 | + } | |
| 1203 | + } | |
| 1204 | + | |
| 1205 | + $from_shortcodes = ''; | |
| 1206 | + if (is_string($shortcodes) && strpos($shortcodes, '[') !== false) { | |
| 1207 | + $from_shortcodes = self::text_from_shortcodes($shortcodes); | |
| 1208 | + } | |
| 1209 | + | |
| 1210 | + if (self::is_blank($from_json) && self::is_blank($from_shortcodes)) { | |
| 1211 | + continue; | |
| 1212 | + } | |
| 1213 | + | |
| 1214 | + return self::visible_word_count($from_json) >= self::visible_word_count($from_shortcodes) | |
| 1215 | + ? $from_json | |
| 1216 | + : $from_shortcodes; | |
| 1217 | + } | |
| 1218 | + | |
| 1219 | + return ''; | |
| 1220 | + } | |
| 1221 | + | |
| 1222 | + /** | |
| 1223 | + * Rough count of the words a visitor would read in extracted text. | |
| 1224 | + * | |
| 1225 | + * Only used to compare two extractions of the same page, so it needs to | |
| 1226 | + * be consistent rather than locale-exact. | |
| 1227 | + * | |
| 1228 | + * @since 2.10.0 | |
| 1229 | + * | |
| 1230 | + * @param string $text Extracted text or markup. | |
| 1231 | + * @return int | |
| 1232 | + */ | |
| 1233 | + private static function visible_word_count(string $text): int { | |
| 1234 | + $plain = trim((string) preg_replace('/\s+/u', ' ', wp_strip_all_tags($text))); | |
| 1235 | + | |
| 1236 | + return '' === $plain ? 0 : count(explode(' ', $plain)); | |
| 1237 | + } | |
| 1238 | + | |
| 1239 | + /** | |
| 1240 | + * Shortcode attributes that carry copy a visitor reads. | |
| 1241 | + * | |
| 1242 | + * An allow-list, not a deny-list. Oxygen Classic tags carry far more | |
| 1243 | + * attributes than they do copy — `id`, `class`, `selector`, `url`, | |
| 1244 | + * `ct_options` and friends — and a deny-list silently admits every | |
| 1245 | + * attribute a future builder release invents, which is how markup ends up | |
| 1246 | + * being counted as prose. | |
| 1247 | + * | |
| 1248 | + * @var string[] | |
| 1249 | + */ | |
| 1250 | + private const SHORTCODE_TEXT_ATTRIBUTES = [ | |
| 1251 | + 'text', | |
| 1252 | + 'content', | |
| 1253 | + 'heading', | |
| 1254 | + 'title', | |
| 1255 | + 'subtitle', | |
| 1256 | + 'label', | |
| 1257 | + 'caption', | |
| 1258 | + 'description', | |
| 1259 | + 'alt', | |
| 1260 | + 'button_text', | |
| 1261 | + 'link_text', | |
| 1262 | + ]; | |
| 1263 | + | |
| 1264 | + /** | |
| 1265 | + * Extract readable text from a shortcode tree, without rendering it. | |
| 1266 | + * | |
| 1267 | + * Oxygen Classic is the only builder whose storage is shortcodes rather | |
| 1268 | + * than JSON, and the previous implementation handed the string to | |
| 1269 | + * `do_shortcode()`. That silently depends on Oxygen having registered its | |
| 1270 | + * `ct_*` handlers in the current request — which it has on a front-end | |
| 1271 | + * view, and has not during bulk analysis, the post-list column, cron or | |
| 1272 | + * REST/MCP. With no handlers registered `do_shortcode()` returns its input | |
| 1273 | + * unchanged, so the raw shortcode source was scored as if it were the | |
| 1274 | + * page's prose: `[ct_section`, `id="section-1"` and the rest counted toward | |
| 1275 | + * the word count, while the actual copy sitting in `text="..."` attributes | |
| 1276 | + * was never counted at all (#776). | |
| 1277 | + * | |
