← All changes | includes/seo/class-sitemap-generator.php +3259 -328 1.0.0 → 2.11.0 View file →

		@@ -14,8 +14,20 @@
14	14	declare(strict_types=1);
15	15
16	16	namespace ThinkRank\SEO;
17	17
	18	+use InvalidArgumentException;
	19	+
	20	+// Prevent direct access.
	21	+if ( ! defined( 'ABSPATH' ) ) {
	22	+ exit;
	23	+}
	24	+
	25	+// Filename derivation and web-root removal are shared with the deactivator and
	26	+// with uninstall.php, which runs without an autoloader — so they live in a
	27	+// plain function file both can require. See includes/cleanup-webroot.php.
	28	+require_once __DIR__ . '/../cleanup-webroot.php';
	29	+
18	30	/**
19	31	* Sitemap Generator Class
20	32	*
21	33	* Generates and manages XML sitemaps for search engine optimization.
		@@ -26,13 +38,246 @@
26	38	*/
27	39	class Sitemap_Generator extends Abstract_SEO_Manager {
28	40
29	41	/**
	42	+ * Memoised WooCommerce page IDs kept out of the sitemap. Null until resolved.
	43	+ *
	44	+ * @since 2.0.1
	45	+ * @var int[]|null
	46	+ */
	47	+ private ?array $woocommerce_excluded_page_ids = null;
	48	+
	49	+ /**
	50	+ * Inclusion flag -> the child sitemap it controls.
	51	+ *
	52	+ * Shared by the child-list builder and by the "did the caller change what
	53	+ * the sitemap includes?" check in maybe_promote_to_index().
	54	+ *
	55	+ * @since 1.31.0
	56	+ * @var array<string, string>
	57	+ */
	58	+ private const INCLUSION_CHILD_TYPES = [
	59	+ 'include_posts' => 'posts',
	60	+ 'include_pages' => 'pages',
	61	+ 'include_categories' => 'categories',
	62	+ 'include_tags' => 'tags',
	63	+ ];
	64	+
	65	+ /**
	66	+ * Child sitemap type -> the object that actually supplies its entries.
	67	+ *
	68	+ * A child normally carries its object's own slug, but the sitemap presets UI
	69	+ * writes a display name for the WooCommerce taxonomy child
	70	+ * ('product_categories', not 'product_cat'), so a saved child list can name a
	71	+ * type no post type or taxonomy answers to. Resolving through here lets those
	72	+ * children take the same generic path as every other custom taxonomy instead
	73	+ * of needing a case of their own (#690).
	74	+ *
	75	+ * @since 2.7.0
	76	+ * @var array<string, string>
	77	+ */
	78	+ private const CHILD_TYPE_ALIASES = [
	79	+ 'product_categories' => 'product_cat',
	80	+ ];
	81	+
	82	+ /**
30	83	* Supported sitemap types
31	84	*
32	85	* @since 1.0.0
33	86	* @var array
34	87	*/
	88	+ /**
	89	+ * How many post IDs to hydrate at a time while walking the sitemap set.
	90	+ *
	91	+ * @since 2.0.1
	92	+ * @var int
	93	+ */
	94	+ private const ID_WALK_CHUNK = 500;
	95	+
	96	+ /**
	97	+ * How long a content or settings change is debounced before the sitemap is
	98	+ * rebuilt, in seconds. Coalesces bulk edits into a single regeneration.
	99	+ *
	100	+ * @since 2.2.1
	101	+ * @var int
	102	+ */
	103	+ private const REGENERATION_DEBOUNCE = 30;
	104	+
	105	+ /**
	106	+ * How far past its due time the scheduled rebuild may sit before a request
	107	+ * takes it over, in seconds.
	108	+ *
	109	+ * WP-Cron is request-driven, so on a site running DISABLE_WP_CRON, blocking
	110	+ * loopback requests, or seeing very little traffic the event never fires and
	111	+ * the sitemap silently stops updating (#629). The grace keeps the fast path
	112	+ * (cron) in charge under normal conditions.
	113	+ *
	114	+ * @since 2.2.1
	115	+ * @var int
	116	+ */
	117	+ private const REGENERATION_TAKEOVER_GRACE = 120;
	118	+
	119	+ /**
	120	+ * Longest backoff between takeover attempts after a failed rebuild, so a
	121	+ * persistently failing generation cannot run on every admin request.
	122	+ *
	123	+ * @since 2.2.1
	124	+ * @var int
	125	+ */
	126	+ private const REGENERATION_MAX_BACKOFF = 3600;
	127	+
	128	+ /**
	129	+ * Option holding the rebuild that content/settings changes are still waiting
	130	+ * on: `since`, `source` ('content'|'settings'), `attempts`, `next_attempt`.
	131	+ * Absent means the served sitemap is up to date with what triggered it.
	132	+ *
	133	+ * @since 2.2.1
	134	+ * @var string
	135	+ */
	136	+ public const REGENERATION_PENDING_OPTION = 'thinkrank_sitemap_regeneration_pending';
	137	+
	138	+ /**
	139	+ * Option holding the last automatic-regeneration failure (`message`,
	140	+ * `source`, `time`), so the failure is visible instead of swallowed.
	141	+ *
	142	+ * @since 2.2.1
	143	+ * @var string
	144	+ */
	145	+ public const REGENERATION_ERROR_OPTION = 'thinkrank_sitemap_regeneration_error';
	146	+
	147	+ /**
	148	+ * Delivery modes accepted by the `delivery_mode` setting.
	149	+ *
	150	+ * Mirrors LLMs_Txt_Manager::DELIVERY_MODES, which solved the same problem
	151	+ * for llms.txt. `auto` is the only value a site should normally need.
	152	+ *
	153	+ * @since 2.9.0
	154	+ * @var string[]
	155	+ */
	156	+ public const DELIVERY_MODES = ['auto', 'static', 'dynamic'];
	157	+
	158	+ /**
	159	+ * Cache group for documents rendered on the dynamic path.
	160	+ *
	161	+ * @since 2.9.0
	162	+ * @var string
	163	+ */
	164	+ private const DYNAMIC_CACHE_PREFIX = 'thinkrank_sitemap_doc_';
	165	+
	166	+ /**
	167	+ * How long a dynamically rendered document is cached.
	168	+ *
	169	+ * Invalidated by content and settings changes through
	170	+ * {@see self::flush_dynamic_cache()}, so this is only the backstop for a
	171	+ * change nothing hooked.
	172	+ *
	173	+ * @since 2.9.0
	174	+ * @var int
	175	+ */
	176	+ private const DYNAMIC_CACHE_TTL = 12 * HOUR_IN_SECONDS;
	177	+
	178	+ /**
	179	+ * How long the render lock is held before it is assumed abandoned.
	180	+ *
	181	+ * Long enough for a large site's full build, short enough that a request
	182	+ * killed mid-build does not lock the endpoint out for meaningfully long.
	183	+ *
	184	+ * @since 2.9.0
	185	+ * @var int
	186	+ */
	187	+ private const RENDER_LOCK_TTL = 60;
	188	+
	189	+ /**
	190	+ * Cached stand-in for "this site does not publish that name".
	191	+ *
	192	+ * published_document_names() lists what the configuration *could* produce,
	193	+ * but a child whose type is excluded produces nothing. Without a negative
	194	+ * entry those names miss the cache forever, so every request for one
	195	+ * rebuilt the entire sitemap — the same cost the positive cache exists to
	196	+ * avoid, on a public endpoint (#754 review).
	197	+ *
	198	+ * @since 2.9.0
	199	+ * @var string
	200	+ */
	201	+ private const ABSENT_MARKER = "\0thinkrank-absent";
	202	+
	203	+ /**
	204	+ * How many times, and how long, a losing request waits for the winner.
	205	+ *
	206	+ * Bounded at roughly a second in total: past that, building a second copy
	207	+ * costs less than making a crawler wait.
	208	+ *
	209	+ * @since 2.9.0
	210	+ * @var int
	211	+ */
	212	+ private const RENDER_LOCK_WAIT_ATTEMPTS = 4;
	213	+
	214	+ /**
	215	+ * @since 2.9.0
	216	+ * @var int
	217	+ */
	218	+ private const RENDER_LOCK_WAIT_MICROSECONDS = 250000;
	219	+
	220	+ /**
	221	+ * Where generated documents go instead of disk, when set.
	222	+ *
	223	+ * Every sitemap document this class produces — segments, the index, the
	224	+ * single flat file and local-sitemap.xml — is published through the one
	225	+ * writer, {@see self::save_sitemap_to_file()}. Swapping that writer for a
	226	+ * collector is therefore all it takes to render the same bytes without a
	227	+ * filesystem, which is what dynamic delivery needs (#752). Doing it here
	228	+ * rather than duplicating the build pipeline is deliberate: a second
	229	+ * pipeline would drift from this one, and the index in particular is
	230	+ * assembled from whatever the children actually produced.
