| @@ -40,8 +40,36 @@ | ||
| 40 | 40 | */ |
| 41 | 41 | class Site_Identity_Manager extends Abstract_SEO_Manager { |
| 42 | 42 | |
| 43 | 43 | /** |
| 44 | + * The per-context title formats as ThinkRank ships them. | |
| 45 | + * | |
| 46 | + * These are not in get_default_settings(): the admin screen seeds them on | |
| 47 | + * first save, so on a real install they are stored values, indistinguishable | |
| 48 | + * from a template the user typed. The migration needs to tell those two | |
| 49 | + * apart — it may overwrite a shipped default with an imported template, and | |
| 50 | + * must never overwrite a choice the user made — so this is the record of | |
| 51 | + * what "untouched" looks like. | |
| 52 | + * | |
| 53 | + * Keep in step with getDefaultSettings() in | |
| 54 | + * src/admin/components/essential-seo/SiteIdentityTab.js. SiteIdentityTitleFormatDefaultsTest | |
| 55 | + * fails when the two drift. | |
| 56 | + * | |
| 57 | + * @since 2.8.0 | |
| 58 | + * @var array<string, string> | |
| 59 | + */ | |
| 60 | + public const TITLE_FORMAT_DEFAULTS = [ | |
| 61 | + 'homepage_title' => '%site_title% %sep% %site_description%', | |
| 62 | + 'post_title' => '%post_title% %sep% %site_title%', | |
| 63 | + 'page_title' => '%page_title% %sep% %site_title%', | |
| 64 | + 'category_title' => '%category_title% %sep% %site_title%', | |
| 65 | + 'tag_title' => '%tag_title% %sep% %site_title%', | |
| 66 | + 'author_title' => '%author_name% %sep% %site_title%', | |
| 67 | + 'search_title' => 'Search Results for "%search_term%" %sep% %site_title%', | |
| 68 | + 'archive_title' => '%archive_title% %sep% %site_title%', | |
| 69 | + ]; | |
| 70 | + | |
| 71 | + /** | |
| 44 | 72 | * WordPress filesystem instance |
| 45 | 73 | * |
| 46 | 74 | * @since 1.0.0 |
| 47 | 75 | * @var \WP_Filesystem_Base|null |
| @@ -303,8 +331,91 @@ | ||
| 303 | 331 | } |
| 304 | 332 | } |
| 305 | 333 | |
| 306 | 334 | /** |
| 335 | + * Save settings, then refresh what a new canonical scheme invalidates. | |
| 336 | + * | |
| 337 | + * The static sitemap files are written with the scheme in force when they | |
| 338 | + * were built, and nothing else rebuilds them until a post or term changes. | |
| 339 | + * So a change of scheme left every `<loc>` on the old one while canonical | |
| 340 | + * and og:url had already moved (#736). Every writer (the settings route, | |
| 341 | + * the robots route, the MCP abilities, an import) lands here. | |
| 342 | + * | |
| 343 | + * @since 2.7.0 | |
| 344 | + * | |
| 345 | + * @param string $context_type Context type. | |
| 346 | + * @param int|null $context_id Context ID. | |
| 347 | + * @param array $settings Settings to save. | |
| 348 | + * @return bool | |
| 349 | + */ | |
| 350 | + public function save_settings(string $context_type, ?int $context_id, array $settings): bool { | |
| 351 | + if (!self::touches_canonical_scheme($context_type, $context_id, $settings)) { | |
| 352 | + return parent::save_settings($context_type, $context_id, $settings); | |
| 353 | + } | |
| 354 | + | |
| 355 | + $before = Url_Scheme::preference(); | |
| 356 | + $saved = parent::save_settings($context_type, $context_id, $settings); | |
| 357 | + | |
| 358 | + if ($saved) { | |
| 359 | + $this->on_canonical_scheme_saved($before); | |
| 360 | + } | |
| 361 | + | |
| 362 | + return $saved; | |
| 363 | + } | |
| 364 | + | |
| 365 | + /** | |
| 366 | + * Whether a save can change the site-wide canonical scheme. | |
| 367 | + * | |
