| @@ -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 +273,424 @@ | ||
| 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 | + | |
| 322 | + /** | |
| 323 | + * Whether the robots.txt resync listener is registered for this request. | |
| 324 | + * | |
| 325 | + * @since 2.14.0 | |
| 326 | + * @var bool | |
| 327 | + */ | |
| 328 | + private static bool $robots_sync_listener_registered = false; | |
| 329 | + | |
| 330 | + /** | |
| 331 | + * Flag set when a plugin change may have altered the sitemap set. | |
| 332 | + * | |
| 333 | + * @since 2.14.0 | |
| 334 | + * @var string | |
| 335 | + */ | |
| 336 | + public const ROBOTS_RESYNC_OPTION = 'thinkrank_robots_txt_resync_pending'; | |
| 337 | + | |
| 244 | 338 | public function __construct() { |
| 245 | 339 | parent::__construct('site_identity'); |
| 340 | + | |
| 341 | + if (!self::$icon_sizes_listener_registered) { | |
| 342 | + self::$icon_sizes_listener_registered = true; | |
| 343 | + add_action('thinkrank_seo_settings_saved', [$this, 'generate_icon_sizes_on_save'], 10, 2); | |
| 344 | + // Admin only: resizing is not front-end work, and admin traffic is | |
| 345 | + // enough to run a one-time backfill promptly. | |
| 346 | + add_action('admin_init', [self::class, 'maybe_backfill_icon_sizes']); | |
| 347 | + } | |
| 348 | + | |
| 349 | + if (!self::$robots_sync_listener_registered) { | |
| 350 | + self::$robots_sync_listener_registered = true; | |
| 351 | + | |
| 352 | + // A physical robots.txt bypasses PHP entirely, so composing the | |
| 353 | + // Sitemap block at render time fixes the served output only on | |
| 354 | + // sites with no file. Activating or deactivating a sitemap | |
| 355 | + // contributor changes the set, and until #835 nothing rewrote the | |
| 356 | + // file: the deactivated plugin's sitemap stayed advertised, serving | |
| 357 | + // HTML to anything that followed it. | |
| 358 | + add_action('activated_plugin', [self::class, 'flag_robots_txt_resync']); | |
| 359 | + add_action('deactivated_plugin', [self::class, 'flag_robots_txt_resync']); | |
| 360 | + add_action('init', [self::class, 'maybe_resync_robots_txt'], 99); | |
| 361 | + } | |
| 246 | 362 | } |
| 247 | 363 | |
| 248 | 364 | /** |
| 365 | + * Note that the set of sitemap contributors may have changed. | |
| 366 | + * | |
| 367 | + * Deliberately unconditional about which plugin: a contributor is anything | |
| 368 | + * hooking `thinkrank_additional_sitemaps`, which is resolved at runtime and | |
| 369 | + * cannot be inspected for a plugin that is on its way out. | |
| 370 | + * | |
| 371 | + * The rewrite is not done here. `deactivated_plugin` fires inside the | |
| 372 | + * request that deactivated it, while that plugin's filters are still | |
| 373 | + * attached, so rendering now still sees the sitemap that is going away — | |
| 374 | + * measured, not assumed: the first version of this fix wrote the | |
| 375 | + * deactivated plugin's sitemap straight back into the file. The next | |
| 376 | + * request has the real plugin set loaded, so the work waits for it. | |
| 377 | + * | |
| 378 | + * @since 2.14.0 | |
| 379 | + * @return void | |
| 380 | + */ | |
| 381 | + public static function flag_robots_txt_resync(): void { | |
| 382 | + if (!file_exists(ABSPATH . 'robots.txt')) { | |
| 383 | + return; | |
| 384 | + } | |
| 385 | + | |
| 386 | + update_option(self::ROBOTS_RESYNC_OPTION, 1, false); | |
| 387 | + } | |
| 388 | + | |
| 389 | + /** | |
| 390 | + * Rewrite the physical robots.txt once, on the request after a change. | |
| 391 | + * | |
| 392 | + * @since 2.14.0 | |
| 393 | + * @return void | |
| 394 | + */ | |
| 395 | + public static function maybe_resync_robots_txt(): void { | |
| 396 | + if (!get_option(self::ROBOTS_RESYNC_OPTION)) { | |
| 397 | + return; | |
| 398 | + } | |
| 399 | + | |
| 400 | + // Cleared first, so a render that fatals cannot retry on every request | |
| 401 | + // for the rest of the site's life. | |
| 402 | + delete_option(self::ROBOTS_RESYNC_OPTION); | |
| 403 | + | |
| 404 | + if (!file_exists(ABSPATH . 'robots.txt')) { | |
| 405 | + return; | |
| 406 | + } | |
| 407 | + | |
| 408 | + (new self())->sync_robots_txt_file(); | |
| 409 | + } | |
| 410 | + | |
| 411 | + /** | |
| 412 | + * Save settings, then refresh what a new canonical scheme invalidates. | |
| 413 | + * | |
| 414 | + * The static sitemap files are written with the scheme in force when they | |
| 415 | + * were built, and nothing else rebuilds them until a post or term changes. | |
| 416 | + * So a change of scheme left every `<loc>` on the old one while canonical | |
| 417 | + * and og:url had already moved (#736). Every writer (the settings route, | |
| 418 | + * the robots route, the MCP abilities, an import) lands here. | |
| 419 | + * | |
| 420 | + * @since 2.7.0 | |
| 421 | + * | |
| 422 | + * @param string $context_type Context type. | |
| 423 | + * @param int|null $context_id Context ID. | |
| 424 | + * @param array $settings Settings to save. | |
| 425 | + * @return bool | |
| 426 | + */ | |
| 427 | + public function save_settings(string $context_type, ?int $context_id, array $settings): bool { | |
| 428 | + if (!self::touches_canonical_scheme($context_type, $context_id, $settings)) { | |
| 429 | + return parent::save_settings($context_type, $context_id, $settings); | |
| 430 | + } | |
| 431 | + | |
| 432 | + $before = Url_Scheme::preference(); | |
| 433 | + $saved = parent::save_settings($context_type, $context_id, $settings); | |
| 434 | + | |
| 435 | + if ($saved) { | |
| 436 | + $this->on_canonical_scheme_saved($before); | |
| 437 | + } | |
| 438 | + | |
| 439 | + return $saved; | |
| 440 | + } | |
| 441 | + | |
| 442 | + /** | |
| 443 | + * Whether a save can change the site-wide canonical scheme. | |
| 444 | + * | |
| 445 | + * @since 2.7.0 | |
| 446 | + * | |
| 447 | + * @param string $context_type Context type. | |
| 448 | + * @param int|null $context_id Context ID. | |
| 449 | + * @param array $settings Settings being saved. | |
| 450 | + * @return bool | |
| 451 | + */ | |
| 452 | + public static function touches_canonical_scheme(string $context_type, ?int $context_id, array $settings): bool { | |
| 453 | + return 'site' === sanitize_key($context_type) | |
| 454 | + && empty($context_id) | |
| 455 | + && array_key_exists('canonical_scheme', $settings); | |
| 456 | + } | |
| 457 | + | |
| 458 | + /** | |
| 459 | + * Rebuild the static sitemaps when the effective scheme changed. | |
| 460 | + * | |
| 461 | + * Compares the effective preference, filter included, so a site whose | |
| 462 | + * scheme is pinned by `thinkrank_canonical_scheme` does not rebuild on a | |
| 463 | + * stored value that changes nothing it publishes. | |
| 464 | + * | |
| 465 | + * @since 2.7.0 | |
| 466 | + * | |
| 467 | + * @param string $before Effective scheme before the save. | |
| 468 | + * @return void | |
| 469 | + */ | |
| 470 | + protected function on_canonical_scheme_saved(string $before): void { | |
| 471 | + // The preference is cached for the request; the save just changed it. | |
| 472 | + Url_Scheme::reset(); | |
| 473 | + | |
| 474 | + if (Url_Scheme::preference() === $before) { | |
| 475 | + return; | |
| 476 | + } | |
| 477 | + | |
| 478 | + $this->schedule_sitemap_rebuild(); | |
| 479 | + } | |
| 480 | + | |
| 481 | + /** | |
| 482 | + * Queue a settings-driven sitemap rebuild. | |
| 483 | + * | |
| 484 | + * Debounced and run after the response, like any other settings change | |
| 485 | + * that alters what the sitemap publishes. | |
| 486 | + * | |
| 487 | + * @since 2.7.0 | |
| 488 | + * @return void | |
| 489 | + */ | |
| 490 | + protected function schedule_sitemap_rebuild(): void { | |
| 491 | + (new Sitemap_Generator(false))->schedule_regeneration(); | |
| 492 | + } | |
| 493 | + | |
| 494 | + /** | |
| 495 | + * Build the icon derivatives for a newly chosen favicon. | |
| 496 | + * | |
| 497 | + * Runs on save, which is the only moment the choice changes and the only | |
| 498 | + * place image work belongs — resolving a size on the front end must stay a | |
| 499 | + * lookup. Failure is silent by design: a missing derivative degrades to the | |
| 500 | + * next best file, so a site whose host cannot resize still renders an icon. | |
| 501 | + * | |
| 502 | + * @since 2.3.1 | |
| 503 | + * | |
| 504 | + * @param string $manager_type Settings category that was saved. | |
