| @@ -15,8 +15,13 @@ | ||
| 15 | 15 | declare(strict_types=1); |
| 16 | 16 | |
| 17 | 17 | namespace ThinkRank\SEO; |
| 18 | 18 | |
| 19 | +// Prevent direct access | |
| 20 | +if (!defined('ABSPATH')) { | |
| 21 | + exit; | |
| 22 | +} | |
| 23 | + | |
| 19 | 24 | // Ensure dependencies are loaded |
| 20 | 25 | if (!class_exists('ThinkRank\\SEO\\Abstract_SEO_Manager')) { |
| 21 | 26 | require_once THINKRANK_PLUGIN_DIR . 'includes/seo/class-abstract-seo-manager.php'; |
| 22 | 27 | } |
| @@ -35,8 +40,36 @@ | ||
| 35 | 40 | */ |
| 36 | 41 | class Site_Identity_Manager extends Abstract_SEO_Manager { |
| 37 | 42 | |
| 38 | 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 | + /** | |
| 39 | 72 | * WordPress filesystem instance |
| 40 | 73 | * |
| 41 | 74 | * @since 1.0.0 |
| 42 | 75 | * @var \WP_Filesystem_Base|null |
| @@ -240,13 +273,358 @@ | ||
| 240 | 273 | * Constructor |
| 241 | 274 | * |
| 242 | 275 | * @since 1.0.0 |
| 243 | 276 | */ |
| 277 | + /** | |
| 278 | + * The square derivatives wp_site_icon() asks for. | |
| 279 | + * | |
| 280 | + * Core generates these only through its own Site Icon crop flow, so an | |
| 281 | + * image chosen as a ThinkRank favicon straight from the media library has | |
| 282 | + * none of them and every sizes="" declaration is a near miss (#571). | |
| 283 | + * | |
| 284 | + * @since 2.3.1 | |
| 285 | + * @var int[] | |
| 286 | + */ | |
| 287 | + public const ICON_SIZES = [32, 180, 192, 270]; | |
| 288 | + | |
| 289 | + /** | |
| 290 | + * Transient holding resolved icon URLs, keyed by configured URL and size. | |
| 291 | + * | |
| 292 | + * The site-icon filter runs in wp_head on every FRONT-END request, and | |
| 293 | + * resolving a URL to its attachment costs an uncached postmeta query. The | |
| 294 | + * mapping only changes when the icon setting does, so it is cached here and | |
| 295 | + * dropped on save. | |
| 296 | + * | |
| 297 | + * @since 2.3.1 | |
| 298 | + * @var string | |
| 299 | + */ | |
| 300 | + public const ICON_URL_TRANSIENT = 'thinkrank_site_icon_urls'; | |
| 301 | + | |
| 302 | + /** | |
| 303 | + * Marker for the one-time derivative backfill on existing installs. | |
| 304 | + * | |
| 305 | + * @since 2.3.1 | |
| 306 | + * @var string | |
| 307 | + */ | |
| 308 | + public const ICON_BACKFILL_OPTION = 'thinkrank_site_icon_sizes_backfilled'; | |
| 309 | + | |
| 310 | + /** | |
| 311 | + * Whether the icon-derivative listener has been registered this request. | |
| 312 | + * | |
| 313 | + * Static because `thinkrank_seo_settings_saved` is a global hook — one | |
| 314 | + * listener serves every instance, and this class is constructed on the | |
| 315 | + * front end as well as in admin. | |
| 316 | + * | |
| 317 | + * @since 2.3.1 | |
| 318 | + * @var bool | |
| 319 | + */ | |
| 320 | + private static bool $icon_sizes_listener_registered = false; | |
| 321 | + | |
| 244 | 322 | public function __construct() { |
| 245 | 323 | parent::__construct('site_identity'); |
| 324 | + | |
| 325 | + if (!self::$icon_sizes_listener_registered) { | |
| 326 | + self::$icon_sizes_listener_registered = true; | |
| 327 | + add_action('thinkrank_seo_settings_saved', [$this, 'generate_icon_sizes_on_save'], 10, 2); | |
| 328 | + // Admin only: resizing is not front-end work, and admin traffic is | |
| 329 | + // enough to run a one-time backfill promptly. | |
| 330 | + add_action('admin_init', [self::class, 'maybe_backfill_icon_sizes']); | |
| 331 | + } | |
| 246 | 332 | } |
| 247 | 333 | |
| 248 | 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 | + /** | |
| 418 | + * Build the icon derivatives for a newly chosen favicon. | |
| 419 | + * | |
| 420 | + * Runs on save, which is the only moment the choice changes and the only | |
| 421 | + * place image work belongs — resolving a size on the front end must stay a | |
| 422 | + * lookup. Failure is silent by design: a missing derivative degrades to the | |
| 423 | + * next best file, so a site whose host cannot resize still renders an icon. | |
| 424 | + * | |
| 425 | + * @since 2.3.1 | |
| 426 | + * | |
| 427 | + * @param string $manager_type Settings category that was saved. | |
| 428 | + * @param array $settings The settings that were written. | |