| 368 | + * @since 2.7.0 | |
| 369 | + * | |
| 370 | + * @param string $context_type Context type. | |
| 371 | + * @param int|null $context_id Context ID. | |
| 372 | + * @param array $settings Settings being saved. | |
| 373 | + * @return bool | |
| 374 | + */ | |
| 375 | + public static function touches_canonical_scheme(string $context_type, ?int $context_id, array $settings): bool { | |
| 376 | + return 'site' === sanitize_key($context_type) | |
| 377 | + && empty($context_id) | |
| 378 | + && array_key_exists('canonical_scheme', $settings); | |
| 379 | + } | |
| 380 | + | |
| 381 | + /** | |
| 382 | + * Rebuild the static sitemaps when the effective scheme changed. | |
| 383 | + * | |
| 384 | + * Compares the effective preference, filter included, so a site whose | |
| 385 | + * scheme is pinned by `thinkrank_canonical_scheme` does not rebuild on a | |
| 386 | + * stored value that changes nothing it publishes. | |
| 387 | + * | |
| 388 | + * @since 2.7.0 | |
| 389 | + * | |
| 390 | + * @param string $before Effective scheme before the save. | |
| 391 | + * @return void | |
| 392 | + */ | |
| 393 | + protected function on_canonical_scheme_saved(string $before): void { | |
| 394 | + // The preference is cached for the request; the save just changed it. | |
| 395 | + Url_Scheme::reset(); | |
| 396 | + | |
| 397 | + if (Url_Scheme::preference() === $before) { | |
| 398 | + return; | |
| 399 | + } | |
| 400 | + | |
| 401 | + $this->schedule_sitemap_rebuild(); | |
| 402 | + } | |
| 403 | + | |
| 404 | + /** | |
| 405 | + * Queue a settings-driven sitemap rebuild. | |
| 406 | + * | |
| 407 | + * Debounced and run after the response, like any other settings change | |
| 408 | + * that alters what the sitemap publishes. | |
| 409 | + * | |
| 410 | + * @since 2.7.0 | |
| 411 | + * @return void | |
| 412 | + */ | |
| 413 | + protected function schedule_sitemap_rebuild(): void { | |
| 414 | + (new Sitemap_Generator(false))->schedule_regeneration(); | |
| 415 | + } | |
| 416 | + | |
| 417 | + /** | |
| 307 | 418 | * Build the icon derivatives for a newly chosen favicon. |
| 308 | 419 | * |
| 309 | 420 | * Runs on save, which is the only moment the choice changes and the only |
| 310 | 421 | * place image work belongs — resolving a size on the front end must stay a |
| @@ -709,8 +820,17 @@ | ||
| 709 | 820 | $body = ($custom !== '' && !$fully_blocked) |
| 710 | 821 | ? $this->strip_robots_header($custom) |
| 711 | 822 | : trim($this->generate_robots_txt()['content']); |
| 712 | 823 | |
| 824 | + // The per-agent AI directives are machine-owned, so they are composed | |
| 825 | + // here rather than stored: the textarea holds the user's body, with | |
| 826 | + // the fenced block stripped out of every read and re-applied on every | |
| 827 | + // render. A site-wide block already disallows everyone, so adding the | |
| 828 | + // per-agent group there would be noise restating the same refusal. | |
| 829 | + if (!$fully_blocked) { | |
| 830 | + $body = $this->apply_ai_crawler_block($body, $settings); | |
| 831 | + } | |
| 832 | + | |
| 713 | 833 | if ($body === '') { |
| 714 | 834 | return ''; |
| 715 | 835 | } |
| 716 | 836 | |
| @@ -786,9 +906,12 @@ | ||
| 786 | 906 | $settings = $this->get_settings('site'); |
| 787 | 907 | |
| 788 | 908 | // Compare bodies, not raw strings: the auto-generated header carries a |
| 789 | 909 | // regeneration timestamp that always differs and means nothing here. |
| 790 | - $served = $this->strip_robots_header($effective['content']); | |