| 505 | + * @param array $settings The settings that were written. | |
| 506 | + * @return void | |
| 507 | + */ | |
| 508 | + public function generate_icon_sizes_on_save(string $manager_type, array $settings): void { | |
| 509 | + if ('site_identity' !== $manager_type) { | |
| 510 | + return; | |
| 511 | + } | |
| 512 | + | |
| 513 | + // The choice, or the derivatives behind it, may have just changed. | |
| 514 | + delete_transient(self::ICON_URL_TRANSIENT); | |
| 515 | + | |
| 516 | + foreach (['favicon_url', 'apple_touch_icon_url'] as $key) { | |
| 517 | + if (empty($settings[$key]) || !is_string($settings[$key])) { | |
| 518 | + continue; | |
| 519 | + } | |
| 520 | + | |
| 521 | + $attachment_id = self::icon_attachment_id($settings[$key]); | |
| 522 | + | |
| 523 | + if ($attachment_id) { | |
| 524 | + self::ensure_icon_sizes($attachment_id); | |
| 525 | + } | |
| 526 | + } | |
| 527 | + | |
| 528 | + // Dropped again after the resizes finish. Resizing is not instant, and a | |
| 529 | + // front-end request arriving mid-generation would otherwise repopulate | |
| 530 | + // the transient with the pre-derivative URLs and pin them for the full | |
| 531 | + // TTL — leaving the sizes= declarations untrue until the next save. | |
| 532 | + delete_transient(self::ICON_URL_TRANSIENT); | |
| 533 | + } | |
| 534 | + | |
| 535 | + /** | |
| 536 | + * Build the derivatives for a site that configured its icons before this | |
| 537 | + * existed. | |
| 538 | + * | |
| 539 | + * generate_icon_sizes_on_save() only fires on a settings write, so every | |
| 540 | + * site with an icon already chosen would keep serving whatever | |
| 541 | + * wp_get_attachment_image_url() could find — in practice the 150x150 | |
| 542 | + * thumbnail behind a sizes="32x32" declaration — until someone happened to | |
| 543 | + * re-save Site Identity. That is the bug this is meant to fix, so the | |
| 544 | + * derivatives are built once on upgrade instead of waiting for a save. | |
| 545 | + * | |
| 546 | + * Guarded by its own option rather than the plugin version so it runs once | |
| 547 | + * and stays cheap: the check is a single autoloaded read on requests after | |
| 548 | + * the first. | |
| 549 | + * | |
| 550 | + * @since 2.3.1 | |
| 551 | + * | |
| 552 | + * @return void | |
| 553 | + */ | |
| 554 | + public static function maybe_backfill_icon_sizes(): void { | |
| 555 | + if (get_option(self::ICON_BACKFILL_OPTION)) { | |
| 556 | + return; | |
| 557 | + } | |
| 558 | + | |
| 559 | + // Written before the work, not after: a host that cannot resize must | |
| 560 | + // not retry on every admin request forever. | |
| 561 | + update_option(self::ICON_BACKFILL_OPTION, time(), true); | |
| 562 | + | |
| 563 | + $settings = (new self())->get_settings('site'); | |
| 564 | + | |
| 565 | + if (!is_array($settings)) { | |
| 566 | + return; | |
| 567 | + } | |
| 568 | + | |
| 569 | + foreach (['favicon_url', 'apple_touch_icon_url'] as $key) { | |
| 570 | + if (empty($settings[$key]) || !is_string($settings[$key])) { | |
| 571 | + continue; | |
| 572 | + } | |
| 573 | + | |
| 574 | + $attachment_id = self::icon_attachment_id($settings[$key]); | |
| 575 | + | |
| 576 | + if ($attachment_id) { | |
| 577 | + self::ensure_icon_sizes($attachment_id); | |
| 578 | + } | |
| 579 | + } | |
| 580 | + | |
| 581 | + delete_transient(self::ICON_URL_TRANSIENT); | |
| 582 | + } | |
| 583 | + | |
| 584 | + /** | |
| 585 | + * Attachment ID behind a configured icon URL, or 0 when it is not ours. | |
| 586 | + * | |
| 587 | + * attachment_url_to_postid() matches _wp_attached_file, which holds the | |
| 588 | + * ORIGINAL upload path, so the URL of a generated derivative | |
| 589 | + * (`logo-512x512.png`) returns 0 — and that is exactly what the media | |
| 590 | + * picker hands back when the user chooses a size. Attachment_Lookup falls | |
| 591 | + * back to the original behind it; the fallback started here and moved | |
| 592 | + * there when every other image lookup turned out to need it (#847). | |
| 593 | + * | |
| 594 | + * Shared with SEO_Manager's site-icon filter so both sides of the feature | |
| 595 | + * agree on which attachment a configured URL means. | |
| 596 | + * | |
| 597 | + * @since 2.3.1 | |
| 598 | + * | |
| 599 | + * @param string $url Configured icon URL. | |
| 600 | + * @return int Attachment ID, or 0. | |
| 601 | + */ | |
| 602 | + public static function icon_attachment_id(string $url): int { | |
| 603 | + return Attachment_Lookup::id_from_url($url); | |
| 604 | + } | |
| 605 | + | |
| 606 | + /** | |
| 607 | + * Which ICON_SIZES derivatives this attachment still needs. | |
| 608 | + * | |
| 609 | + * Split out from the generation so the decision can be asserted on its | |
| 610 | + * own: whether a size is skipped because it already exists or because it | |
| 611 | + * would upscale is invisible once both answers are "nothing was built". | |
| 612 | + * | |
| 613 | + * A source is measured by its SHORTER edge — a 400x40 banner cannot yield | |
| 614 | + * a true 192x192 — and anything reporting no dimensions at all (SVGs) is | |
| 615 | + * left alone. | |
| 616 | + * | |
| 617 | + * @since 2.3.1 | |
| 618 | + * | |
| 619 | + * @param array $meta Attachment metadata. | |
| 620 | + * @return array<string, array{width: int, height: int, crop: bool}> Sizes to build. | |
| 621 | + */ | |
| 622 | + public static function missing_icon_sizes(array $meta): array { | |
| 623 | + $source = min((int) ($meta['width'] ?? 0), (int) ($meta['height'] ?? 0)); | |
| 624 | + | |
| 625 | + if ($source < 1) { | |
| 626 | + return []; | |
| 627 | + } | |
| 628 | + | |
| 629 | + $wanted = []; | |
| 630 | + foreach (self::ICON_SIZES as $size) { | |
| 631 | + // Never upscale: a stretched source behind an accurate sizes="" | |
| 632 | + // label is worse than the honest near miss it would replace. | |
| 633 | + if (isset($meta['sizes']["site_icon-{$size}"]) || $size > $source) { | |
| 634 | + continue; | |
| 635 | + } | |
| 636 | + | |
| 637 | + $wanted["site_icon-{$size}"] = ['width' => $size, 'height' => $size, 'crop' => true]; | |
| 638 | + } | |
| 639 | + | |
| 640 | + return $wanted; | |
| 641 | + } | |
| 642 | + | |
| 643 | + /** | |
| 644 | + * Generate whatever ICON_SIZES derivatives this attachment is missing. | |
| 645 | + * | |
| 646 | + * Only the missing ones, and never one larger than the source: upscaling a | |
| 647 | + * small favicon would put a blurrier file behind an accurate sizes="" label | |
| 648 | + * than the honest near-miss it replaced. | |
| 649 | + * | |
| 650 | + * @since 2.3.1 | |
| 651 | + * | |
| 652 | + * @param int $attachment_id Attachment to build derivatives for. | |
| 653 | + * @return string[] Size names generated, empty when there was nothing to do. | |
| 654 | + */ | |
| 655 | + public static function ensure_icon_sizes(int $attachment_id): array { | |
| 656 | + $meta = wp_get_attachment_metadata($attachment_id); | |
| 657 | + | |
| 658 | + if (!is_array($meta)) { | |
| 659 | + return []; | |
| 660 | + } | |
| 661 | + | |
| 662 | + $wanted = self::missing_icon_sizes($meta); | |
| 663 | + | |
| 664 | + if (empty($wanted)) { | |
| 665 | + return []; | |
| 666 | + } | |
| 667 | + | |
| 668 | + $file = get_attached_file($attachment_id); | |
| 669 | + | |
| 670 | + if (!$file || !file_exists($file)) { | |
| 671 | + return []; | |
| 672 | + } | |
| 673 | + | |
| 674 | + $editor = wp_get_image_editor($file); | |
| 675 | + | |
| 676 | + if (is_wp_error($editor)) { | |
| 677 | + return []; | |
| 678 | + } | |
| 679 | + | |
| 680 | + $generated = $editor->multi_resize($wanted); | |
| 681 | + | |
| 682 | + if (empty($generated)) { | |
| 683 | + return []; | |
| 684 | + } | |
| 685 | + | |
| 686 | + $meta['sizes'] = array_merge($meta['sizes'] ?? [], $generated); | |
| 687 | + wp_update_attachment_metadata($attachment_id, $meta); | |
| 688 | + | |
| 689 | + return array_keys($generated); | |
| 690 | + } | |
| 691 | + | |
| 692 | + /** | |
| 249 | 693 | * Initialize WordPress filesystem |
| 250 | 694 | * |
| 251 | 695 | * @since 1.0.0 |
| 252 | 696 | * @return bool True if filesystem is initialized, false otherwise |
| @@ -442,8 +886,28 @@ | ||
| 442 | 886 | $body = ($custom !== '' && !$fully_blocked) |
| 443 | 887 | ? $this->strip_robots_header($custom) |
| 444 | 888 | : trim($this->generate_robots_txt()['content']); |
| 445 | 889 | |
| 890 | + // The per-agent AI directives are machine-owned, so they are composed | |