| 429 | + * @return void | |
| 430 | + */ | |
| 431 | + public function generate_icon_sizes_on_save(string $manager_type, array $settings): void { | |
| 432 | + if ('site_identity' !== $manager_type) { | |
| 433 | + return; | |
| 434 | + } | |
| 435 | + | |
| 436 | + // The choice, or the derivatives behind it, may have just changed. | |
| 437 | + delete_transient(self::ICON_URL_TRANSIENT); | |
| 438 | + | |
| 439 | + foreach (['favicon_url', 'apple_touch_icon_url'] as $key) { | |
| 440 | + if (empty($settings[$key]) || !is_string($settings[$key])) { | |
| 441 | + continue; | |
| 442 | + } | |
| 443 | + | |
| 444 | + $attachment_id = self::icon_attachment_id($settings[$key]); | |
| 445 | + | |
| 446 | + if ($attachment_id) { | |
| 447 | + self::ensure_icon_sizes($attachment_id); | |
| 448 | + } | |
| 449 | + } | |
| 450 | + | |
| 451 | + // Dropped again after the resizes finish. Resizing is not instant, and a | |
| 452 | + // front-end request arriving mid-generation would otherwise repopulate | |
| 453 | + // the transient with the pre-derivative URLs and pin them for the full | |
| 454 | + // TTL — leaving the sizes= declarations untrue until the next save. | |
| 455 | + delete_transient(self::ICON_URL_TRANSIENT); | |
| 456 | + } | |
| 457 | + | |
| 458 | + /** | |
| 459 | + * Build the derivatives for a site that configured its icons before this | |
| 460 | + * existed. | |
| 461 | + * | |
| 462 | + * generate_icon_sizes_on_save() only fires on a settings write, so every | |
| 463 | + * site with an icon already chosen would keep serving whatever | |
| 464 | + * wp_get_attachment_image_url() could find — in practice the 150x150 | |
| 465 | + * thumbnail behind a sizes="32x32" declaration — until someone happened to | |
| 466 | + * re-save Site Identity. That is the bug this is meant to fix, so the | |
| 467 | + * derivatives are built once on upgrade instead of waiting for a save. | |
| 468 | + * | |
| 469 | + * Guarded by its own option rather than the plugin version so it runs once | |
| 470 | + * and stays cheap: the check is a single autoloaded read on requests after | |
| 471 | + * the first. | |
| 472 | + * | |
| 473 | + * @since 2.3.1 | |
| 474 | + * | |
| 475 | + * @return void | |
| 476 | + */ | |
| 477 | + public static function maybe_backfill_icon_sizes(): void { | |
| 478 | + if (get_option(self::ICON_BACKFILL_OPTION)) { | |
| 479 | + return; | |
| 480 | + } | |
| 481 | + | |
| 482 | + // Written before the work, not after: a host that cannot resize must | |
| 483 | + // not retry on every admin request forever. | |
| 484 | + update_option(self::ICON_BACKFILL_OPTION, time(), true); | |
| 485 | + | |
| 486 | + $settings = (new self())->get_settings('site'); | |
| 487 | + | |
| 488 | + if (!is_array($settings)) { | |
| 489 | + return; | |
| 490 | + } | |
| 491 | + | |
| 492 | + foreach (['favicon_url', 'apple_touch_icon_url'] as $key) { | |
| 493 | + if (empty($settings[$key]) || !is_string($settings[$key])) { | |
| 494 | + continue; | |
| 495 | + } | |
| 496 | + | |
| 497 | + $attachment_id = self::icon_attachment_id($settings[$key]); | |
| 498 | + | |
| 499 | + if ($attachment_id) { | |
| 500 | + self::ensure_icon_sizes($attachment_id); | |
| 501 | + } | |
| 502 | + } | |
| 503 | + | |
| 504 | + delete_transient(self::ICON_URL_TRANSIENT); | |
| 505 | + } | |
| 506 | + | |
| 507 | + /** | |
| 508 | + * Attachment ID behind a configured icon URL, or 0 when it is not ours. | |
| 509 | + * | |
| 510 | + * attachment_url_to_postid() matches _wp_attached_file, which holds the | |
| 511 | + * ORIGINAL upload path, so the URL of a generated derivative | |
| 512 | + * (`logo-512.png`) returns 0 — and that is exactly what the media picker | |
| 513 | + * hands back when the user chooses a size. Strip the dimension suffix and | |
| 514 | + * try the original once. | |
| 515 | + * | |
| 516 | + * Shared with SEO_Manager's site-icon filter so both sides of the feature | |
| 517 | + * agree on which attachment a configured URL means. | |
| 518 | + * | |
| 519 | + * @since 2.3.1 | |
| 520 | + * | |
| 521 | + * @param string $url Configured icon URL. | |
| 522 | + * @return int Attachment ID, or 0. | |
| 523 | + */ | |