	231	+ *
	232	+ * @since 2.9.0
	233	+ * @var callable|null
	234	+ */
	235	+ private $document_sink = null;
	236	+
	237	+ /**
	238	+ * Transient guarding against two generations running at once. Shared with
	239	+ * Sitemap_Endpoint's manual generate route so an automatic rebuild and a
	240	+ * manual one cannot write the same files concurrently.
	241	+ *
	242	+ * @since 2.2.1
	243	+ * @var string
	244	+ */
	245	+ public const GENERATION_LOCK_TRANSIENT = 'thinkrank_sitemap_generation_lock';
	246	+
	247	+ /**
	248	+ * How many term IDs to hydrate at a time while walking a taxonomy.
	249	+ *
	250	+ * @since 2.0.1
	251	+ * @var int
	252	+ */
	253	+ private const TERM_WALK_CHUNK = 1000;
	254	+
	255	+ /**
	256	+ * Exception code for an automatic rebuild stopped short of the memory limit.
	257	+ *
	258	+ * @since 2.10.1
	259	+ * @var int
	260	+ */
	261	+ private const MEMORY_ABORT_CODE = 4290;
	262	+
	263	+ /**
	264	+ * Whether the chunked walks should stop before the memory limit. Only on
	265	+ * for automatic rebuilds, whose failure is recorded and retried.
	266	+ *
	267	+ * @since 2.10.1
	268	+ * @var bool
	269	+ */
	270	+ private bool $memory_guard = false;
	271	+
	272	+ /**
	273	+ * Largest memory cost of one walked chunk in this rebuild, in bytes.
	274	+ *
	275	+ * @since 2.10.1
	276	+ * @var int
	277	+ */
	278	+ private int $walk_chunk_cost = 0;
	279	+
35	280	private array $sitemap_types = [
36	281	'posts' => [
37	282	'name' => 'Posts',
38	283	'post_types' => ['post'],
		@@ -59,39 +304,123 @@
59	304	]
60	305	];
61	306
62	307	/**
	308	+ * The generator the content-change listeners share, built on the first
	309	+ * change of a request.
	310	+ *
	311	+ * @since 2.10.1
	312	+ * @var self|null
	313	+ */
	314	+ private static ?self $listener = null;
	315	+
	316	+ /**
63	317	* Constructor
64	318	*
65	319	* @since 1.0.0
	320	+ * @since 2.10.1 Registers no hooks, whatever `$register_hooks` says. The
	321	+ * content-change listeners are registered once, at bootstrap,
	322	+ * by register_content_listeners().
	323	+ *
	324	+ * @param bool $register_hooks Unused since 2.10.1; kept so existing callers,
	325	+ * Pro's included, keep working.
66	326	*/
67		- public function __construct() {
	327	+ public function __construct(bool $register_hooks = true) {
68	328	parent::__construct('sitemap');
	329	+ }
69	330
70		- // Initialize auto-generation hooks
71		- $this->init_auto_generation_hooks();
	331	+ /**
	332	+ * Filter the args of a sitemap post query.
	333	+ *
	334	+ * Exists so integrations can widen what the sitemap sees — the multilingual
	335	+ * manager uses it to include every language, since these queries otherwise
	336	+ * run in whichever language happened to be active at generation time.
	337	+ *
	338	+ * @since 1.23.0
	339	+ *
	340	+ * @param array $args get_posts() arguments.
	341	+ * @return array Filtered arguments.
	342	+ */
	343	+ private function filter_query_args(array $args): array {
	344	+ /**
	345	+ * Filter the arguments of a sitemap post query.
	346	+ *
	347	+ * @since 1.23.0
	348	+ *
	349	+ * @param array $args get_posts() arguments.
	350	+ */
	351	+ return (array) apply_filters('thinkrank_sitemap_query_args', $args);
72	352	}
73	353
74	354	/**
75		- * Initialize WordPress hooks for auto-generation
	355	+ * Filter the args of a sitemap term query.
76	356	*
77		- * @since 1.0.0
	357	+ * @since 1.23.0
	358	+ *
	359	+ * @param array $args get_terms() arguments.
	360	+ * @return array Filtered arguments.
	361	+ */
	362	+ private function filter_term_query_args(array $args): array {
	363	+ /**
	364	+ * Filter the arguments of a sitemap term query.
	365	+ *
	366	+ * @since 1.23.0
	367	+ *
	368	+ * @param array $args get_terms() arguments.
	369	+ */
	370	+ return (array) apply_filters('thinkrank_sitemap_term_query_args', $args);
	371	+ }
	372	+
	373	+ /**
	374	+ * Register the content-change listeners that queue an automatic rebuild.
	375	+ *
	376	+ * Called once per request, at plugin bootstrap, next to the WP-Cron
	377	+ * listeners that run the rebuild these queue (both in
	378	+ * Plugin::register_sitemap_cron_listeners(), on plugins_loaded). The
	379	+ * constructor used to register them, so they existed only in a request
	380	+ * that happened to build a generator: the REST endpoint, the setup wizard,
	381	+ * an MCP ability. Block-editor saves go through REST and were heard. A
	382	+ * scheduled post published by WP-Cron, Quick Edit, the classic editor and
	383	+ * WP-CLI were not, so the post stayed out of the sitemap with nothing
	384	+ * pending to recover it (#824). Each instance also added its own set, so
	385	+ * one REST save ran the handlers once per generator built.
	386	+ *
	387	+ * The generator is built on the first change and shared for the rest of
	388	+ * the request, so a bulk edit does not construct one per post.
	389	+ *
	390	+ * @since 2.10.1
78	391	* @return void
79	392	*/
80		- private function init_auto_generation_hooks(): void {
81		- // Content change hooks - use priority 20 to run after other plugins
82		- add_action('save_post', [$this, 'handle_content_change'], 20, 2);
83		- add_action('delete_post', [$this, 'handle_content_deletion'], 20);
84		- add_action('wp_trash_post', [$this, 'handle_content_deletion'], 20);
85		- add_action('untrash_post', [$this, 'handle_content_change_by_id'], 20);
	393	+ public static function register_content_listeners(): void {
	394	+ // Priority 20, to run after other plugins.
	395	+ add_action('save_post', static function (int $post_id, \WP_Post $post): void {
	396	+ self::listener()->handle_content_change($post_id, $post);
	397	+ }, 20, 2);
	398	+ add_action('delete_post', static function (int $post_id): void {
	399	+ self::listener()->handle_content_deletion($post_id);
	400	+ }, 20);
	401	+ add_action('wp_trash_post', static function (int $post_id): void {
	402	+ self::listener()->handle_content_deletion($post_id);
	403	+ }, 20);
	404	+ add_action('untrash_post', static function (int $post_id): void {
	405	+ self::listener()->handle_content_change_by_id($post_id);
	406	+ }, 20);
86	407
87		- // Taxonomy change hooks
88		- add_action('created_term', [$this, 'handle_taxonomy_change'], 20, 3);
89		- add_action('edited_term', [$this, 'handle_taxonomy_change'], 20, 3);
90		- add_action('delete_term', [$this, 'handle_taxonomy_change'], 20, 3);
	408	+ foreach (['created_term', 'edited_term', 'delete_term'] as $hook) {
	409	+ add_action($hook, static function (int $term_id, int $tt_id, string $taxonomy): void {
	410	+ self::listener()->handle_taxonomy_change($term_id, $tt_id, $taxonomy);
	411	+ }, 20, 3);
	412	+ }
	413	+ }
91	414
92		- // Debounced regeneration to prevent multiple rapid-fire generations
93		- add_action('thinkrank_regenerate_sitemap', [$this, 'auto_regenerate_sitemap']);
	415	+ /**
	416	+ * The generator the content-change listeners share.
	417	+ *
	418	+ * @since 2.10.1
	419	+ * @return self
	420	+ */
	421	+ private static function listener(): self {
	422	+ return self::$listener ??= new self(false);
94	423	}
95	424
96	425	/**
97	426	* Generate XML sitemap
		@@ -103,15 +432,10 @@
103	432	*/
104	433	public function generate_sitemap(array $options = []): string {
105	434	$settings = $this->get_settings('site');
106	435
107		- $xml = '<?xml version="1.0" encoding="UTF-8"?>' . "\n";
	436	+ $xml = $this->xml_prolog($settings, 'sitemap');
108	437
109		- // Add XSL stylesheet only if styling is enabled
110		- if (!empty($settings['enable_styling'])) {
111		- $xml .= '<?xml-stylesheet type="text/xsl" href="' . home_url('/wp-content/plugins/thinkrank/assets/sitemap.xsl') . '"?>' . "\n";
112		- }
113		-
114	438	// Add image namespace if images are enabled
115	439	if (!empty($settings['include_images'])) {
116	440	$xml .= '<urlset xmlns="http://www.sitemaps.org/schemas/sitemap/0.9" xmlns:image="http://www.google.com/schemas/sitemap-image/1.1">' . "\n";
117	441	} else {
		@@ -117,15 +441,19 @@
117	441	} else {
118	442	$xml .= '<urlset xmlns="http://www.sitemaps.org/schemas/sitemap/0.9">' . "\n";
119	443	}
120	444
121		- // Add homepage
122		- $xml .= $this->generate_url_entry(home_url('/'), gmdate('c'), 1.0, 'daily');
	445	+ // Add homepage. Use the static front page's real modified time when set,
	446	+ // so lastmod reflects real changes rather than the generation time.