| 891 | + // here rather than stored: the textarea holds the user's body, with | |
| 892 | + // the fenced block stripped out of every read and re-applied on every | |
| 893 | + // render. A site-wide block already disallows everyone, so adding the | |
| 894 | + // per-agent group there would be noise restating the same refusal. | |
| 895 | + // Composed here rather than read from storage, for the same reason as | |
| 896 | + // the AI block below: the set of sitemaps an install publishes is a | |
| 897 | + // runtime fact. `robots_txt_content` is a snapshot of it taken at the | |
| 898 | + // last save, and nothing invalidated that snapshot, so deactivating a | |
| 899 | + // sitemap provider left its URL advertised and serving HTML (#835). | |
| 900 | + // Composing it on every render means the advertisement agrees with what | |
| 901 | + // the install publishes, in both directions, with no cache to expire. | |
| 902 | + if (!$fully_blocked) { | |
| 903 | + $body = $this->apply_sitemap_block($body); | |
| 904 | + } | |
| 905 | + | |
| 906 | + if (!$fully_blocked) { | |
| 907 | + $body = $this->apply_ai_crawler_block($body, $settings); | |
| 908 | + } | |
| 909 | + | |
| 446 | 910 | if ($body === '') { |
| 447 | 911 | return ''; |
| 448 | 912 | } |
| 449 | 913 | |
| @@ -450,8 +914,87 @@ | ||
| 450 | 914 | return $this->robots_txt_header() . $body . "\n"; |
| 451 | 915 | } |
| 452 | 916 | |
| 453 | 917 | /** |
| 918 | + * Replace the generated Sitemap block with the one this install publishes. | |
| 919 | + * | |
| 920 | + * @since 2.14.0 | |
| 921 | + * @param string $body Robots.txt body, without the header. | |
| 922 | + * @return string | |
| 923 | + */ | |
| 924 | + private function apply_sitemap_block(string $body): string { | |
| 925 | + $stripped = $this->strip_generated_sitemap_block($body); | |
| 926 | + $urls = $this->get_sitemap_urls_for_robots(); | |
| 927 | + | |
| 928 | + if (empty($urls)) { | |
| 929 | + return $stripped; | |
| 930 | + } | |
| 931 | + | |
| 932 | + $block = ''; | |
| 933 | + foreach ($urls as $url) { | |
| 934 | + $block .= 'Sitemap: ' . $url . "\n"; | |
| 935 | + } | |
| 936 | + | |
| 937 | + if ('' === trim($stripped)) { | |
| 938 | + return trim($block); | |
| 939 | + } | |
| 940 | + | |
| 941 | + // The grammar build_robots_txt_content() writes: one blank line before | |
| 942 | + // the block, none inside it. A blank line terminates a record in the | |
| 943 | + // robots.txt grammar, so a line between every directive is invalid. | |
| 944 | + return rtrim($stripped) . "\n\n" . trim($block); | |
| 945 | + } | |
| 946 | + | |
| 947 | + /** | |
| 948 | + * Remove the plugin-written Sitemap block from a stored body. | |
| 949 | + * | |
| 950 | + * Only the trailing run of `Sitemap:` lines is removed, which is the exact | |
| 951 | + * shape `build_robots_txt_content()` writes: a blank line, then nothing but | |
| 952 | + * `Sitemap:` lines to the end of the body. A `Sitemap:` line anywhere else | |
| 953 | + * was typed by the site owner and is left exactly where they put it, which | |
| 954 | + * is why this cannot simply strip every matching line. | |
| 955 | + * | |
| 956 | + * @since 2.14.0 | |
| 957 | + * @param string $body Robots.txt body. | |
| 958 | + * @return string | |
| 959 | + */ | |
| 960 | + private function strip_generated_sitemap_block(string $body): string { | |
| 961 | + $lines = preg_split('/\R/', $body); | |
| 962 | + | |
| 963 | + if (!is_array($lines)) { | |
| 964 | + return $body; | |
| 965 | + } | |
| 966 | + | |
| 967 | + $cut = count($lines); | |
| 968 | + | |
| 969 | + // Walk back over the trailing block: sitemap lines, and the blank lines | |
| 970 | + // that separate or pad it. Anything else ends the block. | |
| 971 | + for ($i = count($lines) - 1; $i >= 0; $i--) { | |
| 972 | + $line = trim($lines[$i]); | |
| 973 | + | |
| 974 | + if ('' === $line) { | |
| 975 | + $cut = $i; | |
| 976 | + continue; | |
| 977 | + } | |
| 978 | + | |
| 979 | + if (0 === stripos($line, 'sitemap:')) { | |
| 980 | + $cut = $i; | |
| 981 | + continue; | |
| 982 | + } | |
| 983 | + | |
| 984 | + break; | |
| 985 | + } | |
| 986 | + | |
| 987 | + if ($cut >= count($lines)) { | |
| 988 | + return $body; | |
| 989 | + } | |
| 990 | + | |
| 991 | + // Nothing but sitemap lines in the whole body means there is no owner | |
| 992 | + // content to keep. | |
| 993 | + return rtrim(implode("\n", array_slice($lines, 0, $cut))); | |
| 994 | + } | |
| 995 | + | |
| 996 | + /** | |
| 454 | 997 | * Resolve the robots.txt actually served to crawlers, with its origin. |
| 455 | 998 | * |
| 456 | 999 | * Lets an API/MCP consumer see the effective output without crawling the |
| 457 | 1000 | * URL. Mirrors serving precedence: a physical robots.txt in the web root is |
| @@ -468,8 +1011,9 @@ | ||
| 468 | 1011 | $robots_file = ABSPATH . 'robots.txt'; |
| 469 | 1012 | if (file_exists($robots_file) && is_readable($robots_file)) { |
| 470 | 1013 | // phpcs:ignore WordPress.WP.AlternativeFunctions.file_get_contents_file_get_contents -- Reading a public web-root file; WP_Filesystem is not available on front-end requests. |
| 471 | 1014 | return [ |
| 1015 | + // phpcs:ignore WordPress.WP.AlternativeFunctions.file_get_contents_file_get_contents -- reads a local file the plugin just located; WP_Filesystem would need credentials on some hosts. | |
| 472 | 1016 | 'content' => (string) file_get_contents($robots_file), |
| 473 | 1017 | 'is_default' => false, |
| 474 | 1018 | 'source' => 'file', |
| 475 | 1019 | ]; |
| @@ -496,8 +1040,68 @@ | ||
| 496 | 1040 | ]; |
| 497 | 1041 | } |
| 498 | 1042 | |
| 499 | 1043 | /** |
| 1044 | + * Describe how /robots.txt is actually delivered, and whether that still | |
| 1045 | + * matches the saved settings. | |
| 1046 | + * | |
| 1047 | + * The admin screen edits settings, but a physical robots.txt in the web root | |
| 1048 | + * is served directly by the web server and bypasses the `robots_txt` filter | |
| 1049 | + * entirely. When those two drift, the editor is showing content no crawler | |
| 1050 | + * ever sees — the conflict this exists to surface. | |
| 1051 | + * | |
| 1052 | + * @since 1.31.0 | |
| 1053 | + * | |
| 1054 | + * @return array{content: string, source: string, is_default: bool, in_sync: bool, out_of_sync_reason: string, url: string} | |
| 1055 | + * The served content and its origin, whether it still reflects the | |
| 1056 | + * body the editor is showing, why it does not when it does not | |
| 1057 | + * ('file_drift' or 'crawl_blocked'), and the public URL it is | |
| 1058 | + * served from. | |
| 1059 | + */ | |
| 1060 | + public function get_robots_txt_delivery(): array { | |
| 1061 | + $effective = $this->get_effective_robots_txt(); | |
| 1062 | + $settings = $this->get_settings('site'); | |
| 1063 | + | |
| 1064 | + // Compare bodies, not raw strings: the auto-generated header carries a | |
| 1065 | + // regeneration timestamp that always differs and means nothing here. | |
| 1066 | + // The AI crawler block is composed at render time on both sides, so it | |
| 1067 | + // is identical by construction and comparing it would only ever report | |
| 1068 | + // a false drift the admin cannot act on. | |
| 1069 | + $served = $this->strip_ai_crawler_block($this->strip_robots_header($effective['content'])); | |
| 1070 | + | |
| 1071 | + // Measure against the body the editor is displaying — get_served_robots_body() | |
| 1072 | + // — not against render_robots_txt(). Two things made the old comparison | |
| 1073 | + // report "in sync" while the screen showed rules no crawler receives: | |
| 1074 | + // a physical file was compared to a freshly rendered body rather than | |
| 1075 | + // to the stored override the textarea shows, and a site-wide crawl | |
| 1076 | + // block makes render_robots_txt() return the generated "Disallow: /" | |
| 1077 | + // on both sides of the comparison, so it always matched. | |
| 1078 | + $expected = $this->get_served_robots_body(); | |
| 1079 | + | |
| 1080 | + // Management off: WordPress serves its own default and the editor is not | |
| 1081 | + // claiming anything is live, so there is nothing to be out of sync with. | |
| 1082 | + $managed = !empty($settings['robots_txt_enabled']); | |
| 1083 | + $in_sync = !$managed || $served === $expected; | |
| 1084 | + | |
| 1085 | + $reason = ''; | |
| 1086 | + if (!$in_sync) { | |