| 910 | + // The AI crawler block is composed at render time on both sides, so it | |
| 911 | + // is identical by construction and comparing it would only ever report | |
| 912 | + // a false drift the admin cannot act on. | |
| 913 | + $served = $this->strip_ai_crawler_block($this->strip_robots_header($effective['content'])); | |
| 791 | 914 | |
| 792 | 915 | // Measure against the body the editor is displaying — get_served_robots_body() |
| 793 | 916 | // — not against render_robots_txt(). Two things made the old comparison |
| 794 | 917 | // report "in sync" while the screen showed rules no crawler receives: |
| @@ -1166,17 +1289,19 @@ | ||
| 1166 | 1289 | $home_text = $settings['breadcrumb_home_text'] ?? 'Home'; |
| 1167 | 1290 | if (empty($home_text)) { |
| 1168 | 1291 | $optimization['warnings'][] = 'Empty home text reduces accessibility for screen readers'; |
| 1169 | 1292 | $optimization['score'] -= 15; |
| 1170 | - } elseif (strlen($home_text) > 20) { | |
| 1171 | - $optimization['suggestions'][] = 'Keep home text concise (current: ' . strlen($home_text) . ' chars)'; | |
| 1293 | + } elseif (mb_strlen($home_text) > 20) { | |
| 1294 | + // mb_strlen: this number is shown to the user as "chars" (#687). | |
| 1295 | + $optimization['suggestions'][] = 'Keep home text concise (current: ' . mb_strlen($home_text) . ' chars)'; | |
| 1172 | 1296 | $optimization['score'] -= 5; |
| 1173 | 1297 | } |
| 1174 | 1298 | |
| 1175 | 1299 | // Check prefix usage |
| 1176 | 1300 | $prefix = $settings['breadcrumb_prefix'] ?? ''; |
| 1177 | - if (!empty($prefix) && strlen($prefix) > 50) { | |
| 1178 | - $optimization['suggestions'][] = 'Breadcrumb prefix is quite long (' . strlen($prefix) . ' chars) - consider shortening'; | |
| 1301 | + if (!empty($prefix) && mb_strlen($prefix) > 50) { | |
| 1302 | + // mb_strlen: this number is shown to the user as "chars" (#687). | |
| 1303 | + $optimization['suggestions'][] = 'Breadcrumb prefix is quite long (' . mb_strlen($prefix) . ' chars) - consider shortening'; | |
| 1179 | 1304 | $optimization['score'] -= 5; |
| 1180 | 1305 | } |
| 1181 | 1306 | |
| 1182 | 1307 | // Current page display |
| @@ -2747,8 +2872,47 @@ | ||
| 2747 | 2872 | * @since 2.0.1 |
| 2748 | 2873 | * |
| 2749 | 2874 | * @return string[] |
| 2750 | 2875 | */ |
| 2876 | + /** | |
| 2877 | + * The stored alternate name(s), shaped for schema output. | |
| 2878 | + * | |
| 2879 | + * schema.org and Google both allow `alternateName` to carry one value or | |
| 2880 | + * several, and the store already round-trips either shape, so this accepts | |
| 2881 | + * both and normalises: null when there is nothing to publish, a bare string | |
| 2882 | + * for one name, a list for more. Emitting a one-element array would be | |
| 2883 | + * valid but noisier than it needs to be. | |
| 2884 | + * | |
| 2885 | + * Shared because both WebSite producers need it and must agree — a property | |
| 2886 | + * added to one and not the other is how #688 happened. | |
| 2887 | + * | |
| 2888 | + * @since 2.7.0 | |
| 2889 | + * | |
| 2890 | + * @param mixed $value Stored alternate_name value. | |
| 2891 | + * @return string|string[]|null | |
| 2892 | + */ | |
| 2893 | + public static function alternate_name_for_schema($value) { | |
| 2894 | + $names = []; | |
| 2895 | + | |
| 2896 | + foreach ((array) $value as $name) { | |
| 2897 | + if (!is_scalar($name)) { | |
| 2898 | + continue; | |
| 2899 | + } | |
| 2900 | + | |