	447	+ $xml .= $this->generate_url_entry(home_url('/'), $this->get_homepage_lastmod(), 1.0, 'daily');
123	448
124		- // MEMORY-EFFICIENT: Generate posts using chunked processing
	449	+ // MEMORY-EFFICIENT: Generate posts using chunked processing.
	450	+ // Note: the single "general" sitemap includes every URL (no per-file cap);
	451	+ // per-file chunking applies to the segmented/index model where each type
	452	+ // gets its own paginated file (see generate_multiple_sitemaps).
125	453	$enabled_post_types = $this->get_enabled_post_types($settings);
126	454	if (!empty($enabled_post_types)) {
127		- $xml .= $this->generate_posts_chunked($enabled_post_types, $settings);
	455	+ $xml .= implode('', $this->collect_post_entries($enabled_post_types, $settings));
128	456	}
129	457
130	458	// OPTIMIZED: Get all enabled taxonomies and fetch in single query
131	459	$enabled_taxonomies = $this->get_enabled_taxonomies($settings);
		@@ -219,8 +547,34 @@
219	547	];
220	548	}
221	549
222	550	/**
	551	+ * Sitemap keys outside the defaults.
	552	+ *
	553	+ * @since 2.0.1
	554	+ *
	555	+ * @return string[]
	556	+ */
	557	+ protected function additional_setting_keys(): array {
	558	+ return ['selected_preset'];
	559	+ }
	560	+
	561	+ /**
	562	+ * Inclusion flags are per post type and per taxonomy.
	563	+ *
	564	+ * A site registering a `product` post type stores `include_product`; an
	565	+ * enumerated list would go stale on the next registration, so the family
	566	+ * is matched instead.
	567	+ *
	568	+ * @since 2.0.1
	569	+ *
	570	+ * @return string[]
	571	+ */
	572	+ protected function dynamic_setting_key_patterns(): array {
	573	+ return ['/^include_[a-z0-9_]+$/', '/^exclude_[a-z0-9_]+$/'];
	574	+ }
	575	+
	576	+ /**
223	577	* Get default settings for a context type (implements interface)
224	578	*
225	579	* @since 1.0.0
226	580	*
		@@ -243,8 +597,15 @@
243	597	]
244	598	],
245	599	'use_sitemap_index' => false,
246	600
	601	+ // How the sitemap reaches crawlers. 'auto' keeps the historical
	602	+ // behaviour wherever the web root is writable, and only falls back
	603	+ // to serving the sitemap from PHP where writing a file is
	604	+ // impossible — previously a hard failure with nothing served
	605	+ // (#752).
	606	+ 'delivery_mode' => 'auto',
	607	+
247	608	// General Settings
248	609	'links_per_sitemap' => 1000,
249	610	'include_images' => true,
250	611	'include_featured_images' => false,
		@@ -266,8 +627,17 @@
266	627	// Advanced Options
267	628	'enable_styling' => true,
268	629	'custom_url_pattern' => 'sitemap-{type}.xml',
269	630
	631	+ // Stylesheet branding (#639). Both colours default to empty, not
	632	+ // to the stock hexes: empty means the stylesheet's own value
	633	+ // stands, so a site that never opens this screen renders exactly
	634	+ // as it did before the setting existed.
	635	+ 'styling_logo' => false,
	636	+ 'styling_logo_url' => '',
	637	+ 'styling_color_main' => '',
	638	+ 'styling_color_accent' => '',
	639	+
270	640	// Generation tracking
271	641	'last_generated' => ''
272	642	];
273	643	}
		@@ -272,8 +642,49 @@
272	642	];
273	643	}
274	644
275	645	/**
	646	+ * Normalize the stylesheet branding values on the way into the store.
	647	+ *
	648	+ * The generic sanitizer only runs `sanitize_text_field()` over a string,
	649	+ * which happily keeps "red" or "rebeccapurple" as a colour. Nothing
	650	+ * downstream can use those — {@see Sitemap_Stylesheet::render()} skips any
	651	+ * value it cannot read as a hex colour — so storing them would report a
	652	+ * successful save of a setting that changes nothing, and `get-sitemap-
	653	+ * settings` would hand an agent back a colour the sitemap does not use.
	654	+ * Reducing here instead keeps the store and the rendering in agreement.
	655	+ *
	656	+ * @since 2.7.0
	657	+ *
	658	+ * @param array $settings Settings to sanitize.
	659	+ * @param string $context_type Context the save is for.
	660	+ * @return array
	661	+ */
	662	+ protected function sanitize_settings(array $settings, string $context_type = 'site'): array {
	663	+ $sanitized = parent::sanitize_settings($settings, $context_type);
	664	+
	665	+ foreach (['styling_color_main', 'styling_color_accent'] as $key) {
	666	+ if (array_key_exists($key, $sanitized)) {
	667	+ $sanitized[$key] = Sitemap_Stylesheet::hex($sanitized[$key]);
	668	+ }
	669	+ }
	670	+
	671	+ if (array_key_exists('styling_logo_url', $sanitized)) {
	672	+ $sanitized['styling_logo_url'] = esc_url_raw((string) $sanitized['styling_logo_url']);
	673	+ }
	674	+
	675	+ // A mode this build cannot act on has to be stored as the fallback
	676	+ // rather than kept verbatim, or get-sitemap-settings reports a delivery
	677	+ // mode the site does not actually apply.
	678	+ if (array_key_exists('delivery_mode', $sanitized)) {
	679	+ $mode = sanitize_key((string) $sanitized['delivery_mode']);
	680	+ $sanitized['delivery_mode'] = in_array($mode, self::DELIVERY_MODES, true) ? $mode : 'auto';
	681	+ }
	682	+
	683	+ return $sanitized;
	684	+ }
	685	+
	686	+ /**
276	687	* Get settings schema definition (implements interface)
277	688	*
278	689	* @since 1.0.0
279	690	*
		@@ -287,8 +698,15 @@
287	698	'title' => 'Enable Sitemap',
288	699	'description' => 'Generate XML sitemap for search engines',
289	700	'default' => true
290	701	],
	702	+ 'delivery_mode' => [
	703	+ 'type' => 'string',
	704	+ 'title' => 'Sitemap Delivery',
	705	+ 'description' => 'How the sitemap is served: auto picks static when the WordPress root is writable and dynamic when it is not, static writes files to the web root, dynamic serves the sitemap from WordPress with no files written',
	706	+ 'enum' => self::DELIVERY_MODES,
	707	+ 'default' => 'auto'
	708	+ ],
291	709	'include_posts' => [
292	710	'type' => 'boolean',
293	711	'title' => 'Include Posts',
294	712	'description' => 'Include blog posts in sitemap',
		@@ -329,8 +747,32 @@
329	747	'type' => 'string',
330	748	'title' => 'Last Generated',
331	749	'description' => 'Timestamp of last sitemap generation',
332	750	'default' => ''
	751	+ ],
	752	+ 'styling_logo' => [
	753	+ 'type' => 'boolean',
	754	+ 'title' => 'Show Logo On Sitemap',
	755	+ 'description' => 'Show a logo above the sitemap heading',
	756	+ 'default' => false
	757	+ ],
	758	+ 'styling_logo_url' => [
	759	+ 'type' => 'string',
	760	+ 'title' => 'Sitemap Logo',
	761	+ 'description' => 'Logo image URL. Empty falls back to the site icon',
	762	+ 'default' => ''
	763	+ ],
	764	+ 'styling_color_main' => [
	765	+ 'type' => 'string',
	766	+ 'title' => 'Sitemap Main Color',
	767	+ 'description' => 'Hex color for the sitemap header, links and table head. Empty keeps the stock palette',
	768	+ 'default' => ''
	769	+ ],
	770	+ 'styling_color_accent' => [
	771	+ 'type' => 'string',
	772	+ 'title' => 'Sitemap Accent Color',
	773	+ 'description' => 'Hex color for the header gradient and link hovers. Empty keeps the stock palette',
	774	+ 'default' => ''
333	775	]
334	776	];
335	777	}
336	778
		@@ -347,10 +789,19 @@
347	789	* @return string XML URL entry
348	790	*/
349	791	private function generate_url_entry(string $url, string $lastmod, float $priority, string $changefreq, array $images = []): string {
350	792	$xml = " <url>\n";
351		- $xml .= " <loc>" . esc_url($url) . "</loc>\n";
352		- $xml .= " <lastmod>" . esc_html($lastmod) . "</lastmod>\n";
	793	+ // Every <loc> in every sitemap passes through here, which is why the
	794	+ // scheme preference is applied at this one point rather than at each
	795	+ // of the dozen collectors that build URLs (#638). An http sitemap on
	796	+ // an https site hands search engines the wrong address for the whole
	797	+ // site at once.