| 1087 | + // A crawl block is a deliberate override, not a stale file, and the | |
| 1088 | + // admin needs to be told which of the two they are looking at. | |
| 1089 | + $blocked = empty($settings['allow_search_engines'] ?? true) || !get_option('blog_public'); | |
| 1090 | + $reason = $blocked ? 'crawl_blocked' : 'file_drift'; | |
| 1091 | + } | |
| 1092 | + | |
| 1093 | + return [ | |
| 1094 | + 'content' => $effective['content'], | |
| 1095 | + 'source' => $effective['source'], | |
| 1096 | + 'is_default' => $effective['is_default'], | |
| 1097 | + 'in_sync' => $in_sync, | |
| 1098 | + 'out_of_sync_reason' => $reason, | |
| 1099 | + 'url' => home_url('/robots.txt'), | |
| 1100 | + ]; | |
| 1101 | + } | |
| 1102 | + | |
| 1103 | + /** | |
| 500 | 1104 | * Keep the physical robots.txt file in step with the saved settings. |
| 501 | 1105 | * |
| 502 | 1106 | * When management is enabled the physical file is the source of truth the |
| 503 | 1107 | * web server serves, so this makes sure it exists and matches the effective |
| @@ -534,8 +1138,9 @@ | ||
| 534 | 1138 | // do_robots()/robots_txt filter entirely, so without this those lines |
| 535 | 1139 | // are silently dropped. ThinkRank's own filter_robots_txt callback just |
| 536 | 1140 | // re-returns this same content (it calls render_robots_txt(), which does |
| 537 | 1141 | // not re-apply the filter), so there is no recursion or double-append. |
| 1142 | + // phpcs:ignore WordPress.NamingConventions.PrefixAllGlobals.NonPrefixedHooknameFound -- WPML/core hook, not ours to name. | |
| 538 | 1143 | $content = (string) apply_filters('robots_txt', $content, (bool) get_option('blog_public')); |
| 539 | 1144 | if ($content === '') { |
| 540 | 1145 | return false; |
| 541 | 1146 | } |
| @@ -840,17 +1445,19 @@ | ||
| 840 | 1445 | $home_text = $settings['breadcrumb_home_text'] ?? 'Home'; |
| 841 | 1446 | if (empty($home_text)) { |
| 842 | 1447 | $optimization['warnings'][] = 'Empty home text reduces accessibility for screen readers'; |
| 843 | 1448 | $optimization['score'] -= 15; |
| 844 | - } elseif (strlen($home_text) > 20) { | |
| 845 | - $optimization['suggestions'][] = 'Keep home text concise (current: ' . strlen($home_text) . ' chars)'; | |
| 1449 | + } elseif (mb_strlen($home_text) > 20) { | |
| 1450 | + // mb_strlen: this number is shown to the user as "chars" (#687). | |
| 1451 | + $optimization['suggestions'][] = 'Keep home text concise (current: ' . mb_strlen($home_text) . ' chars)'; | |
| 846 | 1452 | $optimization['score'] -= 5; |
| 847 | 1453 | } |
| 848 | 1454 | |
| 849 | 1455 | // Check prefix usage |
| 850 | 1456 | $prefix = $settings['breadcrumb_prefix'] ?? ''; |
| 851 | - if (!empty($prefix) && strlen($prefix) > 50) { | |
| 852 | - $optimization['suggestions'][] = 'Breadcrumb prefix is quite long (' . strlen($prefix) . ' chars) - consider shortening'; | |
| 1457 | + if (!empty($prefix) && mb_strlen($prefix) > 50) { | |
| 1458 | + // mb_strlen: this number is shown to the user as "chars" (#687). | |
| 1459 | + $optimization['suggestions'][] = 'Breadcrumb prefix is quite long (' . mb_strlen($prefix) . ' chars) - consider shortening'; | |
| 853 | 1460 | $optimization['score'] -= 5; |
| 854 | 1461 | } |
| 855 | 1462 | |
| 856 | 1463 | // Current page display |
| @@ -981,13 +1588,16 @@ | ||
| 981 | 1588 | } |
| 982 | 1589 | |
| 983 | 1590 | // Additional logo analysis for local images |
| 984 | 1591 | if (!empty($logo_url) && filter_var($logo_url, FILTER_VALIDATE_URL)) { |
| 985 | - $attachment_id = attachment_url_to_postid($logo_url); | |
| 1592 | + $attachment_id = Attachment_Lookup::id_from_url($logo_url); | |
| 986 | 1593 | if ($attachment_id) { |
| 987 | 1594 | $image_meta = wp_get_attachment_metadata($attachment_id); |
| 988 | - $width = isset($image_meta['width']) ? (int) $image_meta['width'] : 0; | |
| 989 | - $height = isset($image_meta['height']) ? (int) $image_meta['height'] : 0; | |
| 1595 | + // The configured file's own size — a logo picked at a generated | |
| 1596 | + // size is not as large as the upload behind it. | |
| 1597 | + $logo_file = Attachment_Lookup::describe($attachment_id, $logo_url); | |
| 1598 | + $width = $logo_file['width']; | |
| 1599 | + $height = $logo_file['height']; | |
| 990 | 1600 | |
| 991 | 1601 | // SVG logos store 0x0 metadata — no dimension/ratio analysis |
| 992 | 1602 | // is possible (and dividing by 0 is fatal). |
| 993 | 1603 | if ($image_meta && $width > 0 && $height > 0) { |
| @@ -1090,11 +1700,12 @@ | ||
| 1090 | 1700 | $optimization['score'] -= 15; |
| 1091 | 1701 | } |
| 1092 | 1702 | } |
| 1093 | 1703 | |
| 1094 | - // Business type validation | |
| 1095 | - if (empty($settings['business_type'])) { | |
| 1096 | - $optimization['suggestions'][] = 'Select a specific business type for better schema markup'; | |
| 1704 | + // Business type validation (shared rule, one message — #622). | |
| 1705 | + $business_type = $this->business_type_status($settings); | |
| 1706 | + if ('suggestion' === $business_type['status']) { | |
| 1707 | + $optimization['suggestions'][] = $business_type['message']; | |
| 1097 | 1708 | $optimization['score'] -= 5; |
| 1098 | 1709 | } |
| 1099 | 1710 | |
| 1100 | 1711 | // Email validation |
| @@ -1624,24 +2235,17 @@ | ||
| 1624 | 2235 | 'icon' => '✗' |
| 1625 | 2236 | ]; |
| 1626 | 2237 | } |
| 1627 | 2238 | |
| 1628 | - // Business Type validation | |
| 1629 | - if (!empty($settings['business_type']) && $settings['business_type'] !== 'LocalBusiness') { | |
| 1630 | - $field_details[] = [ | |
| 1631 | - 'field' => 'business_type', | |
| 1632 | - 'label' => 'Business type is selected for proper schema markup.', | |
| 1633 | - 'status' => 'valid', | |
| 1634 | - 'icon' => '✓' | |
| 1635 | - ]; | |
| 1636 | - } else { | |
| 1637 | - $field_details[] = [ | |
| 1638 | - 'field' => 'business_type', | |
| 1639 | - 'label' => 'Specific business type selection recommended for better schema markup.', | |
| 1640 | - 'status' => 'suggestion', | |
| 1641 | - 'icon' => '⚠' | |
| 1642 | - ]; | |
| 1643 | - } | |
| 2239 | + // Business Type validation — see business_type_status() for why there | |
| 2240 | + // is exactly one rule here now (#622). | |
| 2241 | + $business_type = $this->business_type_status($settings); | |
| 2242 | + $field_details[] = [ | |
| 2243 | + 'field' => 'business_type', | |
| 2244 | + 'label' => $business_type['message'], | |
| 2245 | + 'status' => $business_type['status'], | |
| 2246 | + 'icon' => 'valid' === $business_type['status'] ? '✓' : '⚠', | |
| 2247 | + ]; | |
| 1644 | 2248 | |
| 1645 | 2249 | // Address validation (NAP consistency) |
| 1646 | 2250 | $address_fields = ['business_address', 'business_city', 'business_state', 'business_country']; |
| 1647 | 2251 | $address_complete = true; |
| @@ -2260,12 +2864,15 @@ | ||
| 2260 | 2864 | 'warnings' => [], |
| 2261 | 2865 | 'suggestions' => [] |
| 2262 | 2866 | ]; |
| 2263 | 2867 | |
| 2264 | - // Validate business name (required for local SEO) | |
| 2868 | + // Business name is what makes the LocalBusiness schema useful, but it | |
| 2869 | + // cannot be a blocking error: the toggle is what reveals the business | |
| 2870 | + // fields, so requiring the name up front makes enabling Local SEO | |
| 2871 | + // impossible. The frontend already skips the output while the name is | |
| 2872 | + // empty (see Seo_Manager::output_local_seo_meta_tags()). | |
| 2265 | 2873 | if (empty($settings['business_name'])) { |
| 2266 | - $validation['errors'][] = 'Business name is required when local SEO is enabled'; | |
| 2267 | - $validation['valid'] = false; | |
| 2874 | + $validation['warnings'][] = 'Business name is missing - required before local business schema is output'; | |
| 2268 | 2875 | } elseif (strlen($settings['business_name']) > 100) { |
| 2269 | 2876 | $validation['warnings'][] = 'Business name is very long, consider shortening for better display'; |
| 2270 | 2877 | } |
| 2271 | 2878 | |
| @@ -2322,11 +2929,15 @@ | ||
| 2322 | 2929 | } else { |
| 2323 | 2930 | $validation['suggestions'][] = 'Add business hours to improve local search visibility'; |
| 2324 | 2931 | } |
| 2325 | 2932 | |
| 2326 | - // Validate business type | |
| 2327 | - if (empty($settings['business_type'])) { | |
| 2328 | - $validation['suggestions'][] = 'Select a specific business type for better schema markup'; | |
| 2933 | + // Business type, through the shared rule (#622). This is the only place | |