| 524 | + public static function icon_attachment_id(string $url): int { | |
| 525 | + $attachment_id = (int) attachment_url_to_postid($url); | |
| 526 | + | |
| 527 | + if ($attachment_id) { | |
| 528 | + return $attachment_id; | |
| 529 | + } | |
| 530 | + | |
| 531 | + $original = preg_replace('/-\d+x\d+(?=\.[a-zA-Z0-9]+$)/', '', $url); | |
| 532 | + | |
| 533 | + if (is_string($original) && $original !== $url) { | |
| 534 | + return (int) attachment_url_to_postid($original); | |
| 535 | + } | |
| 536 | + | |
| 537 | + return 0; | |
| 538 | + } | |
| 539 | + | |
| 540 | + /** | |
| 541 | + * Which ICON_SIZES derivatives this attachment still needs. | |
| 542 | + * | |
| 543 | + * Split out from the generation so the decision can be asserted on its | |
| 544 | + * own: whether a size is skipped because it already exists or because it | |
| 545 | + * would upscale is invisible once both answers are "nothing was built". | |
| 546 | + * | |
| 547 | + * A source is measured by its SHORTER edge — a 400x40 banner cannot yield | |
| 548 | + * a true 192x192 — and anything reporting no dimensions at all (SVGs) is | |
| 549 | + * left alone. | |
| 550 | + * | |
| 551 | + * @since 2.3.1 | |
| 552 | + * | |
| 553 | + * @param array $meta Attachment metadata. | |
| 554 | + * @return array<string, array{width: int, height: int, crop: bool}> Sizes to build. | |
| 555 | + */ | |
| 556 | + public static function missing_icon_sizes(array $meta): array { | |
| 557 | + $source = min((int) ($meta['width'] ?? 0), (int) ($meta['height'] ?? 0)); | |
| 558 | + | |
| 559 | + if ($source < 1) { | |
| 560 | + return []; | |
| 561 | + } | |
| 562 | + | |
| 563 | + $wanted = []; | |
| 564 | + foreach (self::ICON_SIZES as $size) { | |
| 565 | + // Never upscale: a stretched source behind an accurate sizes="" | |
| 566 | + // label is worse than the honest near miss it would replace. | |
| 567 | + if (isset($meta['sizes']["site_icon-{$size}"]) || $size > $source) { | |
| 568 | + continue; | |
| 569 | + } | |
| 570 | + | |
| 571 | + $wanted["site_icon-{$size}"] = ['width' => $size, 'height' => $size, 'crop' => true]; | |
| 572 | + } | |
| 573 | + | |
| 574 | + return $wanted; | |
| 575 | + } | |
| 576 | + | |
| 577 | + /** | |
| 578 | + * Generate whatever ICON_SIZES derivatives this attachment is missing. | |
| 579 | + * | |
| 580 | + * Only the missing ones, and never one larger than the source: upscaling a | |
| 581 | + * small favicon would put a blurrier file behind an accurate sizes="" label | |
| 582 | + * than the honest near-miss it replaced. | |
| 583 | + * | |
| 584 | + * @since 2.3.1 | |
| 585 | + * | |
| 586 | + * @param int $attachment_id Attachment to build derivatives for. | |
| 587 | + * @return string[] Size names generated, empty when there was nothing to do. | |
| 588 | + */ | |
| 589 | + public static function ensure_icon_sizes(int $attachment_id): array { | |
| 590 | + $meta = wp_get_attachment_metadata($attachment_id); | |
| 591 | + | |
| 592 | + if (!is_array($meta)) { | |
| 593 | + return []; | |
| 594 | + } | |
| 595 | + | |
| 596 | + $wanted = self::missing_icon_sizes($meta); | |
| 597 | + | |
| 598 | + if (empty($wanted)) { | |
| 599 | + return []; | |
| 600 | + } | |
| 601 | + | |
| 602 | + $file = get_attached_file($attachment_id); | |
| 603 | + | |
| 604 | + if (!$file || !file_exists($file)) { | |
| 605 | + return []; | |
| 606 | + } | |
| 607 | + | |
| 608 | + $editor = wp_get_image_editor($file); | |
| 609 | + | |
| 610 | + if (is_wp_error($editor)) { | |
| 611 | + return []; | |
| 612 | + } | |
| 613 | + | |
| 614 | + $generated = $editor->multi_resize($wanted); | |
| 615 | + | |
| 616 | + if (empty($generated)) { | |
| 617 | + return []; | |
| 618 | + } | |
| 619 | + | |
| 620 | + $meta['sizes'] = array_merge($meta['sizes'] ?? [], $generated); | |
| 621 | + wp_update_attachment_metadata($attachment_id, $meta); | |
| 622 | + | |
| 623 | + return array_keys($generated); | |
| 624 | + } | |
| 625 | + | |
| 626 | + /** | |
| 249 | 627 | * Initialize WordPress filesystem |
| 250 | 628 | * |
| 251 | 629 | * @since 1.0.0 |
| 252 | 630 | * @return bool True if filesystem is initialized, false otherwise |
| @@ -442,8 +820,17 @@ | ||