| 1278 | + * `strip_shortcodes()` is no help either — it also only knows registered | |
| 1279 | + * shortcodes, so it leaves the same text untouched. | |
| 1280 | + * | |
| 1281 | + * Reading the stored tree directly is what every other builder here already | |
| 1282 | + * does, and it matches the class's stated design: no render engine, no | |
| 1283 | + * dependency on load order, safe during a bulk run. | |
| 1284 | + * | |
| 1285 | + * Parsing unconditionally, rather than rendering when Oxygen happens to be | |
| 1286 | + * loaded and parsing otherwise, is deliberate. It makes the extracted text | |
| 1287 | + * the same in every context, so the score in the editor matches the score | |
| 1288 | + * from a bulk run or from MCP. The old code produced whichever of the two | |
| 1289 | + * the request happened to allow, which is why the same post could report | |
| 1290 | + * two different word counts depending on how it was asked. | |
| 1291 | + * | |
| 1292 | + * The trade-off is that rendered output (resolved images, links, anything | |
| 1293 | + * Oxygen pulls in from a reusable part) is no longer reflected here. For | |
| 1294 | + * what this text feeds — word count, content scoring, meta-description | |
| 1295 | + * fallbacks and schema text — that markup was never the point, and counting | |
| 1296 | + * it only when the builder happened to be booted was the bug. | |
| 1297 | + * | |
| 1298 | + * @since 2.10.0 | |
| 1299 | + * | |
| 1300 | + * @param string $stored Raw shortcode source. | |
| 1301 | + * @return string Extracted text. | |
| 1302 | + */ | |
| 1303 | + private static function text_from_shortcodes(string $stored): string { | |
| 1304 | + // Oxygen stores each element's settings as a JSON blob in `ct_options`. | |
| 1305 | + // It is configuration, never copy, and it contains braces and brackets | |
| 1306 | + // that would otherwise confuse the tag scan below, so it goes first. | |
| 1307 | + // | |
| 1308 | + // The blob is matched as a balanced JSON object, not as "up to the | |
| 1309 | + // next quote". Oxygen wraps it in single quotes but does not escape | |
| 1310 | + // an apostrophe inside it (`"nicename":"Bob's Plumbing"`), so the | |
| 1311 | + // quote-to-quote match stopped mid-value and the rest of the blob, | |
| 1312 | + // `s Plumbing"}'` and all, was left in the tag and leaked into the | |
| 1313 | + // text. Strings inside the object are skipped whole, so neither a quote | |
| 1314 | + // nor a brace inside a value can end the match early. | |
| 1315 | + $source = (string) preg_replace( | |
| 1316 | + '/\sct_options\s*=\s*\'(?<obj>\{(?:[^{}"]++|"(?:[^"\\\\]|\\\\.)*+"|(?&obj))*+\})\'/s', | |
| 1317 | + '', | |
| 1318 | + $stored | |
| 1319 | + ); | |
| 1320 | + | |
| 1321 | + // Anything not shaped like Oxygen's JSON blob keeps the old, | |
| 1322 | + // quote-delimited strip. | |
| 1323 | + $source = (string) preg_replace( | |
| 1324 | + '/\sct_options\s*=\s*(["\']).*?\1/s', | |
| 1325 | + '', | |
| 1326 | + $source | |
| 1327 | + ); | |
| 1328 | + | |
| 1329 | + $attributes = implode('|', array_map( | |
| 1330 | + static fn(string $name): string => preg_quote($name, '/'), | |
| 1331 | + self::SHORTCODE_TEXT_ATTRIBUTES | |
| 1332 | + )); | |
| 1333 | + | |
| 1334 | + // Replace each shortcode tag with whatever readable copy its attributes | |