	798	+ $xml .= " <loc>" . esc_url(Url_Scheme::apply($url)) . "</loc>\n";
	799	+ // Omit <lastmod> when unknown (empty) — a fabricated timestamp is worse
	800	+ // than no timestamp, and an absent lastmod is valid per the spec.
	801	+ if (!empty($lastmod)) {
	802	+ $xml .= " <lastmod>" . esc_html($lastmod) . "</lastmod>\n";
	803	+ }
353	804	$xml .= " <priority>" . number_format($priority, 1) . "</priority>\n";
354	805	$xml .= " <changefreq>" . esc_html($changefreq) . "</changefreq>\n";
355	806
356	807	// Add image entries if provided
		@@ -355,9 +806,9 @@
355	806
356	807	// Add image entries if provided
357	808	foreach ($images as $image) {
358	809	$xml .= " <image:image>\n";
359		- $xml .= " <image:loc>" . esc_url($image['url']) . "</image:loc>\n";
	810	+ $xml .= " <image:loc>" . esc_url(Url_Scheme::apply((string) $image['url'])) . "</image:loc>\n";
360	811
361	812	if (!empty($image['title'])) {
362	813	$xml .= " <image:title>" . esc_html($image['title']) . "</image:title>\n";
363	814	}
		@@ -374,33 +825,62 @@
374	825	return $xml;
375	826	}
376	827
377	828	/**
378		- * Generate post entries for sitemap (MEMORY-EFFICIENT VERSION)
	829	+ * Collect every post <url> entry for the given post type(s) as an array of
	830	+ * XML strings, using memory-efficient chunked fetching.
379	831	*
380		- * @since 1.0.0
	832	+ * Unlike the old capped generator, this returns ALL matching entries — the
	833	+ * links-per-sitemap limit is applied later by paginating this array into
	834	+ * separate files (see generate_multiple_sitemaps), matching how Rank Math
	835	+ * splits large sitemaps instead of truncating them.
381	836	*
382		- * @param string $post_type Post type
383		- * @param array $settings Sitemap settings
384		- * @return string XML entries
	837	+ * @since 1.14.0
	838	+ *
	839	+ * @param array $post_types Array of post types to fetch
	840	+ * @param array $settings Sitemap settings
	841	+ * @return array<string> Individual <url>…</url> entry strings
385	842	*/
386		- private function generate_post_entries(string $post_type, array $settings): string {
387		- // Use memory-efficient chunked processing for single post type
388		- return $this->generate_posts_chunked([$post_type], $settings);
	843	+ /**
	844	+ * lastmod for the homepage <url> entry: the static front page's real
	845	+ * modification time when one is set, otherwise the generation time.
	846	+ *
	847	+ * @return string ISO-8601 date.
	848	+ */
	849	+ private function get_homepage_lastmod(): string {
	850	+ if (get_option('show_on_front') === 'page') {
	851	+ $front_id = (int) get_option('page_on_front');
	852	+ if ($front_id) {
	853	+ $modified = get_post_field('post_modified_gmt', $front_id);
	854	+ if (!empty($modified) && $modified !== '0000-00-00 00:00:00') {
	855	+ return gmdate('c', strtotime($modified));
	856	+ }
	857	+ }
	858	+ }
	859	+ return gmdate('c');
389	860	}
390	861
	862	+ private function collect_post_entries(array $post_types, array $settings): array {
	863	+ // Materialize the streaming source for callers that need the full set
	864	+ // (e.g. the single un-paginated "general" sitemap). The paginated index
	865	+ // path streams collect_post_entries_iter() directly to bound memory.
	866	+ return iterator_to_array($this->collect_post_entries_iter($post_types, $settings), false);
	867	+ }
	868	+
391	869	/**
392		- * Generate posts XML using memory-efficient chunked processing
	870	+ * Stream <url> entries for the given post types, yielding one at a time.
393	871	*
394		- * @since 1.0.0
	872	+ * Same chunked query, filtering, and ordering as before, but yields each
	873	+ * entry instead of accumulating the whole set — so the paginated index
	874	+ * generator never holds every URL of a large post type in memory at once.
395	875	*
396		- * @param array $post_types Array of post types to fetch
397		- * @param array $settings Sitemap settings
398		- * @return string XML entries for all posts
	876	+ * @param array $post_types Post types to include.
	877	+ * @param array $settings Sitemap settings.
	878	+ * @return \Generator<string> <url> entry strings.
399	879	*/
400		- private function generate_posts_chunked(array $post_types, array $settings): string {
	880	+ private function collect_post_entries_iter(array $post_types, array $settings): \Generator {
401	881	if (empty($post_types)) {
402		- return '';
	882	+ return;
403	883	}
404	884
405	885	// Parse and validate exclude_posts setting
406	886	$exclude_ids = [];
		@@ -411,66 +891,283 @@
411	891	// Continue with empty array on validation failure - error details available in exception
412	892	}
413	893	}
414	894
415		- // Use links_per_sitemap setting to limit posts per sitemap
416		- $links_per_sitemap = !empty($settings['links_per_sitemap']) ?
417		- intval($settings['links_per_sitemap']) : 1000;
	895	+ // The static front page is emitted once as the explicit homepage entry,
	896	+ // so exclude it here to avoid a duplicate <loc> (its permalink equals
	897	+ // home_url('/')).
	898	+ if (get_option('show_on_front') === 'page') {
	899	+ $front_id = (int) get_option('page_on_front');
	900	+ if ($front_id) {
	901	+ $exclude_ids[] = $front_id;
	902	+ }
	903	+ }
418	904
419		- // Ensure value is within reasonable bounds
420		- $links_per_sitemap = max(1, min(50000, $links_per_sitemap));
	905	+ // Resolve the ordered ID list in one indexed query, then hydrate in
	906	+ // chunks via post__in. This avoids large OFFSET windows (which MySQL
	907	+ // must scan-and-discard, making a full walk O(n^2)) while still loading
	908	+ // only one chunk of full post objects into memory at a time. The
	909	+ // original order is preserved (the id query and post__in hydration both
	910	+ // use it).
	911	+ $all_ids = get_posts($this->filter_query_args([
	912	+ 'post_type' => $post_types,
	913	+ 'post_status' => 'publish',
	914	+ 'numberposts' => -1,
	915	+ 'exclude' => $exclude_ids,
	916	+ 'orderby' => 'post_type post_date',
	917	+ 'order' => 'ASC DESC',
	918	+ 'fields' => 'ids',
	919	+ ]));
421	920
422		- $xml = '';
423		- $chunk_size = 500; // Process 500 posts at a time
424		- $offset = 0;
425		- $total_processed = 0;
	921	+ if (empty($all_ids)) {
	922	+ return;
	923	+ }
426	924
427		- do {
428		- // Fetch posts in chunks to prevent memory issues
429		- $posts = get_posts([
430		- 'post_type' => $post_types,
	925	+ // Walk the ID list with a moving window rather than array_chunk().
	926	+ // array_chunk() builds a second array holding every element again, so
	927	+ // peak memory was twice the ID list — on a 100k-post site that is ~16MB
	928	+ // where ~8MB is needed, and this walk is the one part of an otherwise
	929	+ // well-bounded routine with no ceiling (#402).
	930	+ $total = count($all_ids);
	931	+
	932	+ for ($offset = 0; $offset < $total; $offset += self::ID_WALK_CHUNK) {
	933	+ $this->assert_memory_headroom();
	934	+ $chunk_start = memory_get_usage(true);
	935	+
	936	+ $chunk = array_slice($all_ids, $offset, self::ID_WALK_CHUNK);
	937	+
	938	+ $posts = get_posts($this->filter_query_args([
	939	+ 'post_type' => $post_types,
431	940	'post_status' => 'publish',
432		- 'numberposts' => $chunk_size,
433		- 'offset' => $offset,
434		- 'exclude' => $exclude_ids,
435		- 'orderby' => 'post_type post_date',
436		- 'order' => 'ASC DESC'
437		- ]);
	941	+ 'numberposts' => count($chunk),
	942	+ 'post__in' => $chunk,
	943	+ 'orderby' => 'post__in', // preserve the resolved order
	944	+ ]));
438	945
439		- // Process each post in the chunk
	946	+ // get_featured_image() hydrates each featured image under the
	947	+ // attachment's own ID, which the chunk's post IDs do not reach.
	948	+ $attachment_ids = [];
	949	+
440	950	foreach ($posts as $post) {
441		- if ($total_processed >= $links_per_sitemap) {
442		- break 2; // Exit both loops when limit reached
443		- }
444		-
445	951	if ($this->should_include_in_sitemap($post, $settings)) {
446		- $url = get_permalink($post);
	952	+ /**
	953	+ * Filter a sitemap entry's permalink.