| 2901 | + $name = trim((string) $name); | |
| 2902 | + | |
| 2903 | + if ('' !== $name && !in_array($name, $names, true)) { | |
| 2904 | + $names[] = $name; | |
| 2905 | + } | |
| 2906 | + } | |
| 2907 | + | |
| 2908 | + if (empty($names)) { | |
| 2909 | + return null; | |
| 2910 | + } | |
| 2911 | + | |
| 2912 | + return 1 === count($names) ? $names[0] : $names; | |
| 2913 | + } | |
| 2914 | + | |
| 2751 | 2915 | protected function additional_setting_keys(): array { |
| 2752 | 2916 | return [ |
| 2753 | 2917 | // Title formats, one per context. |
| 2754 | 2918 | 'homepage_title', 'post_title', 'page_title', 'category_title', |
| @@ -2762,8 +2926,10 @@ | ||
| 2762 | 2926 | // Schema toggles that live on this screen. |
| 2763 | 2927 | 'organization_schema', 'knowledge_graph', |
| 2764 | 2928 | // Robots rules composed by the Robots.txt panel. |
| 2765 | 2929 | 'custom_robots_rules', |
| 2930 | + // Per-agent AI crawler allow/block map (#657). | |
| 2931 | + 'ai_crawler_rules', | |
| 2766 | 2932 | // Hero section. |
| 2767 | 2933 | 'hero_title', 'hero_subtitle', 'hero_cta_text', 'hero_cta_url', |
| 2768 | 2934 | 'hero_background_image', |
| 2769 | 2935 | // Local SEO / business details. |
| @@ -2775,8 +2941,44 @@ | ||
| 2775 | 2941 | ]; |
| 2776 | 2942 | } |
| 2777 | 2943 | |
| 2778 | 2944 | /** |
| 2945 | + * Sanitize settings, normalising the AI crawler rule map. | |
| 2946 | + * | |
| 2947 | + * The generic array sanitizer keeps the shape but says nothing about the | |
| 2948 | + * values: a payload could store `ai_crawler_rules[gptbot] = "maybe"`, or a | |
| 2949 | + * slug no crawler answers to, and both would round-trip through every | |
| 2950 | + * later response. Normalising here rather than in the REST handler puts it | |
| 2951 | + * on the one path every writer shares — the settings route, the robots | |
| 2952 | + * route and the MCP abilities all land in save_settings() (#657). | |
| 2953 | + * | |
| 2954 | + * @since 2.5.0 | |
| 2955 | + * | |
| 2956 | + * @param array $settings Settings to sanitize. | |
| 2957 | + * @param string $context_type Context type. | |
| 2958 | + * @return array Sanitized settings. | |
| 2959 | + */ | |
| 2960 | + protected function sanitize_settings(array $settings, string $context_type = 'site'): array { | |
| 2961 | + $sanitized = parent::sanitize_settings($settings, $context_type); | |
| 2962 | + | |
| 2963 | + if (array_key_exists('ai_crawler_rules', $sanitized)) { | |
| 2964 | + $sanitized['ai_crawler_rules'] = AI_Crawlers::normalize_rules($sanitized['ai_crawler_rules']); | |
| 2965 | + } | |
| 2966 | + | |
| 2967 | + // Same reasoning one key up, for the scheme override (#638). Anything | |
| 2968 | + // that is not one of the three modes means "follow WordPress", and is | |
| 2969 | + // stored as that rather than kept verbatim — otherwise get-site-identity | |
| 2970 | + // -settings would report a scheme the site does not actually publish. | |
| 2971 | + if (array_key_exists('canonical_scheme', $sanitized)) { | |
| 2972 | + $sanitized['canonical_scheme'] = in_array($sanitized['canonical_scheme'], Url_Scheme::MODES, true) | |
| 2973 | + ? $sanitized['canonical_scheme'] | |
| 2974 | + : Url_Scheme::AUTOMATIC; | |
| 2975 | + } | |
| 2976 | + | |
| 2977 | + return $sanitized; | |
| 2978 | + } | |
| 2979 | + | |
| 2980 | + /** | |
| 2779 | 2981 | * Get default settings for a context type (implements interface) |
| 2780 | 2982 | * |
| 2781 | 2983 | * @since 1.0.0 |
| 2782 | 2984 | * |