| 2934 | + // it is reported on the generic path: validate_settings() with no tab | |
| 2935 | + // context attaches basic-info field details, not business-info ones, so | |
| 2936 | + // without this the setting would go unreported there entirely. | |
| 2937 | + $business_type = $this->business_type_status($settings); | |
| 2938 | + if ('suggestion' === $business_type['status']) { | |
| 2939 | + $validation['suggestions'][] = $business_type['message']; | |
| 2329 | 2940 | } |
| 2330 | 2941 | |
| 2331 | 2942 | return $validation; |
| 2332 | 2943 | } |
| @@ -2406,8 +3017,201 @@ | ||
| 2406 | 3017 | return $output; |
| 2407 | 3018 | } |
| 2408 | 3019 | |
| 2409 | 3020 | /** |
| 3021 | + * Keys the Site Identity screens store beyond the 16 defaults. | |
| 3022 | + * | |
| 3023 | + * Title formats, breadcrumb configuration, the hero fields, the business | |
| 3024 | + * block and the wizard's identity fields are all real settings written by | |
| 3025 | + * this manager, none of which get_default_settings() names — it seeds only | |
| 3026 | + * the values a fresh install needs. Gating on defaults alone would stop | |
| 3027 | + * every one of them saving (#452). | |
| 3028 | + * | |
| 3029 | + * @since 2.0.1 | |
| 3030 | + * | |
| 3031 | + * @return string[] | |
| 3032 | + */ | |
| 3033 | + /** | |
| 3034 | + * The stored alternate name(s), shaped for schema output. | |
| 3035 | + * | |
| 3036 | + * schema.org and Google both allow `alternateName` to carry one value or | |
| 3037 | + * several, and the store already round-trips either shape, so this accepts | |
| 3038 | + * both and normalises: null when there is nothing to publish, a bare string | |
| 3039 | + * for one name, a list for more. Emitting a one-element array would be | |
| 3040 | + * valid but noisier than it needs to be. | |
| 3041 | + * | |
| 3042 | + * Shared because both WebSite producers need it and must agree — a property | |
| 3043 | + * added to one and not the other is how #688 happened. | |
| 3044 | + * | |
| 3045 | + * @since 2.7.0 | |
| 3046 | + * | |
| 3047 | + * @param mixed $value Stored alternate_name value. | |
| 3048 | + * @return string|string[]|null | |
| 3049 | + */ | |
| 3050 | + public static function alternate_name_for_schema($value) { | |
| 3051 | + $names = []; | |
| 3052 | + | |
| 3053 | + foreach ((array) $value as $name) { | |
| 3054 | + if (!is_scalar($name)) { | |
| 3055 | + continue; | |
| 3056 | + } | |
| 3057 | + | |
| 3058 | + $name = trim((string) $name); | |
| 3059 | + | |
| 3060 | + if ('' !== $name && !in_array($name, $names, true)) { | |
| 3061 | + $names[] = $name; | |
| 3062 | + } | |
| 3063 | + } | |
| 3064 | + | |
| 3065 | + if (empty($names)) { | |
| 3066 | + return null; | |
| 3067 | + } | |
| 3068 | + | |
| 3069 | + return 1 === count($names) ? $names[0] : $names; | |
| 3070 | + } | |
| 3071 | + | |
| 3072 | + protected function additional_setting_keys(): array { | |
| 3073 | + return [ | |
| 3074 | + // Title formats, one per context. | |
| 3075 | + 'homepage_title', 'post_title', 'page_title', 'category_title', | |
| 3076 | + 'tag_title', 'author_title', 'search_title', 'archive_title', | |
| 3077 | + // The blog-index homepage's meta description (#897). | |
| 3078 | + 'homepage_description', | |
| 3079 | + // Breadcrumbs. | |
| 3080 | + 'breadcrumb_prefix', 'show_current_page', 'breadcrumb_use_seo_title', | |
| 3081 | + // Identity, as written by the setup wizard and the importers. | |
| 3082 | + 'alternate_name', 'identity_type', 'represents', | |
| 3083 | + 'default_meta_description', 'default_social_image', | |
| 3084 | + 'social_media_accounts', | |
| 3085 | + // Schema toggles that live on this screen. | |
| 3086 | + 'organization_schema', 'knowledge_graph', | |
| 3087 | + // Robots rules composed by the Robots.txt panel. | |
| 3088 | + 'custom_robots_rules', | |
| 3089 | + // Per-agent AI crawler allow/block map (#657). | |
| 3090 | + 'ai_crawler_rules', | |
| 3091 | + // Hero section. | |
| 3092 | + 'hero_title', 'hero_subtitle', 'hero_cta_text', 'hero_cta_url', | |
| 3093 | + 'hero_background_image', | |
| 3094 | + // Local SEO / business details. | |
| 3095 | + 'local_seo_enabled', 'business_type', 'business_name', | |
| 3096 | + 'business_address', 'business_city', 'business_state', | |
| 3097 | + 'business_postal_code', 'business_country', 'business_phone', | |
| 3098 | + 'business_email', 'business_latitude', 'business_longitude', | |
| 3099 | + 'business_price_range', 'business_hours', | |
| 3100 | + ]; | |
| 3101 | + } | |
| 3102 | + | |
| 3103 | + /** | |
| 3104 | + * Sanitize settings, normalising the AI crawler rule map. | |
| 3105 | + * | |
| 3106 | + * The generic array sanitizer keeps the shape but says nothing about the | |
| 3107 | + * values: a payload could store `ai_crawler_rules[gptbot] = "maybe"`, or a | |
| 3108 | + * slug no crawler answers to, and both would round-trip through every | |
| 3109 | + * later response. Normalising here rather than in the REST handler puts it | |
| 3110 | + * on the one path every writer shares — the settings route, the robots | |
| 3111 | + * route and the MCP abilities all land in save_settings() (#657). | |
| 3112 | + * | |
| 3113 | + * @since 2.5.0 | |
| 3114 | + * | |
| 3115 | + * @param array $settings Settings to sanitize. | |
| 3116 | + * @param string $context_type Context type. | |
| 3117 | + * @return array Sanitized settings. | |
| 3118 | + */ | |
| 3119 | + protected function sanitize_settings(array $settings, string $context_type = 'site'): array { | |
| 3120 | + $sanitized = parent::sanitize_settings($settings, $context_type); | |
| 3121 | + | |
| 3122 | + if (array_key_exists('ai_crawler_rules', $sanitized)) { | |
| 3123 | + $sanitized['ai_crawler_rules'] = AI_Crawlers::normalize_rules($sanitized['ai_crawler_rules']); | |
| 3124 | + } | |
| 3125 | + | |
| 3126 | + // Same reasoning one key up, for the scheme override (#638). Anything | |
| 3127 | + // that is not one of the three modes means "follow WordPress", and is | |
| 3128 | + // stored as that rather than kept verbatim — otherwise get-site-identity | |
| 3129 | + // -settings would report a scheme the site does not actually publish. | |
| 3130 | + if (array_key_exists('canonical_scheme', $sanitized)) { | |
| 3131 | + $sanitized['canonical_scheme'] = in_array($sanitized['canonical_scheme'], Url_Scheme::MODES, true) | |
| 3132 | + ? $sanitized['canonical_scheme'] | |
| 3133 | + : Url_Scheme::AUTOMATIC; | |
| 3134 | + } | |
| 3135 | + | |
| 3136 | + // Same reasoning again for the business type. It goes straight into | |
| 3137 | + // LocalBusiness schema, so a type that is not in the schema.org | |
| 3138 | + // vocabulary is invalid structured data — and storing it verbatim would | |
| 3139 | + // have get-site-identity-settings report a type the site cannot | |
| 3140 | + // actually publish. An empty value keeps meaning "not set"; anything | |
| 3141 | + // else unrecognised falls back to the general-purpose root (#623). | |
| 3142 | + if (array_key_exists('business_type', $sanitized)) { | |
| 3143 | + $type = (string) $sanitized['business_type']; | |
| 3144 | + | |
| 3145 | + if ('' !== $type && !\ThinkRank\Config\Local_Business_Types_Config::is_valid($type)) { | |
| 3146 | + $type = \ThinkRank\Config\Local_Business_Types_Config::ROOT; | |
| 3147 | + } | |
| 3148 | + | |
| 3149 | + $sanitized['business_type'] = $type; | |
| 3150 | + } | |
| 3151 | + | |
| 3152 | + return $sanitized; | |
| 3153 | + } | |
| 3154 | + | |
| 3155 | + /** | |
| 3156 | + * schema.org's general-purpose LocalBusiness type. | |
| 3157 | + * | |
| 3158 | + * The default, the first option in the control, and a valid answer in its | |
| 3159 | + * own right — which is the whole point of #622. | |
| 3160 | + * | |
| 3161 | + * @since 2.10.0 | |
| 3162 | + * @var string | |
| 3163 | + */ | |
| 3164 | + private const GENERAL_BUSINESS_TYPE = 'LocalBusiness'; | |
| 3165 | + | |
| 3166 | + /** | |
| 3167 | + * The one rule for whether a business type needs the user's attention. | |
| 3168 | + * | |
| 3169 | + * There were three, with two wordings and two different conditions. Two | |
| 3170 | + * fired when the value was empty; the third fired when it WAS | |
| 3171 | + * `LocalBusiness` — which is the default, the first option in the control | |
| 3172 | + * and a perfectly valid schema.org type. So the warning appeared out of the | |