	954	+ *
	955	+ * The multilingual manager uses this to generate each
	956	+ * translation's URL in its OWN language: the sitemap query
	957	+ * deliberately runs with suppress_filters, and the cron
	958	+ * rebuild runs with no language context at all, so a bare
	959	+ * get_permalink() resolved every translation to the
	960	+ * default-language URL — N entries sharing one <loc> (#409).
	961	+ *
	962	+ * @since 2.0.1
	963	+ * @param string $url Permalink as WordPress resolved it.
	964	+ * @param \WP_Post $post Post the entry describes.
	965	+ */
	966	+ $url = apply_filters('thinkrank_sitemap_post_permalink', get_permalink($post), $post);
447	967	$lastmod = gmdate('c', strtotime($post->post_modified_gmt));
448	968	$priority = $this->calculate_intelligent_priority($post, $post->post_type);
449	969	$changefreq = $this->calculate_change_frequency($post, $post->post_type);
450	970	$images = $this->extract_post_images($post, $settings);
451	971
452		- $xml .= $this->generate_url_entry($url, $lastmod, $priority, $changefreq, $images);
453		- $total_processed++;
	972	+ $thumbnail_id = (int) get_post_thumbnail_id($post);
	973	+ if ($thumbnail_id > 0) {
	974	+ $attachment_ids[] = $thumbnail_id;
	975	+ }
	976	+
	977	+ yield $this->generate_url_entry($url, $lastmod, $priority, $changefreq, $images);
454	978	}
455	979	}
456	980
457		- $offset += $chunk_size;
	981	+ // Free the hydrated chunk before loading the next one — including
	982	+ // the copies get_posts() left in the runtime object cache.
	983	+ unset($posts);
	984	+ $this->release_walk_memory(
	985	+ $chunk_start,
	986	+ $this->chunk_post_cache_groups($post_types),
	987	+ array_merge($chunk, $attachment_ids)
	988	+ );
	989	+ }
	990	+ }
458	991
459		- // Store count before freeing memory
460		- $posts_count = count($posts);
	992	+ /**
	993	+ * Resolve and bound the configured links-per-sitemap limit.
	994	+ *
	995	+ * @since 1.14.0
	996	+ *
	997	+ * @param array $settings Sitemap settings
	998	+ * @return int Links per sitemap file (1–50000)
	999	+ */
	1000	+ private function get_links_per_sitemap(array $settings): int {
	1001	+ $limit = !empty($settings['links_per_sitemap']) ? intval($settings['links_per_sitemap']) : 1000;
461	1002
462		- // Free memory after each chunk
463		- unset($posts);
	1003	+ return max(1, min(50000, $limit));
	1004	+ }
464	1005
465		- } while ($posts_count === $chunk_size && $total_processed < $links_per_sitemap);
	1006	+ /**
	1007	+ * Stream <url> entries for a taxonomy's terms, yielding one at a time and
	1008	+ * fetching terms in bounded chunks (number/offset) — so the paginated index
	1009	+ * generator never holds every term of a large taxonomy in memory at once.
	1010	+ *
	1011	+ * @param string $taxonomy Taxonomy name.
	1012	+ * @param array $settings Sitemap settings.
	1013	+ * @return \Generator<string> <url> entry strings.
	1014	+ */
	1015	+ private function collect_taxonomy_entries_iter(string $taxonomy, array $settings): \Generator {
	1016	+ $exclude_term_ids = [];
	1017	+ if (!empty($settings['exclude_terms'])) {
	1018	+ try {
	1019	+ $exclude_term_ids = $this->validate_exclude_terms($settings['exclude_terms']);
	1020	+ } catch (InvalidArgumentException $e) {
	1021	+ // Continue with an empty exclude list on validation failure.
	1022	+ }
	1023	+ }
466	1024
	1025	+ // product_cat may legitimately have empty terms (products added later);
	1026	+ // every other taxonomy hides empties — matching fetch_taxonomies_optimized().
	1027	+ $hide_empty = $taxonomy !== 'product_cat';
	1028	+ $priority = $taxonomy === 'category' ? 0.6 : 0.4;
	1029	+
	1030	+ // Resolve the ordered term IDs in one query, then hydrate in chunks via
	1031	+ // include. Avoids large OFFSET windows (O(n^2) over a full walk) while
	1032	+ // holding only one chunk of full term objects at a time.
	1033	+ $all_ids = get_terms($this->filter_term_query_args([
	1034	+ 'taxonomy' => $taxonomy,
	1035	+ 'hide_empty' => $hide_empty,
	1036	+ 'exclude' => $exclude_term_ids,
	1037	+ 'orderby' => 'count',
	1038	+ 'order' => 'DESC',
	1039	+ 'fields' => 'ids',
	1040	+ ]));
	1041	+
	1042	+ if (is_wp_error($all_ids) || empty($all_ids)) {
	1043	+ return;
	1044	+ }
	1045	+
	1046	+ // Same moving window as the post walk above, for the same reason.
	1047	+ $total = count($all_ids);
	1048	+
	1049	+ for ($offset = 0; $offset < $total; $offset += self::TERM_WALK_CHUNK) {
	1050	+ $this->assert_memory_headroom();
	1051	+ $chunk_start = memory_get_usage(true);
	1052	+
	1053	+ $chunk = array_slice($all_ids, $offset, self::TERM_WALK_CHUNK);
	1054	+
	1055	+ $terms = get_terms($this->filter_term_query_args([
	1056	+ 'taxonomy' => $taxonomy,
	1057	+ 'include' => $chunk,
	1058	+ 'orderby' => 'include', // preserve the resolved order
	1059	+ 'hide_empty' => false, // already filtered by the id query
	1060	+ ]));
	1061	+
	1062	+ if (is_wp_error($terms) || empty($terms)) {
	1063	+ continue;
	1064	+ }
	1065	+
	1066	+ foreach ($terms as $term) {
	1067	+ // A term the user marked noindex must not be advertised in the
	1068	+ // sitemap: the robots tag now honours term meta, so listing it
	1069	+ // here would have the sitemap contradict the page's own tag.
	1070	+ if ($this->term_is_noindexed((int) $term->term_id)) {
	1071	+ continue;
	1072	+ }
	1073	+
	1074	+ $url = get_term_link($term);
	1075	+ if (!is_wp_error($url)) {
	1076	+ // Omit lastmod for terms — the generation time is not a real
	1077	+ // modification time and would mislabel every term as just-changed.
	1078	+ yield $this->generate_url_entry($url, '', $priority, 'weekly');
	1079	+ }
	1080	+ }
	1081	+
	1082	+ unset($terms);
	1083	+ $this->release_walk_memory($chunk_start, ['terms', 'term_meta'], $chunk);
	1084	+ }
	1085	+ }
	1086	+
	1087	+ /**
	1088	+ * The XML declaration, ownership marker and optional stylesheet every
	1089	+ * sitemap document opens with.
	1090	+ *
	1091	+ * The marker is written unconditionally, and that is the point: removal on
	1092	+ * deactivate and uninstall deletes a web-root sitemap only when the file
	1093	+ * says it is ours, and our filenames are the canonical ones another SEO
	1094	+ * plugin writes too (#515). Tying the proof to `enable_styling` — the one
	1095	+ * marker older versions left — would mean a site with styling off either
	1096	+ * kept a shadowing file behind (#510) or had a competitor's deleted.
	1097	+ *
	1098	+ * @since 2.1.1
	1099	+ *
	1100	+ * The stylesheet URL is served by {@see Sitemap_Stylesheet}, not read off
	1101	+ * disk by the web server, because a static file cannot carry the site's own
	1102	+ * logo and colours (#639). It is a fixed URL: the palette is applied per
	1103	+ * request, so changing a brand colour needs no regeneration and shows up on
	1104	+ * sitemaps published long before.
	1105	+ *
	1106	+ * @param array $settings Sitemap settings (read for `enable_styling`).
	1107	+ * @param string $variant Stylesheet variant, `sitemap` or `index`.
	1108	+ * @return string Prolog lines, newline-terminated.
	1109	+ */
	1110	+ private function xml_prolog(array $settings, string $variant): string {
	1111	+ $xml = '<?xml version="1.0" encoding="UTF-8"?>' . "\n";
	1112	+ $xml .= THINKRANK_SITEMAP_MARKER . "\n";
	1113	+
	1114	+ // The stylesheet is presentation only, so it stays opt-in.
	1115	+ if (!empty($settings['enable_styling'])) {
	1116	+ $xml .= '<?xml-stylesheet type="text/xsl" href="' . esc_url(Sitemap_Stylesheet::url($variant)) . '"?>' . "\n";
	1117	+ }
	1118	+
467	1119	return $xml;
468	1120	}
469	1121
	1122	+ /**
	1123	+ * Wrap a set of <url> entry strings in a complete <urlset> document.