| @@ -2796,9 +2998,32 @@ | ||
| 2796 | 2998 | 'breadcrumb_home_text' => 'Home', |
| 2797 | 2999 | 'breadcrumb_separator' => '>', |
| 2798 | 3000 | 'robots_txt_enabled' => true, |
| 2799 | 3001 | 'allow_search_engines' => true, |
| 3002 | + // Answer 404 when a content selector in the URL resolved to | |
| 3003 | + // nothing (#634). On by default, unlike the other new settings | |
| 3004 | + // here: it changes no URL a visitor or a correct crawler uses, only | |
| 3005 | + // ones where WordPress resolved nothing and served the blog listing | |
| 3006 | + // at 200 anyway. | |
| 3007 | + 'query_protection' => true, | |
| 3008 | + | |
| 3009 | + // Feed controls (#635). All three off, so an upgrade changes | |
| 3010 | + // nothing about what an existing site already sends its | |
| 3011 | + // subscribers; a brand-new install is seeded with the signature and | |
| 3012 | + // the noindex on, in Activator::seed_feed_defaults(). | |
| 3013 | + 'feed_excerpt_only' => false, | |
| 3014 | + 'feed_source_link' => false, | |
| 3015 | + 'feed_noindex' => false, | |
| 3016 | + | |
| 3017 | + // The scheme self-referential URLs go out with (#638). 'automatic' | |
| 3018 | + // means substitute nothing and follow WordPress, which is what | |
| 3019 | + // every site did before the setting existed. | |
| 3020 | + 'canonical_scheme' => Url_Scheme::AUTOMATIC, | |
| 2800 | 3021 | 'robots_txt_content' => '', |
| 3022 | + // Empty map = every AI crawler allowed. Defaults must stay | |
| 3023 | + // permissive so an upgrade never starts blocking a crawler a site | |
| 3024 | + // was happily serving (#657). | |
| 3025 | + 'ai_crawler_rules' => [], | |
| 2801 | 3026 | 'logo_url' => '', |
| 2802 | 3027 | 'favicon_url' => '', |
| 2803 | 3028 | 'apple_touch_icon_url' => '' |
| 2804 | 3029 | ]; |
| @@ -3007,16 +3232,17 @@ | ||
| 3007 | 3232 | // Remove extra whitespace |
| 3008 | 3233 | $title = preg_replace('/\s+/', ' ', $title); |
| 3009 | 3234 | $title = trim($title); |
| 3010 | 3235 | |
| 3011 | - // Ensure title is not too long (60 characters max for SEO) | |
| 3012 | - if (strlen($title) > 60) { | |
| 3013 | - // Try to truncate at word boundary | |
| 3014 | - $title = wp_trim_words($title, 8, '...'); | |
| 3015 | - if (strlen($title) > 60) { | |
| 3016 | - $title = substr($title, 0, 57) . '...'; | |
| 3017 | - } | |
| 3018 | - } | |
| 3236 | + // Ensure title is not too long (60 characters max for SEO). | |
| 3237 | + // All three units here were wrong for non-Latin text: strlen() counts | |
| 3238 | + // BYTES so the gate fired at 20 Thai characters, wp_trim_words() counts | |
| 3239 | + // CHARACTERS on th/ja/zh_* so `8` cut the title to 8 of them, and | |
| 3240 | + // substr() cuts bytes so it split a character mid-sequence (#687). | |
| 3241 | + $title = \ThinkRank\Core\Seo_Text::trim_to_length( | |
| 3242 | + $title, | |
| 3243 | + \ThinkRank\Core\Seo_Text::TITLE_MAX_LENGTH | |
| 3244 | + ); | |
| 3019 | 3245 | |
| 3020 | 3246 | // Ensure title is not empty |
| 3021 | 3247 | if (empty($title)) { |
| 3022 | 3248 | $title = get_bloginfo('name'); |
| @@ -3406,8 +3632,29 @@ | ||
| 3406 | 3632 | * @since 1.0.0 |
| 3407 | 3633 | * @return array Array of sitemap URLs |
| 3408 | 3634 | */ |
| 3409 | 3635 | private function get_sitemap_urls_for_robots(): array { |
| 3636 | + // One wrapper over every return path below, including the #104 extras. | |
| 3637 | + // The Sitemap: line is the only absolute URL of ours in robots.txt and | |