| 442 | 820 | $body = ($custom !== '' && !$fully_blocked) |
| 443 | 821 | ? $this->strip_robots_header($custom) |
| 444 | 822 | : trim($this->generate_robots_txt()['content']); |
| 445 | 823 | |
| 824 | + // The per-agent AI directives are machine-owned, so they are composed | |
| 825 | + // here rather than stored: the textarea holds the user's body, with | |
| 826 | + // the fenced block stripped out of every read and re-applied on every | |
| 827 | + // render. A site-wide block already disallows everyone, so adding the | |
| 828 | + // per-agent group there would be noise restating the same refusal. | |
| 829 | + if (!$fully_blocked) { | |
| 830 | + $body = $this->apply_ai_crawler_block($body, $settings); | |
| 831 | + } | |
| 832 | + | |
| 446 | 833 | if ($body === '') { |
| 447 | 834 | return ''; |
| 448 | 835 | } |
| 449 | 836 | |
| @@ -507,27 +894,53 @@ | ||
| 507 | 894 | * ever sees — the conflict this exists to surface. |
| 508 | 895 | * |
| 509 | 896 | * @since 1.31.0 |
| 510 | 897 | * |
| 511 | - * @return array{content: string, source: string, is_default: bool, in_sync: bool, url: string} | |
| 898 | + * @return array{content: string, source: string, is_default: bool, in_sync: bool, out_of_sync_reason: string, url: string} | |
| 512 | 899 | * The served content and its origin, whether it still reflects the |
| 513 | - * saved settings, and the public URL it is served from. | |
| 900 | + * body the editor is showing, why it does not when it does not | |
| 901 | + * ('file_drift' or 'crawl_blocked'), and the public URL it is | |
| 902 | + * served from. | |
| 514 | 903 | */ |
| 515 | 904 | public function get_robots_txt_delivery(): array { |
| 516 | 905 | $effective = $this->get_effective_robots_txt(); |
| 906 | + $settings = $this->get_settings('site'); | |
| 517 | 907 | |
| 518 | 908 | // Compare bodies, not raw strings: the auto-generated header carries a |
| 519 | 909 | // regeneration timestamp that always differs and means nothing here. |
| 520 | - $served = $this->strip_robots_header($effective['content']); | |
| 521 | - $expected = $this->strip_robots_header($this->render_robots_txt()); | |
| 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'])); | |
| 522 | 914 | |
| 915 | + // Measure against the body the editor is displaying — get_served_robots_body() | |
| 916 | + // — not against render_robots_txt(). Two things made the old comparison | |
| 917 | + // report "in sync" while the screen showed rules no crawler receives: | |
| 918 | + // a physical file was compared to a freshly rendered body rather than | |
| 919 | + // to the stored override the textarea shows, and a site-wide crawl | |
| 920 | + // block makes render_robots_txt() return the generated "Disallow: /" | |
| 921 | + // on both sides of the comparison, so it always matched. | |
| 922 | + $expected = $this->get_served_robots_body(); | |
| 923 | + | |
| 924 | + // Management off: WordPress serves its own default and the editor is not | |
| 925 | + // claiming anything is live, so there is nothing to be out of sync with. | |
| 926 | + $managed = !empty($settings['robots_txt_enabled']); | |
| 927 | + $in_sync = !$managed || $served === $expected; | |
| 928 | + | |
| 929 | + $reason = ''; | |
| 930 | + if (!$in_sync) { | |
| 931 | + // A crawl block is a deliberate override, not a stale file, and the | |
| 932 | + // admin needs to be told which of the two they are looking at. | |
| 933 | + $blocked = empty($settings['allow_search_engines'] ?? true) || !get_option('blog_public'); | |
| 934 | + $reason = $blocked ? 'crawl_blocked' : 'file_drift'; | |
| 935 | + } | |
| 936 | + | |
| 523 | 937 | return [ |
| 524 | 938 | 'content' => $effective['content'], |
| 525 | 939 | 'source' => $effective['source'], |
| 526 | 940 | 'is_default' => $effective['is_default'], |
| 527 | - // Only a physical file can drift. Every other source is rendered | |
| 528 | - // from the settings on demand, so it is in sync by construction. | |
| 529 | - 'in_sync' => $effective['source'] !== 'file' || $served === $expected, | |
| 941 | + 'in_sync' => $in_sync, | |
| 942 | + 'out_of_sync_reason' => $reason, | |
| 530 | 943 | 'url' => home_url('/robots.txt'), |
| 531 | 944 | ]; |
| 532 | 945 | } |