	1124	+ *
	1125	+ * @since 1.14.0
	1126	+ *
	1127	+ * @param array<string> $entries Entry strings
	1128	+ * @param array $settings Sitemap settings
	1129	+ * @param bool $with_image_ns Include the image sitemap namespace
	1130	+ * @return string Full sitemap XML
	1131	+ */
	1132	+ private function wrap_urlset(array $entries, array $settings, bool $with_image_ns): string {
	1133	+ $xml = $this->xml_prolog($settings, 'sitemap');
470	1134
	1135	+ if ($with_image_ns && !empty($settings['include_images'])) {
	1136	+ $xml .= '<urlset xmlns="http://www.sitemaps.org/schemas/sitemap/0.9" xmlns:image="http://www.google.com/schemas/sitemap-image/1.1">' . "\n";
	1137	+ } else {
	1138	+ $xml .= '<urlset xmlns="http://www.sitemaps.org/schemas/sitemap/0.9">' . "\n";
	1139	+ }
471	1140
	1141	+ $xml .= implode('', $entries);
	1142	+ $xml .= '</urlset>';
	1143	+
	1144	+ return $xml;
	1145	+ }
	1146	+
472	1147	/**
	1148	+ * Derive the filename/URL for a given pagination page.
	1149	+ *
	1150	+ * Page 1 keeps the base URL (e.g. /sitemap-posts.xml); pages 2+ insert the
	1151	+ * page number before the extension (e.g. /sitemap-posts-2.xml).
	1152	+ *
	1153	+ * @since 1.14.0
	1154	+ *
	1155	+ * @param string $url Base sitemap URL
	1156	+ * @param int $page 1-based page number
	1157	+ * @return string Paginated URL
	1158	+ */
	1159	+ private function paginate_url(string $url, int $page): string {
	1160	+ if ($page <= 1) {
	1161	+ return $url;
	1162	+ }
	1163	+
	1164	+ return (string) preg_replace('/\.xml$/i', '-' . $page . '.xml', $url);
	1165	+ }
	1166	+
	1167	+
	1168	+
	1169	+ /**
473	1170	* Extract images from post for sitemap
474	1171	*
475	1172	* @since 1.0.0
476	1173	*
		@@ -501,9 +1198,9 @@
501	1198	// Remove duplicates based on URL
502	1199	$unique_images = [];
503	1200	$seen_urls = [];
504	1201	foreach ($images as $image) {
505		- if (!in_array($image['url'], $seen_urls)) {
	1202	+ if (!in_array($image['url'], $seen_urls, true)) {
506	1203	$unique_images[] = $image;
507	1204	$seen_urls[] = $image['url'];
508	1205	}
509	1206	}
		@@ -587,28 +1284,8 @@
587	1284	return $images;
588	1285	}
589	1286
590	1287	/**
591		- * Generate taxonomy entries for sitemap (OPTIMIZED VERSION)
592		- *
593		- * @since 1.0.0
594		- *
595		- * @param string $taxonomy Taxonomy name
596		- * @param array $settings Sitemap settings
597		- * @return string XML entries
598		- */
599		- private function generate_taxonomy_entries(string $taxonomy, array $settings): string {
600		- // Use optimized batch fetching for single taxonomy
601		- $taxonomy_data = $this->fetch_taxonomies_optimized([$taxonomy], $settings);
602		-
603		- if (empty($taxonomy_data[$taxonomy])) {
604		- return '';
605		- }
606		-
607		- return $this->process_taxonomies_to_xml($taxonomy_data[$taxonomy], $taxonomy, $settings);
608		- }
609		-
610		- /**
611	1288	* Get enabled taxonomies based on settings
612	1289	*
613	1290	* @since 1.0.0
614	1291	* @param array $settings Sitemap settings
		@@ -632,9 +1309,16 @@
632	1309	'_builtin' => false
633	1310	], 'names');
634	1311
635	1312	foreach ($custom_taxonomies as $taxonomy) {
636		- if ($this->should_include_taxonomy($taxonomy)) {
	1313	+ if (!$this->should_include_taxonomy($taxonomy)) {
	1314	+ continue;
	1315	+ }
	1316	+
	1317	+ // Same as the post-type walk above: an explicit per-taxonomy flag
	1318	+ // now decides, and an unset flag keeps the previous "included"
	1319	+ // behaviour (#660).
	1320	+ if (\ThinkRank\SEO\Content_Type_Settings::is_included_in_sitemap('taxonomy', $taxonomy, $settings)) {
637	1321	$taxonomies[] = $taxonomy;
638	1322	}
639	1323	}
640	1324
		@@ -677,20 +1361,28 @@
677	1361	}
678	1362	}
679	1363
680	1364	// OPTIMIZED: Single query for multiple taxonomies
681		- $all_terms = get_terms([
	1365	+ $all_terms = get_terms($this->filter_term_query_args([
682	1366	'taxonomy' => $taxonomies, // ✅ Multiple taxonomies in single query
683	1367	'hide_empty' => $hide_empty,
684	1368	'exclude' => $exclude_term_ids,
685	1369	'orderby' => 'taxonomy count',
686	1370	'order' => 'ASC DESC' // Order by taxonomy ASC, then count DESC
687		- ]);
	1371	+ ]));
688	1372
689	1373	if (is_wp_error($all_terms)) {
690	1374	return [];
691	1375	}
692	1376
	1377	+ // Drop terms the user marked noindex. This path feeds the single general
	1378	+ // sitemap while collect_taxonomy_entries_iter() feeds the segmented ones,
	1379	+ // so both need the filter or the two disagree about the same term.
	1380	+ $all_terms = array_values(array_filter(
	1381	+ $all_terms,
	1382	+ fn($term) => !$this->term_is_noindexed((int) $term->term_id)
	1383	+ ));
	1384	+
693	1385	// Group terms by taxonomy
694	1386	return $this->group_terms_by_taxonomy($all_terms);
695	1387	}
696	1388
		@@ -733,9 +1425,11 @@
733	1425
734	1426	foreach ($terms as $term) {
735	1427	$url = get_term_link($term);
736	1428	if (!is_wp_error($url)) {
737		- $lastmod = gmdate('c');
	1429	+ // Omit lastmod for terms (generation time is not a real
	1430	+ // modification time).
	1431	+ $lastmod = '';
738	1432	$priority = $taxonomy === 'category' ? 0.6 : 0.4;
739	1433	$changefreq = 'weekly';
740	1434
741	1435	$xml .= $this->generate_url_entry($url, $lastmod, $priority, $changefreq);
		@@ -778,12 +1472,12 @@
778	1472
779	1473	// Special page types get higher priority
780	1474	if ($post_type === 'page') {
781	1475	$page_template = get_page_template_slug($post->ID);
782		- if (in_array($page_template, ['page-home.php', 'front-page.php']) ||
783		- $post->ID == get_option('page_on_front')) {
	1476	+ if (in_array($page_template, ['page-home.php', 'front-page.php'], true) ||
	1477	+ (int) $post->ID === (int) get_option('page_on_front')) {
784	1478	$base_priority = 1.0; // Homepage gets maximum priority
785		- } elseif (in_array($page_template, ['page-contact.php', 'page-about.php'])) {
	1479	+ } elseif (in_array($page_template, ['page-contact.php', 'page-about.php'], true)) {
786	1480	$adjustments += 0.05; // Important pages boost
787	1481	}
788	1482	}
789	1483
		@@ -805,11 +1499,11 @@
805	1499	private function calculate_change_frequency(\WP_Post $post, string $post_type): string {
806	1500	// Pages typically change less frequently
807	1501	if ($post_type === 'page') {
808	1502	$page_template = get_page_template_slug($post->ID);
809		- if ($post->ID == get_option('page_on_front')) {
	1503	+ if ((int) $post->ID === (int) get_option('page_on_front')) {
810	1504	return 'daily'; // Homepage changes frequently
811		- } elseif (in_array($page_template, ['page-contact.php', 'page-about.php'])) {
	1505	+ } elseif (in_array($page_template, ['page-contact.php', 'page-about.php'], true)) {
812	1506	return 'monthly'; // Static pages change monthly
813	1507	}
814	1508	return 'yearly'; // Other pages change rarely
815	1509	}
		@@ -837,8 +1531,17 @@
837	1531	* @param array $settings Sitemap settings
838	1532	* @return bool Whether to include in sitemap
839	1533	*/
840	1534	private function should_include_in_sitemap(\WP_Post $post, array $settings): bool {
	1535	+ // The WooCommerce cart, checkout and account pages are transactional,
	1536	+ // never indexable, and generate_default_robots_rules() already emits a
	1537	+ // Disallow for each of them. Listing them here submitted URLs our own
	1538	+ // robots.txt blocks, which Search Console reports as "Submitted URL
	1539	+ // blocked by robots.txt". Yoast and Rank Math exclude the same three.