| 3638 | + // the one a crawler follows to find everything else, so it has to carry | |
| 3639 | + // the site's scheme preference (#638). Applied here rather than where | |
| 3640 | + // the body is assembled, because that path also renders a robots.txt a | |
| 3641 | + // site owner typed themselves, and their text is not ours to rewrite. | |
| 3642 | + return array_map( | |
| 3643 | + static function (string $url): string { | |
| 3644 | + return Url_Scheme::apply($url); | |
| 3645 | + }, | |
| 3646 | + $this->collect_sitemap_urls_for_robots() | |
| 3647 | + ); | |
| 3648 | + } | |
| 3649 | + | |
| 3650 | + /** | |
| 3651 | + * The sitemap URLs robots.txt advertises, before the scheme preference. | |
| 3652 | + * | |
| 3653 | + * @since 1.0.0 | |
| 3654 | + * @return array Array of sitemap URLs | |
| 3655 | + */ | |
| 3656 | + private function collect_sitemap_urls_for_robots(): array { | |
| 3410 | 3657 | try { |
| 3411 | 3658 | // Get sitemap settings |
| 3412 | 3659 | $sitemap_generator = new \ThinkRank\SEO\Sitemap_Generator(); |
| 3413 | 3660 | $sitemap_settings = $sitemap_generator->get_settings('site'); |
| @@ -3456,9 +3703,22 @@ | ||
| 3456 | 3703 | // business sitemap and the sitemaps other plugins register both land |
| 3457 | 3704 | // here for the same reason, so they go through one list (#104). |
| 3458 | 3705 | $extra = []; |
| 3459 | 3706 | |
| 3460 | - if (file_exists(ABSPATH . 'local-sitemap.xml')) { | |
| 3707 | + // Not a file test. Under dynamic delivery the local sitemap is | |
| 3708 | + // served from PHP and no file is ever written, so file_exists() | |
| 3709 | + // silently dropped a sitemap the site really does publish (#752). | |
| 3710 | + // On static sites the file is still what proves it, so both count. | |
| 3711 | + $local_sitemap_published = file_exists(ABSPATH . 'local-sitemap.xml'); | |
| 3712 | + | |
| 3713 | + if (!$local_sitemap_published && class_exists('ThinkRank\\SEO\\Sitemap_Generator')) { | |
| 3714 | + $generator = new \ThinkRank\SEO\Sitemap_Generator(false); | |
| 3715 | + | |
| 3716 | + $local_sitemap_published = 'dynamic' === $generator->resolve_delivery_mode() | |
| 3717 | + && $generator->publishes_local_sitemap(); | |
| 3718 | + } | |
| 3719 | + | |
| 3720 | + if ($local_sitemap_published) { | |
| 3461 | 3721 | $extra[] = '/local-sitemap.xml'; |
| 3462 | 3722 | } |
| 3463 | 3723 | |
| 3464 | 3724 | foreach (\ThinkRank\SEO\Sitemap_Generator::additional_sitemaps() as $path) { |
| @@ -3654,8 +3914,116 @@ | ||
| 3654 | 3914 | return $rules; |
| 3655 | 3915 | } |
| 3656 | 3916 | |
| 3657 | 3917 | /** |
| 3918 | + * Opening fence of the machine-owned AI crawler region. | |
| 3919 | + * | |
| 3920 | + * @since 2.5.0 | |
| 3921 | + * @var string | |
| 3922 | + */ | |
| 3923 | + public const AI_BLOCK_BEGIN = '# BEGIN ThinkRank AI crawlers'; | |
| 3924 | + | |
| 3925 | + /** | |
| 3926 | + * Closing fence of the machine-owned AI crawler region. | |
| 3927 | + * | |
| 3928 | + * @since 2.5.0 | |
| 3929 | + * @var string | |
| 3930 | + */ | |
| 3931 | + public const AI_BLOCK_END = '# END ThinkRank AI crawlers'; | |
| 3932 | + | |
| 3933 | + /** | |
| 3934 | + * Render the fenced AI crawler region for the current settings. | |
| 3935 | + * | |
| 3936 | + * One `User-agent:` / `Disallow: /` record per blocked crawler. Allowed | |
| 3937 | + * crawlers emit nothing at all: `Disallow:` with an empty value is the | |
| 3938 | + * robots.txt way of saying "allow everything", but writing eighteen such | |