| 1335 | + // carry. Text between tags is left exactly where it is, so the result | |
| 1336 | + // keeps the page's reading order rather than hoisting all the headings | |
| 1337 | + // to the front. | |
| 1338 | + // The attribute blob is matched quote-aware rather than as "anything up | |
| 1339 | + // to the first `]`". Oxygen copy contains brackets often enough to | |
| 1340 | + // matter — "Best tools [2026]", "[Updated] our policy" — and a naive | |
| 1341 | + // scan ends the tag inside the `text` attribute, dropping the copy | |
| 1342 | + // before the bracket and leaking the stray `"]` after it into the | |
| 1343 | + // prose. Which is this bug's own failure mode: the wrong text scored. | |
| 1344 | + // | |
| 1345 | + // A tag name must start with a letter or underscore. `[2026]` is not a | |
| 1346 | + // shortcode anyone can register, and scanning it as one dropped the | |
| 1347 | + // year out of "Best tools [2026]". | |
| 1348 | + $text = (string) preg_replace_callback( | |
| 1349 | + '/\[\/?[a-zA-Z_][a-zA-Z0-9_-]*((?:[^\]"\']|"[^"]*"|\'[^\']*\')*)\]/', | |
| 1350 | + static function (array $matches) use ($attributes): string { | |
| 1351 | + if ('' === trim($matches[1])) { | |
| 1352 | + return ' '; | |
| 1353 | + } | |
| 1354 | + | |
| 1355 | + if (!preg_match_all( | |
| 1356 | + '/\b(' . $attributes . ')\s*=\s*(["\'])(.*?)\2/s', | |
| 1357 | + $matches[1], | |
| 1358 | + $found, | |
| 1359 | + PREG_SET_ORDER | |
| 1360 | + )) { | |
| 1361 | + return ' '; | |
| 1362 | + } | |
| 1363 | + | |
| 1364 | + $parts = []; | |
| 1365 | + foreach ($found as $attribute) { | |
| 1366 | + $value = trim($attribute[3]); | |
| 1367 | + | |
| 1368 | + // An attribute holding markup or a JSON fragment is | |
| 1369 | + // configuration that happens to share a name with a copy | |
| 1370 | + // field, not something a visitor reads. | |
| 1371 | + if ('' === $value || preg_match('/^[\[{<]/', $value)) { | |
| 1372 | + continue; | |
| 1373 | + } | |
| 1374 | + | |
| 1375 | + $parts[] = $value; | |
| 1376 | + } | |
| 1377 | + | |
| 1378 | + return empty($parts) ? ' ' : ' ' . implode(' ', $parts) . ' '; | |
| 1379 | + }, | |
| 1380 | + $source | |
| 1381 | + ); | |
| 1382 | + | |
| 1383 | + // Oxygen escapes square brackets in an element's copy before writing | |
| 1384 | + // it between the tags, so that "Best tools [2026]" cannot be mistaken | |
| 1385 | + // for a shortcode (`oxygen_vsb_filter_shortcode_content_encode()`). | |
| 1386 | + // Decoded only now, after the tag scan, for the same reason; left | |
| 1387 | + // encoded, the placeholders were scored as words of their own. | |
| 1388 | + $text = str_replace( | |
| 1389 | + ['_OXY_OPENING_BRACKET_', '_OXY_CLOSING_BRACKET_'], | |
| 1390 | + ['[', ']'], | |
| 1391 | + $text | |
| 1392 | + ); | |
| 1393 | + | |
| 1394 | + // Entities are stored encoded in attributes (&, ’), and would | |
| 1395 | + // otherwise be counted as words. | |
| 1396 | + $text = html_entity_decode($text, ENT_QUOTES | ENT_HTML5, 'UTF-8'); | |
| 1397 | + | |
| 1398 | + return trim((string) preg_replace('/\s+/u', ' ', $text)); | |
| 1017 | 1399 | } |
| 1018 | 1400 | |
| 1019 | 1401 | /** |
| 1020 | 1402 | * A node's children, whether it stores them as an array or an object. |