	1540	+ if (in_array($post->ID, $this->woocommerce_excluded_page_ids(), true)) {
	1541	+ return false;
	1542	+ }
	1543	+
841	1544	// Respect user setting for password protected content
842	1545	if (!empty($post->post_password) && !empty($settings['exclude_password_protected'])) {
843	1546	return false;
844	1547	}
		@@ -847,46 +1550,93 @@
847	1550	if ($post->post_status === 'private' && !empty($settings['exclude_private_posts'])) {
848	1551	return false;
849	1552	}
850	1553
851		- // Skip very short content (likely test/placeholder)
852		- $content_length = strlen(wp_strip_all_tags($post->post_content));
853		- if ($content_length < 100) {
	1554	+ // Only published (and, per setting, private) content belongs in the
	1555	+ // sitemap. Content-quality heuristics (length, "demo"/"test"/"sample"
	1556	+ // in the title, "lorem ipsum" text) were intentionally removed: an XML
	1557	+ // sitemap should list every indexable published URL. Filtering by
	1558	+ // description length silently dropped legitimate WooCommerce products
	1559	+ // with short descriptions, and the substring title match excluded real
	1560	+ // pages such as "Demo" or "Product Samples". Indexability is governed
	1561	+ // by noindex directives below, not by heuristics.
	1562	+ if (!in_array($post->post_status, ['publish', 'private'], true)) {
854	1563	return false;
855	1564	}
856	1565
857		- // Skip posts with test/demo indicators in title (refined filtering)
858		- $title_indicators = ['test', 'demo', 'sample', 'placeholder'];
859		- $title_lower = strtolower($post->post_title);
	1566	+ // Check if post overrides robots and sets noindex.
	1567	+ if ((bool) get_post_meta($post->ID, '_thinkrank_robots_meta_enabled', true)) {
	1568	+ $raw = get_post_meta($post->ID, '_thinkrank_robots_meta', true);
	1569	+ if (is_string($raw) && $raw !== '') {
	1570	+ $robots = json_decode($raw, true);
	1571	+ if (is_array($robots) && !empty($robots['noindex'])) {
	1572	+ return false;
	1573	+ }
	1574	+ }
	1575	+ }
860	1576
861		- foreach ($title_indicators as $indicator) {
862		- if (strpos($title_lower, $indicator) !== false) {
863		- return false;
864		- }
	1577	+ return true;
	1578	+ }
	1579	+
	1580	+ /**
	1581	+ * WooCommerce pages that must never reach the sitemap.
	1582	+ *
	1583	+ * Resolved through wc_get_page_id() so a store that moved or renamed its
	1584	+ * cart/checkout/account pages is still matched. Returns an empty list when
	1585	+ * WooCommerce is not active. Memoised — should_include_in_sitemap() runs
	1586	+ * once per post.
	1587	+ *
	1588	+ * @since 2.0.1
	1589	+ *
	1590	+ * @return int[] Page IDs to exclude.
	1591	+ */
	1592	+ private function woocommerce_excluded_page_ids(): array {
	1593	+ if ($this->woocommerce_excluded_page_ids !== null) {
	1594	+ return $this->woocommerce_excluded_page_ids;
865	1595	}
866	1596
867		- // Skip content with obvious placeholder text
868		- $content_indicators = ['lorem ipsum'];
869		- $content_lower = strtolower($post->post_content);
	1597	+ $ids = [];
870	1598
871		- foreach ($content_indicators as $indicator) {
872		- if (strpos($content_lower, $indicator) !== false) {
873		- return false;
	1599	+ if (function_exists('wc_get_page_id')) {
	1600	+ foreach (['cart', 'checkout', 'myaccount'] as $page) {
	1601	+ $id = (int) wc_get_page_id($page);
	1602	+ // wc_get_page_id() returns -1 when the page is not configured.
	1603	+ if ($id > 0) {
	1604	+ $ids[] = $id;
	1605	+ }
874	1606	}
875	1607	}
876	1608
877		- // Skip drafts (but allow private posts if user setting permits)
878		- if (!in_array($post->post_status, ['publish', 'private'])) {
	1609	+ $this->woocommerce_excluded_page_ids = $ids;
	1610	+
	1611	+ return $ids;
	1612	+ }
	1613	+
	1614	+ /**
	1615	+ * Whether a term carries an explicit noindex override.
	1616	+ *
	1617	+ * Mirrors the post-side check in should_include_post(); terms store the same
	1618	+ * `_thinkrank_robots_meta_enabled` / `_thinkrank_robots_meta` keys, written
	1619	+ * by the update-term-seo ability and by the SEO importer.
	1620	+ *
	1621	+ * @since 1.31.0
	1622	+ *
	1623	+ * @param int $term_id Term to test.
	1624	+ * @return bool True when the term is marked noindex.
	1625	+ */
	1626	+ private function term_is_noindexed(int $term_id): bool {
	1627	+ if (!(bool) get_term_meta($term_id, '_thinkrank_robots_meta_enabled', true)) {
879	1628	return false;
880	1629	}
881	1630
882		- // Check if post is set to noindex via meta
883		- $noindex = get_post_meta($post->ID, '_thinkrank_noindex', true);
884		- if ($noindex === '1' || $noindex === 'true') {
	1631	+ $raw = get_term_meta($term_id, '_thinkrank_robots_meta', true);
	1632	+ if (!is_string($raw) || $raw === '') {
885	1633	return false;
886	1634	}
887	1635
888		- return true;
	1636	+ $robots = json_decode($raw, true);
	1637	+
	1638	+ return is_array($robots) && !empty($robots['noindex']);
889	1639	}
890	1640
891	1641	/**
892	1642	* Count total URLs in sitemap
		@@ -895,9 +1645,9 @@
895	1645	*
896	1646	* @param array $settings Sitemap settings
897	1647	* @return int Total URL count
898	1648	*/
899		- private function count_sitemap_urls(array $settings): int {
	1649	+ public function count_sitemap_urls(array $settings): int {
900	1650	$count = 1; // Homepage
901	1651
902	1652	if (!empty($settings['include_posts'])) {
903	1653	$count += wp_count_posts('post')->publish;
		@@ -1052,9 +1802,15 @@
1052	1802	if (!empty($settings['include_pages'])) {
1053	1803	$post_types[] = 'page';
1054	1804	}
1055	1805
1056		- // Auto-detect public custom post types that should be included
	1806	+ // Auto-detect public custom post types that should be included.
	1807	+ //
	1808	+ // A custom type's `include_<slug>` / `exclude_<slug>` flag is honoured
	1809	+ // here (#660). It was previously stored — additional_setting_keys()
	1810	+ // has always let those keys through — but never read, so a CPT was in
	1811	+ // the sitemap whatever the setting said. Unset still means included, so
	1812	+ // a site that never touched the flag is unaffected.
1057	1813	$custom_post_types = get_post_types([
1058	1814	'public' => true,
1059	1815	'_builtin' => false
1060	1816	], 'names');
		@@ -1059,9 +1815,13 @@
1059	1815	'_builtin' => false
1060	1816	], 'names');
1061	1817
1062	1818	foreach ($custom_post_types as $post_type) {
1063		- if ($this->should_include_post_type($post_type)) {
	1819	+ if (!$this->should_include_post_type($post_type)) {
	1820	+ continue;
	1821	+ }
	1822	+
	1823	+ if (\ThinkRank\SEO\Content_Type_Settings::is_included_in_sitemap('post_type', $post_type, $settings)) {
1064	1824	$post_types[] = $post_type;
1065	1825	}
1066	1826	}
1067	1827
		@@ -1075,11 +1835,18 @@
1075	1835	* @param string $post_type Post type
1076	1836	* @return bool True if post type should trigger regeneration
1077	1837	*/
1078	1838	private function should_include_post_type(string $post_type): bool {
1079		- // Include all public post types (presets control which sitemaps are created)
1080		- $post_type_obj = get_post_type_object($post_type);
1081		- return $post_type_obj && $post_type_obj->public;
	1839	+ // Match Rank Math: only publicly viewable post types belong in the
	1840	+ // sitemap (public && publicly_queryable). Additionally skip post types
	1841	+ // that opt out of front-end search (exclude_from_search => true) — e.g.
	1842	+ // Templately's internal `templately_library` store — which are template
	1843	+ // records, not standalone indexable URLs. BetterDocs `docs` and
	1844	+ // WooCommerce `product` register exclude_from_search => false, so they
	1845	+ // remain included.
	1846	+ // The predicate lives in Content_Type_Settings so the matrix can ask
	1847	+ // the same question before offering a switch for this post type.