| 3939 | + * records to say what silence already says would triple the file and | |
| 3940 | + * invite the reading that an unlisted crawler is therefore refused. | |
| 3941 | + * | |
| 3942 | + * @since 2.5.0 | |
| 3943 | + * | |
| 3944 | + * @param array $settings Site settings. | |
| 3945 | + * @return string Fenced block, newline-terminated, or '' when nothing is blocked. | |
| 3946 | + */ | |
| 3947 | + private function build_ai_crawler_block(array $settings): string { | |
| 3948 | + $blocked = AI_Crawlers::blocked_slugs($settings['ai_crawler_rules'] ?? []); | |
| 3949 | + | |
| 3950 | + if (empty($blocked)) { | |
| 3951 | + return ''; | |
| 3952 | + } | |
| 3953 | + | |
| 3954 | + $agents = AI_Crawlers::all(); | |
| 3955 | + | |
| 3956 | + $lines = [ | |
| 3957 | + self::AI_BLOCK_BEGIN, | |
| 3958 | + '# Managed by ThinkRank — edits between these lines are overwritten.', | |
| 3959 | + ]; | |
| 3960 | + | |
| 3961 | + foreach ($blocked as $slug) { | |
| 3962 | + $lines[] = ''; | |
| 3963 | + $lines[] = 'User-agent: ' . $agents[$slug]['token']; | |
| 3964 | + $lines[] = 'Disallow: /'; | |
| 3965 | + } | |
| 3966 | + | |
| 3967 | + $lines[] = self::AI_BLOCK_END; | |
| 3968 | + | |
| 3969 | + return implode("\n", $lines) . "\n"; | |
| 3970 | + } | |
| 3971 | + | |
| 3972 | + /** | |
| 3973 | + * Remove the fenced AI crawler region from a robots.txt body. | |
| 3974 | + * | |
| 3975 | + * Tolerates a missing closing fence rather than leaving the rest of the | |
| 3976 | + * file swallowed: a truncated write, or someone deleting the END line by | |
| 3977 | + * hand, would otherwise make every subsequent read drop everything below | |
| 3978 | + * the opening fence. | |
| 3979 | + * | |
| 3980 | + * @since 2.5.0 | |
| 3981 | + * | |
| 3982 | + * @param string $body Robots.txt body. | |
| 3983 | + * @return string Body with the region removed. | |
| 3984 | + */ | |
| 3985 | + public function strip_ai_crawler_block(string $body): string { | |
| 3986 | + if (false === strpos($body, self::AI_BLOCK_BEGIN)) { | |
| 3987 | + return $body; | |
| 3988 | + } | |
| 3989 | + | |
| 3990 | + $pattern = '/\R*' . preg_quote(self::AI_BLOCK_BEGIN, '/') | |
| 3991 | + . '.*?(?:' . preg_quote(self::AI_BLOCK_END, '/') . '|\z)\R*/s'; | |
| 3992 | + | |
| 3993 | + return trim((string) preg_replace($pattern, "\n\n", $body, 1)); | |
| 3994 | + } | |
| 3995 | + | |
| 3996 | + /** | |
| 3997 | + * Put the current AI crawler region into a robots.txt body. | |
| 3998 | + * | |
| 3999 | + * Replaces an existing region in place so the block keeps its position in | |
| 4000 | + * a hand-ordered file, and appends when there is none. Everything outside | |
| 4001 | + * the fences is returned untouched — that is the whole point of fencing | |
| 4002 | + * it, since the body is also a free-text field the user edits. | |
| 4003 | + * | |
| 4004 | + * @since 2.5.0 | |
| 4005 | + * | |
| 4006 | + * @param string $body Robots.txt body (fences optional). | |
| 4007 | + * @param array $settings Site settings. | |
| 4008 | + * @return string Body carrying the current region. | |
| 4009 | + */ | |
| 4010 | + private function apply_ai_crawler_block(string $body, array $settings): string { | |
| 4011 | + $stripped = $this->strip_ai_crawler_block($body); | |
| 4012 | + $block = $this->build_ai_crawler_block($settings); | |
| 4013 | + | |
| 4014 | + if ('' === $block) { | |
| 4015 | + return $stripped; | |
| 4016 | + } | |
| 4017 | + | |
| 4018 | + if ('' === trim($stripped)) { | |