| 3173 | + * box for every site, could not be cleared without choosing a type that | |
| 3174 | + * might be inaccurate, and on an empty value it appeared three times in two | |
| 3175 | + * different phrasings, which is why it was reported as showing twice (#622). | |
| 3176 | + * | |
| 3177 | + * The rule now: a type is expected, and any type in the vocabulary is a | |
| 3178 | + * correct answer. Only an unset value is worth prompting about. | |
| 3179 | + * `LocalBusiness` is the general-purpose answer and is accepted as one — | |
| 3180 | + * with a note that a more specific type sharpens the schema, phrased as the | |
| 3181 | + * guidance it is rather than as a fault the user has to clear. | |
| 3182 | + * | |
| 3183 | + * @since 2.10.0 | |
| 3184 | + * | |
| 3185 | + * @param array $settings Site identity settings. | |
| 3186 | + * @return array{status:string,message:string} `valid` or `suggestion`. | |
| 3187 | + */ | |
| 3188 | + private function business_type_status(array $settings): array { | |
| 3189 | + $type = trim((string) ($settings['business_type'] ?? '')); | |
| 3190 | + | |
| 3191 | + if ('' === $type) { | |
| 3192 | + return [ | |
| 3193 | + 'status' => 'suggestion', | |
| 3194 | + 'message' => __('Select a business type so your local schema describes the right kind of business.', 'thinkrank'), | |
| 3195 | + ]; | |
| 3196 | + } | |
| 3197 | + | |
| 3198 | + // The literal rather than a constant from the expanded type list (#623): | |
| 3199 | + // that lands on its own branch, and this fix must not wait on it. | |
| 3200 | + if (self::GENERAL_BUSINESS_TYPE === $type) { | |
| 3201 | + return [ | |
| 3202 | + 'status' => 'valid', | |
| 3203 | + 'message' => __('Business type is set to Local Business. A more specific type sharpens your schema, if one fits.', 'thinkrank'), | |
| 3204 | + ]; | |
| 3205 | + } | |
| 3206 | + | |
| 3207 | + return [ | |
| 3208 | + 'status' => 'valid', | |
| 3209 | + 'message' => __('Business type is selected for proper schema markup.', 'thinkrank'), | |
| 3210 | + ]; | |
| 3211 | + } | |
| 3212 | + | |
| 3213 | + /** | |
| 2410 | 3214 | * Get default settings for a context type (implements interface) |
| 2411 | 3215 | * |
| 2412 | 3216 | * @since 1.0.0 |
| 2413 | 3217 | * |
| @@ -2427,9 +3231,32 @@ | ||
| 2427 | 3231 | 'breadcrumb_home_text' => 'Home', |
| 2428 | 3232 | 'breadcrumb_separator' => '>', |
| 2429 | 3233 | 'robots_txt_enabled' => true, |
| 2430 | 3234 | 'allow_search_engines' => true, |
| 3235 | + // Answer 404 when a content selector in the URL resolved to | |
| 3236 | + // nothing (#634). On by default, unlike the other new settings | |
| 3237 | + // here: it changes no URL a visitor or a correct crawler uses, only | |
| 3238 | + // ones where WordPress resolved nothing and served the blog listing | |
| 3239 | + // at 200 anyway. | |
| 3240 | + 'query_protection' => true, | |
| 3241 | + | |
| 3242 | + // Feed controls (#635). All three off, so an upgrade changes | |
| 3243 | + // nothing about what an existing site already sends its | |
| 3244 | + // subscribers; a brand-new install is seeded with the signature and | |
| 3245 | + // the noindex on, in Activator::seed_feed_defaults(). | |
| 3246 | + 'feed_excerpt_only' => false, | |
| 3247 | + 'feed_source_link' => false, | |
| 3248 | + 'feed_noindex' => false, | |
| 3249 | + | |
| 3250 | + // The scheme self-referential URLs go out with (#638). 'automatic' | |
| 3251 | + // means substitute nothing and follow WordPress, which is what | |
| 3252 | + // every site did before the setting existed. | |
| 3253 | + 'canonical_scheme' => Url_Scheme::AUTOMATIC, | |
| 2431 | 3254 | 'robots_txt_content' => '', |
| 3255 | + // Empty map = every AI crawler allowed. Defaults must stay | |
| 3256 | + // permissive so an upgrade never starts blocking a crawler a site | |
| 3257 | + // was happily serving (#657). | |
| 3258 | + 'ai_crawler_rules' => [], | |
| 2432 | 3259 | 'logo_url' => '', |
| 2433 | 3260 | 'favicon_url' => '', |
| 2434 | 3261 | 'apple_touch_icon_url' => '' |
| 2435 | 3262 | ]; |
| @@ -2548,10 +3375,13 @@ | ||
| 2548 | 3375 | */ |
| 2549 | 3376 | private function prepare_title_placeholders(array $data, string $context, array $settings): array { |
| 2550 | 3377 | $placeholders = [ |
| 2551 | 3378 | '%title%' => $data['title'] ?? '', |
| 2552 | - '%sitename%' => $settings['site_name'] ?? get_bloginfo('name'), | |
| 2553 | - '%tagline%' => $settings['tagline'] ?? get_bloginfo('description'), | |
| 3379 | + // `?:` rather than `??`: these are persisted as '' rather than left | |
| 3380 | + // unset, and '' is not null, so the null-coalesce never reached the | |
| 3381 | + // WordPress fallback (#398). | |
| 3382 | + '%sitename%' => ($settings['site_name'] ?? '') ?: get_bloginfo('name'), | |
| 3383 | + '%tagline%' => ($settings['tagline'] ?? '') ?: get_bloginfo('description'), | |
| 2554 | 3384 | '%separator%' => '', // Will be replaced with actual separator |
| 2555 | 3385 | '%category%' => '', |
| 2556 | 3386 | '%author%' => '', |
| 2557 | 3387 | '%date%' => '', |
| @@ -2635,16 +3465,17 @@ | ||
| 2635 | 3465 | // Remove extra whitespace |
| 2636 | 3466 | $title = preg_replace('/\s+/', ' ', $title); |
| 2637 | 3467 | $title = trim($title); |
| 2638 | 3468 | |
| 2639 | - // Ensure title is not too long (60 characters max for SEO) | |
| 2640 | - if (strlen($title) > 60) { | |
| 2641 | - // Try to truncate at word boundary | |
| 2642 | - $title = wp_trim_words($title, 8, '...'); | |
| 2643 | - if (strlen($title) > 60) { | |
| 2644 | - $title = substr($title, 0, 57) . '...'; | |
| 2645 | - } | |
| 2646 | - } | |
| 3469 | + // Ensure title is not too long (60 characters max for SEO). | |
| 3470 | + // All three units here were wrong for non-Latin text: strlen() counts | |
| 3471 | + // BYTES so the gate fired at 20 Thai characters, wp_trim_words() counts | |
| 3472 | + // CHARACTERS on th/ja/zh_* so `8` cut the title to 8 of them, and | |
| 3473 | + // substr() cuts bytes so it split a character mid-sequence (#687). | |
| 3474 | + $title = \ThinkRank\Core\Seo_Text::trim_to_length( | |
| 3475 | + $title, | |
| 3476 | + \ThinkRank\Core\Seo_Text::TITLE_MAX_LENGTH | |
| 3477 | + ); | |
| 2647 | 3478 | |
| 2648 | 3479 | // Ensure title is not empty |
| 2649 | 3480 | if (empty($title)) { |
| 2650 | 3481 | $title = get_bloginfo('name'); |
| @@ -3034,8 +3865,29 @@ | ||
| 3034 | 3865 | * @since 1.0.0 |
| 3035 | 3866 | * @return array Array of sitemap URLs |
| 3036 | 3867 | */ |
| 3037 | 3868 | private function get_sitemap_urls_for_robots(): array { |
| 3869 | + // One wrapper over every return path below, including the #104 extras. | |
| 3870 | + // The Sitemap: line is the only absolute URL of ours in robots.txt and | |
| 3871 | + // the one a crawler follows to find everything else, so it has to carry | |
| 3872 | + // the site's scheme preference (#638). Applied here rather than where | |
| 3873 | + // the body is assembled, because that path also renders a robots.txt a | |
| 3874 | + // site owner typed themselves, and their text is not ours to rewrite. | |
| 3875 | + return array_map( | |
| 3876 | + static function (string $url): string { | |
| 3877 | + return Url_Scheme::apply($url); | |
| 3878 | + }, | |
| 3879 | + $this->collect_sitemap_urls_for_robots() | |
| 3880 | + ); | |
| 3881 | + } | |
| 3882 | + | |
| 3883 | + /** | |
| 3884 | + * The sitemap URLs robots.txt advertises, before the scheme preference. | |
| 3885 | + * | |
| 3886 | + * @since 1.0.0 | |
| 3887 | + * @return array Array of sitemap URLs | |
| 3888 | + */ | |
| 3889 | + private function collect_sitemap_urls_for_robots(): array { | |
| 3038 | 3890 | try { |
| 3039 | 3891 | // Get sitemap settings |
| 3040 | 3892 | $sitemap_generator = new \ThinkRank\SEO\Sitemap_Generator(); |
| 3041 | 3893 | $sitemap_settings = $sitemap_generator->get_settings('site'); |
| @@ -3047,29 +3899,88 @@ | ||
| 3047 | 3899 | |
| 3048 | 3900 | $sitemap_urls = []; |
| 3049 | 3901 | $site_url = home_url(); |
| 3050 | 3902 | |
| 3051 | - // Extract enabled sitemap URLs | |
| 3903 | + // Extract enabled sitemap URLs. When the index is enabled it is the | |
| 3904 | + // only entry worth advertising: every child sitemap is already | |
| 3905 | + // listed inside it, so naming them again in robots.txt is pure | |
| 3906 | + // redundancy and drifts out of date as soon as a post type is added. | |
| 3907 | + $index_url = ''; | |