| 533 | 946 | |
| @@ -876,17 +1289,19 @@ | ||
| 876 | 1289 | $home_text = $settings['breadcrumb_home_text'] ?? 'Home'; |
| 877 | 1290 | if (empty($home_text)) { |
| 878 | 1291 | $optimization['warnings'][] = 'Empty home text reduces accessibility for screen readers'; |
| 879 | 1292 | $optimization['score'] -= 15; |
| 880 | - } elseif (strlen($home_text) > 20) { | |
| 881 | - $optimization['suggestions'][] = 'Keep home text concise (current: ' . strlen($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)'; | |
| 882 | 1296 | $optimization['score'] -= 5; |
| 883 | 1297 | } |
| 884 | 1298 | |
| 885 | 1299 | // Check prefix usage |
| 886 | 1300 | $prefix = $settings['breadcrumb_prefix'] ?? ''; |
| 887 | - if (!empty($prefix) && strlen($prefix) > 50) { | |
| 888 | - $optimization['suggestions'][] = 'Breadcrumb prefix is quite long (' . strlen($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'; | |
| 889 | 1304 | $optimization['score'] -= 5; |
| 890 | 1305 | } |
| 891 | 1306 | |
| 892 | 1307 | // Current page display |
| @@ -2296,12 +2711,15 @@ | ||
| 2296 | 2711 | 'warnings' => [], |
| 2297 | 2712 | 'suggestions' => [] |
| 2298 | 2713 | ]; |
| 2299 | 2714 | |
| 2300 | - // Validate business name (required for local SEO) | |
| 2715 | + // Business name is what makes the LocalBusiness schema useful, but it | |
| 2716 | + // cannot be a blocking error: the toggle is what reveals the business | |
| 2717 | + // fields, so requiring the name up front makes enabling Local SEO | |
| 2718 | + // impossible. The frontend already skips the output while the name is | |
| 2719 | + // empty (see Seo_Manager::output_local_seo_meta_tags()). | |
| 2301 | 2720 | if (empty($settings['business_name'])) { |
| 2302 | - $validation['errors'][] = 'Business name is required when local SEO is enabled'; | |
| 2303 | - $validation['valid'] = false; | |
| 2721 | + $validation['warnings'][] = 'Business name is missing - required before local business schema is output'; | |
| 2304 | 2722 | } elseif (strlen($settings['business_name']) > 100) { |
| 2305 | 2723 | $validation['warnings'][] = 'Business name is very long, consider shortening for better display'; |
| 2306 | 2724 | } |
| 2307 | 2725 | |
| @@ -2442,8 +2860,125 @@ | ||
| 2442 | 2860 | return $output; |
| 2443 | 2861 | } |
| 2444 | 2862 | |
| 2445 | 2863 | /** |
| 2864 | + * Keys the Site Identity screens store beyond the 16 defaults. | |
| 2865 | + * | |
| 2866 | + * Title formats, breadcrumb configuration, the hero fields, the business | |
| 2867 | + * block and the wizard's identity fields are all real settings written by | |
| 2868 | + * this manager, none of which get_default_settings() names — it seeds only | |
| 2869 | + * the values a fresh install needs. Gating on defaults alone would stop | |
| 2870 | + * every one of them saving (#452). | |
| 2871 | + * | |
| 2872 | + * @since 2.0.1 | |
| 2873 | + * | |
| 2874 | + * @return string[] | |
| 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 | + | |
| 2915 | + protected function additional_setting_keys(): array { | |
| 2916 | + return [ | |
| 2917 | + // Title formats, one per context. | |
| 2918 | + 'homepage_title', 'post_title', 'page_title', 'category_title', | |
| 2919 | + 'tag_title', 'author_title', 'search_title', 'archive_title', | |
| 2920 | + // Breadcrumbs. | |
| 2921 | + 'breadcrumb_prefix', 'show_current_page', 'breadcrumb_use_seo_title', | |
| 2922 | + // Identity, as written by the setup wizard and the importers. | |
| 2923 | + 'alternate_name', 'identity_type', 'represents', | |
| 2924 | + 'default_meta_description', 'default_social_image', | |
| 2925 | + 'social_media_accounts', | |
| 2926 | + // Schema toggles that live on this screen. | |
| 2927 | + 'organization_schema', 'knowledge_graph', | |
| 2928 | + // Robots rules composed by the Robots.txt panel. | |
| 2929 | + 'custom_robots_rules', | |
| 2930 | + // Per-agent AI crawler allow/block map (#657). | |
| 2931 | + 'ai_crawler_rules', | |
| 2932 | + // Hero section. | |
| 2933 | + 'hero_title', 'hero_subtitle', 'hero_cta_text', 'hero_cta_url', | |
| 2934 | + 'hero_background_image', | |
| 2935 | + // Local SEO / business details. | |
| 2936 | + 'local_seo_enabled', 'business_type', 'business_name', | |