| 4019 | + return trim($block); | |
| 4020 | + } | |
| 4021 | + | |
| 4022 | + return rtrim($stripped) . "\n\n" . trim($block); | |
| 4023 | + } | |
| 4024 | + | |
| 4025 | + /** | |
| 3658 | 4026 | * The auto-generated header prepended to the served robots.txt. |
| 3659 | 4027 | * |
| 3660 | 4028 | * Kept separate from the body so it is only ever added at render time with |
| 3661 | 4029 | * a fresh timestamp, never stored or shown in the editable textarea. |
| @@ -3692,11 +4060,16 @@ | ||
| 3692 | 4060 | */ |
| 3693 | 4061 | public function get_served_robots_body(): string { |
| 3694 | 4062 | $settings = $this->get_settings('site'); |
| 3695 | 4063 | |
| 4064 | + // The AI block is stripped from every one of these paths. A physical | |
| 4065 | + // robots.txt we wrote carries it, and the stored override is whatever | |
| 4066 | + // the textarea last held — so without this the block round-trips into | |
| 4067 | + // the editor, gets saved as ordinary body text, and is then appended | |
| 4068 | + // to a second time on the next render. | |
| 3696 | 4069 | $custom = trim((string) ($settings['robots_txt_content'] ?? '')); |
| 3697 | 4070 | if ($custom !== '') { |
| 3698 | - return $this->strip_robots_header($custom); | |
| 4071 | + return $this->strip_ai_crawler_block($this->strip_robots_header($custom)); | |
| 3699 | 4072 | } |
| 3700 | 4073 | |
| 3701 | 4074 | $robots_file = ABSPATH . 'robots.txt'; |
| 3702 | 4075 | if (file_exists($robots_file)) { |
| @@ -3702,13 +4075,13 @@ | ||
| 3702 | 4075 | if (file_exists($robots_file)) { |
| 3703 | 4076 | // phpcs:ignore WordPress.WP.AlternativeFunctions.file_get_contents_file_get_contents, WordPress.PHP.NoSilencedErrors.Discouraged -- an unreadable robots.txt is an expected state answered with an empty string. |
| 3704 | 4077 | $raw = (string) @file_get_contents($robots_file); |
| 3705 | 4078 | if ($raw !== '') { |
| 3706 | - return $this->strip_robots_header($raw); | |
| 4079 | + return $this->strip_ai_crawler_block($this->strip_robots_header($raw)); | |
| 3707 | 4080 | } |
| 3708 | 4081 | } |
| 3709 | 4082 | |
| 3710 | - return trim($this->generate_robots_txt()['content']); | |
| 4083 | + return $this->strip_ai_crawler_block(trim($this->generate_robots_txt()['content'])); | |
| 3711 | 4084 | } |
| 3712 | 4085 | private function get_site_identity_data(array $settings): array { |
| 3713 | 4086 | return [ |
| 3714 | 4087 | 'site_name' => $settings['site_name'] ?? get_bloginfo('name'), |
| @@ -3770,11 +4143,14 @@ | ||
| 3770 | 4143 | $optimization['validation']['valid'] = false; |
| 3771 | 4144 | } |
| 3772 | 4145 | |
| 3773 | 4146 | if (!empty($value) && isset($config['max_length'])) { |
| 3774 | - if (strlen($value) > $config['max_length']) { | |
| 4147 | + // The warning says "characters", so measure and cut in characters: | |
| 4148 | + // strlen()/substr() fired early on non-Latin values and the | |
| 4149 | + // suggested replacement was cut mid-character (#687). | |
| 4150 | + if (mb_strlen($value) > $config['max_length']) { | |
| 3775 | 4151 | $optimization['validation']['warnings'][] = "{$element} exceeds maximum length of {$config['max_length']} characters"; |
| 3776 | - $optimization['optimized_value'] = substr($value, 0, $config['max_length']); | |
| 4152 | + $optimization['optimized_value'] = \ThinkRank\Core\Seo_Text::trim_to_length($value, (int) $config['max_length']); | |
| 3777 | 4153 | } |
| 3778 | 4154 | } |
| 3779 | 4155 | |
| 3780 | 4156 | // SEO-specific optimizations |