| 3052 | 3908 | if (!empty($sitemap_settings['sitemap_urls']) && is_array($sitemap_settings['sitemap_urls'])) { |
| 3053 | 3909 | foreach ($sitemap_settings['sitemap_urls'] as $sitemap) { |
| 3054 | - if (!empty($sitemap['enabled']) && !empty($sitemap['url'])) { | |
| 3055 | - $sitemap_urls[] = $site_url . $sitemap['url']; | |
| 3910 | + if (empty($sitemap['enabled']) || empty($sitemap['url'])) { | |
| 3911 | + continue; | |
| 3056 | 3912 | } |
| 3913 | + | |
| 3914 | + if (($sitemap['type'] ?? '') === 'index') { | |
| 3915 | + $index_url = $site_url . $sitemap['url']; | |
| 3916 | + continue; | |
| 3917 | + } | |
| 3918 | + | |
| 3919 | + $sitemap_urls[] = $site_url . $sitemap['url']; | |
| 3057 | 3920 | } |
| 3058 | 3921 | } |
| 3059 | 3922 | |
| 3923 | + if ($index_url !== '') { | |
| 3924 | + // The index covers the children and, on a segmented install, | |
| 3925 | + // the local business sitemap too. | |
| 3926 | + // | |
| 3927 | + // It does not cover a sitemap contributed through | |
| 3928 | + // `thinkrank_additional_sitemaps`: the index is built by this | |
| 3929 | + // plugin's own generator and never lists them. Returning the | |
| 3930 | + // index alone therefore left a contributed sitemap with no | |
| 3931 | + // discovery path at all — absent from robots.txt and absent | |
| 3932 | + // from the index — so Pro's News sitemap was unreachable on any | |
| 3933 | + // install with the index enabled, which is the default (#835). | |
| 3934 | + $contributed = []; | |
| 3935 | + | |
| 3936 | + foreach (\ThinkRank\SEO\Sitemap_Generator::additional_sitemaps() as $path) { | |
| 3937 | + $url = home_url($path); | |
| 3938 | + | |
| 3939 | + if ($url !== $index_url && !in_array($url, $contributed, true)) { | |
| 3940 | + $contributed[] = $url; | |
| 3941 | + } | |
| 3942 | + } | |
| 3943 | + | |
| 3944 | + return array_merge([$index_url], $contributed); | |
| 3945 | + } | |
| 3946 | + | |
| 3060 | 3947 | // Fallback to default if no URLs found |
| 3061 | 3948 | if (empty($sitemap_urls)) { |
| 3062 | 3949 | $sitemap_urls[] = home_url('/sitemap.xml'); |
| 3063 | 3950 | } |
| 3064 | 3951 | |
| 3065 | - // Advertise the local business sitemap when it exists. In segmented | |
| 3066 | - // mode it is already listed inside the sitemap index; in single mode | |
| 3067 | - // there is no index, so robots.txt is its discovery path. | |
| 3068 | - if (file_exists(ABSPATH . 'local-sitemap.xml')) { | |
| 3069 | - $local_url = home_url('/local-sitemap.xml'); | |
| 3070 | - if (!in_array($local_url, $sitemap_urls, true)) { | |
| 3071 | - $sitemap_urls[] = $local_url; | |
| 3952 | + // No index on this install, so anything not already listed above has | |
| 3953 | + // no other discovery path — advertise it directly. The local | |
| 3954 | + // business sitemap and the sitemaps other plugins register both land | |
| 3955 | + // here for the same reason, so they go through one list (#104). | |
| 3956 | + $extra = []; | |
| 3957 | + | |
| 3958 | + // Not a file test. Under dynamic delivery the local sitemap is | |
| 3959 | + // served from PHP and no file is ever written, so file_exists() | |
| 3960 | + // silently dropped a sitemap the site really does publish (#752). | |
| 3961 | + // On static sites the file is still what proves it, so both count. | |
| 3962 | + $local_sitemap_published = file_exists(ABSPATH . 'local-sitemap.xml'); | |
| 3963 | + | |
| 3964 | + if (!$local_sitemap_published && class_exists('ThinkRank\\SEO\\Sitemap_Generator')) { | |
| 3965 | + $generator = new \ThinkRank\SEO\Sitemap_Generator(false); | |
| 3966 | + | |
| 3967 | + $local_sitemap_published = 'dynamic' === $generator->resolve_delivery_mode() | |
| 3968 | + && $generator->publishes_local_sitemap(); | |
| 3969 | + } | |
| 3970 | + | |
| 3971 | + if ($local_sitemap_published) { | |
| 3972 | + $extra[] = '/local-sitemap.xml'; | |
| 3973 | + } | |
| 3974 | + | |
| 3975 | + foreach (\ThinkRank\SEO\Sitemap_Generator::additional_sitemaps() as $path) { | |
| 3976 | + $extra[] = $path; | |
| 3977 | + } | |
| 3978 | + | |
| 3979 | + foreach ($extra as $path) { | |
| 3980 | + $url = home_url($path); | |
| 3981 | + if (!in_array($url, $sitemap_urls, true)) { | |
| 3982 | + $sitemap_urls[] = $url; | |
| 3072 | 3983 | } |
| 3073 | 3984 | } |
| 3074 | 3985 | |
| 3075 | 3986 | return $sitemap_urls; |
| @@ -3159,8 +4070,9 @@ | ||
| 3159 | 4070 | // timestamp on every update. |
| 3160 | 4071 | $content = ''; |
| 3161 | 4072 | |
| 3162 | 4073 | $current_user_agent = ''; |
| 4074 | + $sitemap_started = false; | |
| 3163 | 4075 | |
| 3164 | 4076 | foreach ($rules as $rule) { |
| 3165 | 4077 | $directive = $rule['directive'] ?? ''; |
| 3166 | 4078 | $value = $rule['value'] ?? ''; |
| @@ -3181,9 +4093,17 @@ | ||
| 3181 | 4093 | case 'crawl_delay': |
| 3182 | 4094 | $content .= "Crawl-delay: {$value}\n"; |
| 3183 | 4095 | break; |
| 3184 | 4096 | case 'sitemap': |
| 3185 | - $content .= "\nSitemap: {$value}\n"; | |
| 4097 | + // One blank line separates the Sitemap block from the | |
| 4098 | + // preceding group, and none appear inside it. A blank line | |
| 4099 | + // terminates a record in the robots.txt grammar, so putting | |
| 4100 | + // one between every directive was invalid formatting. | |
| 4101 | + if (!$sitemap_started) { | |
| 4102 | + $content .= "\n"; | |
| 4103 | + $sitemap_started = true; | |
| 4104 | + } | |
| 4105 | + $content .= "Sitemap: {$value}\n"; | |
| 3186 | 4106 | break; |
| 3187 | 4107 | } |
| 3188 | 4108 | } |
| 3189 | 4109 | |
| @@ -3190,8 +4110,171 @@ | ||
| 3190 | 4110 | return ltrim($content, "\n"); |
| 3191 | 4111 | } |
| 3192 | 4112 | |
| 3193 | 4113 | /** |
| 4114 | + * Parse a robots.txt body back into the {directive, value} rule shape. | |
| 4115 | + * | |
| 4116 | + * generate_robots_txt() returns `rules` alongside `content`, but callers | |
| 4117 | + * replace `content` with the body actually being served (a stored override | |
| 4118 | + * or a physical file). The generated rules then described something the | |
| 4119 | + * response no longer contained. Re-deriving them from the served body keeps | |
| 4120 | + * the two halves of the payload describing the same document. | |
| 4121 | + * | |
| 4122 | + * @since 2.0.1 | |
| 4123 | + * | |
| 4124 | + * @param string $content Robots.txt body (header optional). | |
| 4125 | + * @return array<int, array{directive: string, value: string}> Parsed rules. | |
| 4126 | + */ | |
| 4127 | + public function parse_robots_txt_rules(string $content): array { | |
| 4128 | + $map = [ | |
| 4129 | + 'user-agent' => 'user_agent', | |
| 4130 | + 'disallow' => 'disallow', | |
| 4131 | + 'allow' => 'allow', | |
| 4132 | + 'crawl-delay' => 'crawl_delay', | |
| 4133 | + 'sitemap' => 'sitemap', | |
| 4134 | + ]; | |
| 4135 | + | |
| 4136 | + $rules = []; | |
| 4137 | + | |
| 4138 | + foreach (preg_split('/\r\n|\r|\n/', $this->strip_robots_header($content)) as $line) { | |
| 4139 | + $line = trim($line); | |
| 4140 | + | |
| 4141 | + // Blank lines separate groups and `#` starts a comment; neither is | |
| 4142 | + // a rule. | |
| 4143 | + if ($line === '' || str_starts_with($line, '#')) { | |
| 4144 | + continue; | |
| 4145 | + } | |
| 4146 | + | |
| 4147 | + $parts = explode(':', $line, 2); | |
| 4148 | + if (count($parts) !== 2) { | |
| 4149 | + continue; | |
| 4150 | + } | |
| 4151 | + | |
| 4152 | + $field = strtolower(trim($parts[0])); | |
| 4153 | + if (!isset($map[$field])) { | |
| 4154 | + continue; | |
| 4155 | + } | |
| 4156 | + | |
| 4157 | + $rules[] = [ | |
| 4158 | + 'directive' => $map[$field], | |
| 4159 | + // Sitemap values are absolute URLs and contain the `:` the | |
| 4160 | + // limited explode above deliberately preserved. | |
| 4161 | + 'value' => trim($parts[1]), | |
| 4162 | + ]; | |
| 4163 | + } | |
| 4164 | + | |
| 4165 | + return $rules; | |
| 4166 | + } | |
| 4167 | + | |
| 4168 | + /** | |
| 4169 | + * Opening fence of the machine-owned AI crawler region. | |
| 4170 | + * | |
| 4171 | + * @since 2.5.0 | |
| 4172 | + * @var string | |
| 4173 | + */ | |
| 4174 | + public const AI_BLOCK_BEGIN = '# BEGIN ThinkRank AI crawlers'; | |
| 4175 | + | |
| 4176 | + /** | |
| 4177 | + * Closing fence of the machine-owned AI crawler region. | |
| 4178 | + * | |
| 4179 | + * @since 2.5.0 | |
| 4180 | + * @var string | |
| 4181 | + */ | |
| 4182 | + public const AI_BLOCK_END = '# END ThinkRank AI crawlers'; | |
| 4183 | + | |
| 4184 | + /** | |
| 4185 | + * Render the fenced AI crawler region for the current settings. | |