| 2937 | + 'business_address', 'business_city', 'business_state', | |
| 2938 | + 'business_postal_code', 'business_country', 'business_phone', | |
| 2939 | + 'business_email', 'business_latitude', 'business_longitude', | |
| 2940 | + 'business_price_range', 'business_hours', | |
| 2941 | + ]; | |
| 2942 | + } | |
| 2943 | + | |
| 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 | + /** | |
| 2446 | 2981 | * Get default settings for a context type (implements interface) |
| 2447 | 2982 | * |
| 2448 | 2983 | * @since 1.0.0 |
| 2449 | 2984 | * |
| @@ -2463,9 +2998,32 @@ | ||
| 2463 | 2998 | 'breadcrumb_home_text' => 'Home', |
| 2464 | 2999 | 'breadcrumb_separator' => '>', |
| 2465 | 3000 | 'robots_txt_enabled' => true, |
| 2466 | 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, | |
| 2467 | 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' => [], | |
| 2468 | 3026 | 'logo_url' => '', |
| 2469 | 3027 | 'favicon_url' => '', |
| 2470 | 3028 | 'apple_touch_icon_url' => '' |
| 2471 | 3029 | ]; |
| @@ -2584,10 +3142,13 @@ | ||
| 2584 | 3142 | */ |
| 2585 | 3143 | private function prepare_title_placeholders(array $data, string $context, array $settings): array { |
| 2586 | 3144 | $placeholders = [ |
| 2587 | 3145 | '%title%' => $data['title'] ?? '', |
| 2588 | - '%sitename%' => $settings['site_name'] ?? get_bloginfo('name'), | |
| 2589 | - '%tagline%' => $settings['tagline'] ?? get_bloginfo('description'), | |
| 3146 | + // `?:` rather than `??`: these are persisted as '' rather than left | |
| 3147 | + // unset, and '' is not null, so the null-coalesce never reached the | |
| 3148 | + // WordPress fallback (#398). | |
| 3149 | + '%sitename%' => ($settings['site_name'] ?? '') ?: get_bloginfo('name'), | |
| 3150 | + '%tagline%' => ($settings['tagline'] ?? '') ?: get_bloginfo('description'), | |
| 2590 | 3151 | '%separator%' => '', // Will be replaced with actual separator |
| 2591 | 3152 | '%category%' => '', |
| 2592 | 3153 | '%author%' => '', |
| 2593 | 3154 | '%date%' => '', |
| @@ -2671,16 +3232,17 @@ | ||
| 2671 | 3232 | // Remove extra whitespace |
| 2672 | 3233 | $title = preg_replace('/\s+/', ' ', $title); |
| 2673 | 3234 | $title = trim($title); |
| 2674 | 3235 | |
| 2675 | - // Ensure title is not too long (60 characters max for SEO) | |
| 2676 | - if (strlen($title) > 60) { | |
| 2677 | - // Try to truncate at word boundary | |
| 2678 | - $title = wp_trim_words($title, 8, '...'); | |
| 2679 | - if (strlen($title) > 60) { | |
| 2680 | - $title = substr($title, 0, 57) . '...'; | |
| 2681 | - } | |
| 2682 | - } | |
| 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 | + $title, | |
| 3243 | + \ThinkRank\Core\Seo_Text::TITLE_MAX_LENGTH | |
| 3244 | + ); | |
| 2683 | 3245 | |
| 2684 | 3246 | // Ensure title is not empty |
| 2685 | 3247 | if (empty($title)) { |
| 2686 | 3248 | $title = get_bloginfo('name'); |
| @@ -3070,8 +3632,29 @@ | ||
| 3070 | 3632 | * @since 1.0.0 |
| 3071 | 3633 | * @return array Array of sitemap URLs |
| 3072 | 3634 | */ |
| 3073 | 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 { | |
| 3074 | 3657 | try { |
| 3075 | 3658 | // Get sitemap settings |
| 3076 | 3659 | $sitemap_generator = new \ThinkRank\SEO\Sitemap_Generator(); |
| 3077 | 3660 | $sitemap_settings = $sitemap_generator->get_settings('site'); |
| @@ -3083,29 +3666,70 @@ | ||
| 3083 | 3666 | |
| 3084 | 3667 | $sitemap_urls = []; |
| 3085 | 3668 | $site_url = home_url(); |
| 3086 | 3669 | |
| 3087 | - // Extract enabled sitemap URLs | |
| 3670 | + // Extract enabled sitemap URLs. When the index is enabled it is the | |
| 3671 | + // only entry worth advertising: every child sitemap is already | |
| 3672 | + // listed inside it, so naming them again in robots.txt is pure | |
| 3673 | + // redundancy and drifts out of date as soon as a post type is added. | |
| 3674 | + $index_url = ''; | |
| 3088 | 3675 | if (!empty($sitemap_settings['sitemap_urls']) && is_array($sitemap_settings['sitemap_urls'])) { |
| 3089 | 3676 | foreach ($sitemap_settings['sitemap_urls'] as $sitemap) { |
| 3090 | - if (!empty($sitemap['enabled']) && !empty($sitemap['url'])) { | |
| 3091 | - $sitemap_urls[] = $site_url . $sitemap['url']; | |
| 3677 | + if (empty($sitemap['enabled']) || empty($sitemap['url'])) { | |
| 3678 | + continue; | |
| 3092 | 3679 | } |
| 3680 | + | |
| 3681 | + if (($sitemap['type'] ?? '') === 'index') { | |
| 3682 | + $index_url = $site_url . $sitemap['url']; | |
| 3683 | + continue; | |
| 3684 | + } | |
| 3685 | + | |
| 3686 | + $sitemap_urls[] = $site_url . $sitemap['url']; | |
| 3093 | 3687 | } |
| 3094 | 3688 | } |
| 3095 | 3689 | |
| 3690 | + if ($index_url !== '') { | |
| 3691 | + // The index alone — it covers the children and, on a segmented | |
| 3692 | + // install, the local business sitemap too. | |
| 3693 | + return [$index_url]; | |
| 3694 | + } | |
| 3695 | + | |
| 3096 | 3696 | // Fallback to default if no URLs found |
| 3097 | 3697 | if (empty($sitemap_urls)) { |
| 3098 | 3698 | $sitemap_urls[] = home_url('/sitemap.xml'); |
| 3099 | 3699 | } |
| 3100 | 3700 | |
| 3101 | - // Advertise the local business sitemap when it exists. In segmented | |
| 3102 | - // mode it is already listed inside the sitemap index; in single mode | |
| 3103 | - // there is no index, so robots.txt is its discovery path. | |
| 3104 | - if (file_exists(ABSPATH . 'local-sitemap.xml')) { | |
| 3105 | - $local_url = home_url('/local-sitemap.xml'); | |
| 3106 | - if (!in_array($local_url, $sitemap_urls, true)) { | |
| 3107 | - $sitemap_urls[] = $local_url; | |
| 3701 | + // No index on this install, so anything not already listed above has | |
| 3702 | + // no other discovery path — advertise it directly. The local | |
| 3703 | + // business sitemap and the sitemaps other plugins register both land | |
| 3704 | + // here for the same reason, so they go through one list (#104). | |
| 3705 | + $extra = []; | |
| 3706 | + | |
| 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) { | |
| 3721 | + $extra[] = '/local-sitemap.xml'; | |
| 3722 | + } | |
| 3723 | + | |
| 3724 | + foreach (\ThinkRank\SEO\Sitemap_Generator::additional_sitemaps() as $path) { | |
| 3725 | + $extra[] = $path; | |
| 3726 | + } | |
| 3727 | + | |
| 3728 | + foreach ($extra as $path) { | |
| 3729 | + $url = home_url($path); | |
| 3730 | + if (!in_array($url, $sitemap_urls, true)) { | |
| 3731 | + $sitemap_urls[] = $url; | |
| 3108 | 3732 | } |
| 3109 | 3733 | } |
| 3110 | 3734 | |
| 3111 | 3735 | return $sitemap_urls; |
| @@ -3195,8 +3819,9 @@ | ||
| 3195 | 3819 | // timestamp on every update. |
| 3196 | 3820 | $content = ''; |
| 3197 | 3821 | |
| 3198 | 3822 | $current_user_agent = ''; |
| 3823 | + $sitemap_started = false; | |
| 3199 | 3824 | |
| 3200 | 3825 | foreach ($rules as $rule) { |
| 3201 | 3826 | $directive = $rule['directive'] ?? ''; |
| 3202 | 3827 | $value = $rule['value'] ?? ''; |
| @@ -3217,9 +3842,17 @@ | ||
| 3217 | 3842 | case 'crawl_delay': |
| 3218 | 3843 | $content .= "Crawl-delay: {$value}\n"; |
| 3219 | 3844 | break; |
| 3220 | 3845 | case 'sitemap': |
| 3221 | - $content .= "\nSitemap: {$value}\n"; | |
| 3846 | + // One blank line separates the Sitemap block from the | |
| 3847 | + // preceding group, and none appear inside it. A blank line | |
| 3848 | + // terminates a record in the robots.txt grammar, so putting | |
| 3849 | + // one between every directive was invalid formatting. | |
| 3850 | + if (!$sitemap_started) { | |
| 3851 | + $content .= "\n"; | |
| 3852 | + $sitemap_started = true; | |
| 3853 | + } | |
| 3854 | + $content .= "Sitemap: {$value}\n"; | |
| 3222 | 3855 | break; |
| 3223 | 3856 | } |
| 3224 | 3857 | } |
| 3225 | 3858 | |
| @@ -3226,8 +3859,171 @@ | ||
| 3226 | 3859 | return ltrim($content, "\n"); |
| 3227 | 3860 | } |
| 3228 | 3861 | |
| 3229 | 3862 | /** |
| 3863 | + * Parse a robots.txt body back into the {directive, value} rule shape. | |
| 3864 | + * | |
| 3865 | + * generate_robots_txt() returns `rules` alongside `content`, but callers | |
| 3866 | + * replace `content` with the body actually being served (a stored override | |
| 3867 | + * or a physical file). The generated rules then described something the | |
| 3868 | + * response no longer contained. Re-deriving them from the served body keeps | |
| 3869 | + * the two halves of the payload describing the same document. | |
| 3870 | + * | |
| 3871 | + * @since 2.0.1 | |
| 3872 | + * | |