| 4186 | + * | |
| 4187 | + * One `User-agent:` / `Disallow: /` record per blocked crawler. Allowed | |
| 4188 | + * crawlers emit nothing at all: `Disallow:` with an empty value is the | |
| 4189 | + * robots.txt way of saying "allow everything", but writing eighteen such | |
| 4190 | + * records to say what silence already says would triple the file and | |
| 4191 | + * invite the reading that an unlisted crawler is therefore refused. | |
| 4192 | + * | |
| 4193 | + * @since 2.5.0 | |
| 4194 | + * | |
| 4195 | + * @param array $settings Site settings. | |
| 4196 | + * @return string Fenced block, newline-terminated, or '' when nothing is blocked. | |
| 4197 | + */ | |
| 4198 | + private function build_ai_crawler_block(array $settings): string { | |
| 4199 | + $blocked = AI_Crawlers::blocked_slugs($settings['ai_crawler_rules'] ?? []); | |
| 4200 | + | |
| 4201 | + if (empty($blocked)) { | |
| 4202 | + return ''; | |
| 4203 | + } | |
| 4204 | + | |
| 4205 | + $agents = AI_Crawlers::all(); | |
| 4206 | + | |
| 4207 | + $lines = [ | |
| 4208 | + self::AI_BLOCK_BEGIN, | |
| 4209 | + '# Managed by ThinkRank — edits between these lines are overwritten.', | |
| 4210 | + ]; | |
| 4211 | + | |
| 4212 | + foreach ($blocked as $slug) { | |
| 4213 | + $lines[] = ''; | |
| 4214 | + $lines[] = 'User-agent: ' . $agents[$slug]['token']; | |
| 4215 | + $lines[] = 'Disallow: /'; | |
| 4216 | + } | |
| 4217 | + | |
| 4218 | + $lines[] = self::AI_BLOCK_END; | |
| 4219 | + | |
| 4220 | + return implode("\n", $lines) . "\n"; | |
| 4221 | + } | |
| 4222 | + | |
| 4223 | + /** | |
| 4224 | + * Remove the fenced AI crawler region from a robots.txt body. | |
| 4225 | + * | |
| 4226 | + * Tolerates a missing closing fence rather than leaving the rest of the | |
| 4227 | + * file swallowed: a truncated write, or someone deleting the END line by | |
| 4228 | + * hand, would otherwise make every subsequent read drop everything below | |
| 4229 | + * the opening fence. | |
| 4230 | + * | |
| 4231 | + * @since 2.5.0 | |
| 4232 | + * | |
| 4233 | + * @param string $body Robots.txt body. | |
| 4234 | + * @return string Body with the region removed. | |
| 4235 | + */ | |
| 4236 | + public function strip_ai_crawler_block(string $body): string { | |
| 4237 | + if (false === strpos($body, self::AI_BLOCK_BEGIN)) { | |
| 4238 | + return $body; | |
| 4239 | + } | |
| 4240 | + | |
| 4241 | + $pattern = '/\R*' . preg_quote(self::AI_BLOCK_BEGIN, '/') | |
| 4242 | + . '.*?(?:' . preg_quote(self::AI_BLOCK_END, '/') . '|\z)\R*/s'; | |
| 4243 | + | |
| 4244 | + return trim((string) preg_replace($pattern, "\n\n", $body, 1)); | |
| 4245 | + } | |
| 4246 | + | |
| 4247 | + /** | |
| 4248 | + * Put the current AI crawler region into a robots.txt body. | |
| 4249 | + * | |
| 4250 | + * Replaces an existing region in place so the block keeps its position in | |
| 4251 | + * a hand-ordered file, and appends when there is none. Everything outside | |
| 4252 | + * the fences is returned untouched — that is the whole point of fencing | |
| 4253 | + * it, since the body is also a free-text field the user edits. | |
| 4254 | + * | |
| 4255 | + * @since 2.5.0 | |
| 4256 | + * | |
| 4257 | + * @param string $body Robots.txt body (fences optional). | |
| 4258 | + * @param array $settings Site settings. | |
| 4259 | + * @return string Body carrying the current region. | |
| 4260 | + */ | |
| 4261 | + private function apply_ai_crawler_block(string $body, array $settings): string { | |
| 4262 | + $stripped = $this->strip_ai_crawler_block($body); | |
| 4263 | + $block = $this->build_ai_crawler_block($settings); | |
| 4264 | + | |
| 4265 | + if ('' === $block) { | |
| 4266 | + return $stripped; | |
| 4267 | + } | |
| 4268 | + | |
| 4269 | + if ('' === trim($stripped)) { | |
| 4270 | + return trim($block); | |
| 4271 | + } | |
| 4272 | + | |
| 4273 | + return rtrim($stripped) . "\n\n" . trim($block); | |
| 4274 | + } | |
| 4275 | + | |
| 4276 | + /** | |
| 3194 | 4277 | * The auto-generated header prepended to the served robots.txt. |
| 3195 | 4278 | * |
| 3196 | 4279 | * Kept separate from the body so it is only ever added at render time with |
| 3197 | 4280 | * a fresh timestamp, never stored or shown in the editable textarea. |
| @@ -3228,22 +4311,28 @@ | ||
| 3228 | 4311 | */ |
| 3229 | 4312 | public function get_served_robots_body(): string { |
| 3230 | 4313 | $settings = $this->get_settings('site'); |
| 3231 | 4314 | |
| 4315 | + // The AI block is stripped from every one of these paths. A physical | |
| 4316 | + // robots.txt we wrote carries it, and the stored override is whatever | |
| 4317 | + // the textarea last held — so without this the block round-trips into | |
| 4318 | + // the editor, gets saved as ordinary body text, and is then appended | |
| 4319 | + // to a second time on the next render. | |
| 3232 | 4320 | $custom = trim((string) ($settings['robots_txt_content'] ?? '')); |
| 3233 | 4321 | if ($custom !== '') { |
| 3234 | - return $this->strip_robots_header($custom); | |
| 4322 | + return $this->strip_ai_crawler_block($this->strip_robots_header($custom)); | |
| 3235 | 4323 | } |
| 3236 | 4324 | |
| 3237 | 4325 | $robots_file = ABSPATH . 'robots.txt'; |
| 3238 | 4326 | if (file_exists($robots_file)) { |
| 4327 | + // 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. | |
| 3239 | 4328 | $raw = (string) @file_get_contents($robots_file); |
| 3240 | 4329 | if ($raw !== '') { |
| 3241 | - return $this->strip_robots_header($raw); | |
| 4330 | + return $this->strip_ai_crawler_block($this->strip_robots_header($raw)); | |
| 3242 | 4331 | } |
| 3243 | 4332 | } |
| 3244 | 4333 | |
| 3245 | - return trim($this->generate_robots_txt()['content']); | |
| 4334 | + return $this->strip_ai_crawler_block(trim($this->generate_robots_txt()['content'])); | |
| 3246 | 4335 | } |
| 3247 | 4336 | private function get_site_identity_data(array $settings): array { |
| 3248 | 4337 | return [ |
| 3249 | 4338 | 'site_name' => $settings['site_name'] ?? get_bloginfo('name'), |
| @@ -3305,11 +4394,14 @@ | ||
| 3305 | 4394 | $optimization['validation']['valid'] = false; |
| 3306 | 4395 | } |
| 3307 | 4396 | |
| 3308 | 4397 | if (!empty($value) && isset($config['max_length'])) { |
| 3309 | - if (strlen($value) > $config['max_length']) { | |
| 4398 | + // The warning says "characters", so measure and cut in characters: | |
| 4399 | + // strlen()/substr() fired early on non-Latin values and the | |
| 4400 | + // suggested replacement was cut mid-character (#687). | |
| 4401 | + if (mb_strlen($value) > $config['max_length']) { | |
| 3310 | 4402 | $optimization['validation']['warnings'][] = "{$element} exceeds maximum length of {$config['max_length']} characters"; |
| 3311 | - $optimization['optimized_value'] = substr($value, 0, $config['max_length']); | |
| 4403 | + $optimization['optimized_value'] = \ThinkRank\Core\Seo_Text::trim_to_length($value, (int) $config['max_length']); | |
| 3312 | 4404 | } |
| 3313 | 4405 | } |
| 3314 | 4406 | |
| 3315 | 4407 | // SEO-specific optimizations |
| @@ -3348,18 +4440,20 @@ | ||
| 3348 | 4440 | return $optimization; |
| 3349 | 4441 | } |
| 3350 | 4442 | |
| 3351 | 4443 | // Check if it's a local image |
| 3352 | - $attachment_id = attachment_url_to_postid($value); | |
| 4444 | + $attachment_id = Attachment_Lookup::id_from_url($value); | |
| 3353 | 4445 | if ($attachment_id) { |
| 3354 | 4446 | $image_meta = wp_get_attachment_metadata($attachment_id); |
| 3355 | 4447 | |
| 3356 | 4448 | if ($image_meta && isset($image_meta['width'], $image_meta['height'])) { |
| 3357 | - // Check recommended size | |
| 4449 | + // Check recommended size, against the configured file itself | |
| 4450 | + // rather than the upload it may have been generated from. | |
| 3358 | 4451 | if (isset($config['recommended_size'])) { |
| 3359 | 4452 | [$rec_width, $rec_height] = explode('x', $config['recommended_size']); |
| 4453 | + $image_file = Attachment_Lookup::describe($attachment_id, $value); | |
| 3360 | 4454 | |
| 3361 | - if ($image_meta['width'] != $rec_width || $image_meta['height'] != $rec_height) { | |
| 4455 | + if ($image_file['width'] !== (int) $rec_width || $image_file['height'] !== (int) $rec_height) { | |
| 3362 | 4456 | $optimization['suggestions'][] = "Consider using {$config['recommended_size']} size for optimal {$element}"; |
| 3363 | 4457 | } |
| 3364 | 4458 | } |
| 3365 | 4459 | |