| 3873 | + * @param string $content Robots.txt body (header optional). | |
| 3874 | + * @return array<int, array{directive: string, value: string}> Parsed rules. | |
| 3875 | + */ | |
| 3876 | + public function parse_robots_txt_rules(string $content): array { | |
| 3877 | + $map = [ | |
| 3878 | + 'user-agent' => 'user_agent', | |
| 3879 | + 'disallow' => 'disallow', | |
| 3880 | + 'allow' => 'allow', | |
| 3881 | + 'crawl-delay' => 'crawl_delay', | |
| 3882 | + 'sitemap' => 'sitemap', | |
| 3883 | + ]; | |
| 3884 | + | |
| 3885 | + $rules = []; | |
| 3886 | + | |
| 3887 | + foreach (preg_split('/\r\n|\r|\n/', $this->strip_robots_header($content)) as $line) { | |
| 3888 | + $line = trim($line); | |
| 3889 | + | |
| 3890 | + // Blank lines separate groups and `#` starts a comment; neither is | |
| 3891 | + // a rule. | |
| 3892 | + if ($line === '' || str_starts_with($line, '#')) { | |
| 3893 | + continue; | |
| 3894 | + } | |
| 3895 | + | |
| 3896 | + $parts = explode(':', $line, 2); | |
| 3897 | + if (count($parts) !== 2) { | |
| 3898 | + continue; | |
| 3899 | + } | |
| 3900 | + | |
| 3901 | + $field = strtolower(trim($parts[0])); | |
| 3902 | + if (!isset($map[$field])) { | |
| 3903 | + continue; | |
| 3904 | + } | |
| 3905 | + | |
| 3906 | + $rules[] = [ | |
| 3907 | + 'directive' => $map[$field], | |
| 3908 | + // Sitemap values are absolute URLs and contain the `:` the | |
| 3909 | + // limited explode above deliberately preserved. | |
| 3910 | + 'value' => trim($parts[1]), | |
| 3911 | + ]; | |
| 3912 | + } | |
| 3913 | + | |
| 3914 | + return $rules; | |
| 3915 | + } | |
| 3916 | + | |
| 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 | + /** | |
| 3230 | 4026 | * The auto-generated header prepended to the served robots.txt. |
| 3231 | 4027 | * |
| 3232 | 4028 | * Kept separate from the body so it is only ever added at render time with |
| 3233 | 4029 | * a fresh timestamp, never stored or shown in the editable textarea. |
| @@ -3264,11 +4060,16 @@ | ||
| 3264 | 4060 | */ |
| 3265 | 4061 | public function get_served_robots_body(): string { |
| 3266 | 4062 | $settings = $this->get_settings('site'); |
| 3267 | 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. | |
| 3268 | 4069 | $custom = trim((string) ($settings['robots_txt_content'] ?? '')); |
| 3269 | 4070 | if ($custom !== '') { |
| 3270 | - return $this->strip_robots_header($custom); | |
| 4071 | + return $this->strip_ai_crawler_block($this->strip_robots_header($custom)); | |
| 3271 | 4072 | } |
| 3272 | 4073 | |
| 3273 | 4074 | $robots_file = ABSPATH . 'robots.txt'; |
| 3274 | 4075 | if (file_exists($robots_file)) { |
| @@ -3274,13 +4075,13 @@ | ||
| 3274 | 4075 | if (file_exists($robots_file)) { |
| 3275 | 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. |
| 3276 | 4077 | $raw = (string) @file_get_contents($robots_file); |
| 3277 | 4078 | if ($raw !== '') { |
| 3278 | - return $this->strip_robots_header($raw); | |
| 4079 | + return $this->strip_ai_crawler_block($this->strip_robots_header($raw)); | |
| 3279 | 4080 | } |
| 3280 | 4081 | } |
| 3281 | 4082 | |
| 3282 | - return trim($this->generate_robots_txt()['content']); | |
| 4083 | + return $this->strip_ai_crawler_block(trim($this->generate_robots_txt()['content'])); | |
| 3283 | 4084 | } |
| 3284 | 4085 | private function get_site_identity_data(array $settings): array { |
| 3285 | 4086 | return [ |
| 3286 | 4087 | 'site_name' => $settings['site_name'] ?? get_bloginfo('name'), |
| @@ -3342,11 +4143,14 @@ | ||
| 3342 | 4143 | $optimization['validation']['valid'] = false; |
| 3343 | 4144 | } |
| 3344 | 4145 | |
| 3345 | 4146 | if (!empty($value) && isset($config['max_length'])) { |
| 3346 | - if (strlen($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']) { | |
| 3347 | 4151 | $optimization['validation']['warnings'][] = "{$element} exceeds maximum length of {$config['max_length']} characters"; |
| 3348 | - $optimization['optimized_value'] = substr($value, 0, $config['max_length']); | |
| 4152 | + $optimization['optimized_value'] = \ThinkRank\Core\Seo_Text::trim_to_length($value, (int) $config['max_length']); | |
| 3349 | 4153 | } |
| 3350 | 4154 | } |
| 3351 | 4155 | |
| 3352 | 4156 | // SEO-specific optimizations |