	1848	+ return \ThinkRank\SEO\Content_Type_Settings::sitemap_accepts_post_type($post_type);
1082	1849	}
1083	1850
1084	1851	/**
1085	1852	* Check if taxonomy should trigger regeneration
		@@ -1088,11 +1855,13 @@
1088	1855	* @param string $taxonomy Taxonomy slug
1089	1856	* @return bool True if taxonomy should trigger regeneration
1090	1857	*/
1091	1858	private function should_include_taxonomy(string $taxonomy): bool {
1092		- // Include all public taxonomies (presets control which sitemaps are created)
1093		- $taxonomy_obj = get_taxonomy($taxonomy);
1094		- return $taxonomy_obj && $taxonomy_obj->public;
	1859	+ // Public taxonomies only, and only the ones this generator can actually
	1860	+ // emit — the same predicate the content-type matrix asks before it
	1861	+ // offers a sitemap switch for one (presets still control which
	1862	+ // sitemaps are created).
	1863	+ return \ThinkRank\SEO\Content_Type_Settings::sitemap_accepts_taxonomy($taxonomy);
1095	1864	}
1096	1865
1097	1866	/**
1098	1867	* Schedule debounced sitemap regeneration
		@@ -1100,16 +1869,870 @@
1100	1869	* @since 1.0.0
1101	1870	* @return void
1102	1871	*/
1103	1872	private function schedule_debounced_regeneration(): void {
1104		- // Clear any existing scheduled regeneration
1105		- wp_clear_scheduled_hook('thinkrank_regenerate_sitemap');
	1873	+ $this->mark_regeneration_pending('content');
	1874	+ $this->debounce_event('thinkrank_regenerate_sitemap');
	1875	+ }
1106	1876
1107		- // Schedule regeneration in 30 seconds to debounce rapid changes
1108		- wp_schedule_single_event(time() + 30, 'thinkrank_regenerate_sitemap');
	1877	+ /**
	1878	+ * Schedule (or keep) the debounced single event behind a regeneration hook.
	1879	+ *
	1880	+ * An event that is already due is left alone. WP-Cron only runs when a
	1881	+ * request arrives, so on a site with DISABLE_WP_CRON, a blocked loopback or
	1882	+ * little traffic an overdue event can sit in the queue for a long time —
	1883	+ * clearing and re-scheduling it on every save pushed the rebuild
	1884	+ * permanently 30 seconds into the future and the sitemap never updated
	1885	+ * (#629). Debouncing only against an event that has not come due yet keeps
	1886	+ * the bulk-edit coalescing without starving the rebuild.
	1887	+ *
	1888	+ * @since 2.2.1
	1889	+ * @param string $hook Regeneration hook to debounce.
	1890	+ * @return void
	1891	+ */
	1892	+ private function debounce_event(string $hook): void {
	1893	+ $next = wp_next_scheduled($hook);
	1894	+
	1895	+ if ($next !== false) {
	1896	+ if ($next <= time()) {
	1897	+ return;
	1898	+ }
	1899	+
	1900	+ wp_clear_scheduled_hook($hook);
	1901	+ }
	1902	+
	1903	+ wp_schedule_single_event(time() + self::REGENERATION_DEBOUNCE, $hook);
1109	1904	}
1110	1905
1111	1906	/**
	1907	+ * Record that a rebuild is outstanding, so an overdue one can be taken over
	1908	+ * by a later request and its staleness surfaced in the UI.
	1909	+ *
	1910	+ * `since` is the *oldest* outstanding change: it is what the takeover grace
	1911	+ * and the admin staleness warning are measured from, so successive edits
	1912	+ * must not push it forward. A settings change outranks a content change —
	1913	+ * it rebuilds regardless of the auto_generate toggle and handles a sitemap
	1914	+ * that has just been disabled — so once one is outstanding it stays the
	1915	+ * recorded source until the rebuild lands.
	1916	+ *
	1917	+ * @since 2.2.1
	1918	+ * @param string $source Either 'content' or 'settings'.
	1919	+ * @return void
	1920	+ */
	1921	+ private function mark_regeneration_pending(string $source): void {
	1922	+ // Whatever made the static files stale made the rendered ones stale
	1923	+ // too. Invalidating here rather than only on the rebuild keeps the two
	1924	+ // delivery modes reacting to exactly the same triggers, which is the
	1925	+ // only way a dynamic site stays as fresh as a static one (#752).
	1926	+ $this->flush_dynamic_cache();
	1927	+
	1928	+ $pending = get_option(self::REGENERATION_PENDING_OPTION, []);
	1929	+ $pending = is_array($pending) ? $pending : [];
	1930	+
	1931	+ $since = !empty($pending['since']) ? (int) $pending['since'] : time();
	1932	+ $current = isset($pending['source']) ? (string) $pending['source'] : '';
	1933	+ $source = ($current === 'settings' || $source === 'settings') ? 'settings' : 'content';
	1934	+
	1935	+ // Merged so the bookkeeping a rebuild keeps on the marker (`started`,
	1936	+ // `memory_limit`) survives an edit made while it is outstanding.
	1937	+ update_option(
	1938	+ self::REGENERATION_PENDING_OPTION,
	1939	+ array_merge($pending, [
	1940	+ 'since' => $since,
	1941	+ 'source' => $source,
	1942	+ 'attempts' => !empty($pending['attempts']) ? (int) $pending['attempts'] : 0,
	1943	+ 'next_attempt' => !empty($pending['next_attempt'])
	1944	+ ? (int) $pending['next_attempt']
	1945	+ : time() + self::REGENERATION_TAKEOVER_GRACE,
	1946	+ // Bumped on every change so a rebuild can tell whether the edit
	1947	+ // it started for is still the newest one outstanding.
	1948	+ 'revision' => (!empty($pending['revision']) ? (int) $pending['revision'] : 0) + 1,
	1949	+ ]),
	1950	+ true
	1951	+ );
	1952	+ }
	1953	+
	1954	+ /**
	1955	+ * The revision of the outstanding rebuild, for
	1956	+ * {@see mark_regeneration_complete()} to compare against once it is done.
	1957	+ *
	1958	+ * @since 2.2.1
	1959	+ * @return int Current revision, 0 when nothing is outstanding.
	1960	+ */
	1961	+ private function current_regeneration_revision(): int {
	1962	+ $pending = get_option(self::REGENERATION_PENDING_OPTION, []);
	1963	+
	1964	+ return (is_array($pending) && !empty($pending['revision'])) ? (int) $pending['revision'] : 0;
	1965	+ }
	1966	+
	1967	+ /**
	1968	+ * Clear the outstanding-rebuild marker and any recorded failure.
	1969	+ *
	1970	+ * Public because a manual generation satisfies whatever the automatic path
	1971	+ * was still waiting to write.
	1972	+ *
	1973	+ * @since 2.2.1
	1974	+ * @return void
	1975	+ */
	1976	+ public function mark_regeneration_complete(?int $revision = null): void {
	1977	+ // The write succeeded, so whatever failure was on record is history.
	1978	+ if (get_option(self::REGENERATION_ERROR_OPTION, null) !== null) {
	1979	+ delete_option(self::REGENERATION_ERROR_OPTION);
	1980	+ }
	1981	+
	1982	+ $pending = get_option(self::REGENERATION_PENDING_OPTION, null);
	1983	+
	1984	+ if ($pending === null) {
	1985	+ return;
	1986	+ }
	1987	+
	1988	+ // A change that landed while this rebuild was running is not covered by
	1989	+ // the files it just wrote, so it has to stay outstanding — otherwise, on
	1990	+ // a site where WP-Cron never fires, clearing the marker would strand it
	1991	+ // exactly the way #629 stranded everything.
	1992	+ if (
	1993	+ $revision !== null
	1994	+ && is_array($pending)
	1995	+ && (int) ($pending['revision'] ?? 0) !== $revision
	1996	+ ) {
	1997	+ // This attempt succeeded, so drop what it claimed: the newer change
	1998	+ // waits the grace a fresh edit gets, not a failure backoff it never
	1999	+ // earned. Not zero: that edit queued its own debounced event, and
	2000	+ // a marker due at once had the next admin request rebuild in its
	2001	+ // shutdown and the event rebuild again seconds later.
	2002	+ unset($pending['started'], $pending['memory_limit']);
	2003	+ $pending['attempts'] = 0;
	2004	+ $pending['next_attempt'] = time() + self::REGENERATION_TAKEOVER_GRACE;
	2005	+
	2006	+ update_option(self::REGENERATION_PENDING_OPTION, $pending, true);
	2007	+ return;
	2008	+ }
	2009	+
	2010	+ delete_option(self::REGENERATION_PENDING_OPTION);
	2011	+ }
	2012	+
	2013	+ /**
	2014	+ * Record the attempt that is about to run before it runs.
	2015	+ *
	2016	+ * A PHP fatal — the memory limit or max_execution_time — is not a
	2017	+ * Throwable, so no catch or finally around the generation runs when the
	2018	+ * process dies, and a failure recorded afterwards was never recorded at
	2019	+ * all: `attempts` stayed 0, the backoff never applied, and the next request
	2020	+ * started the same doomed rebuild again (a fatal every few minutes for as
	2021	+ * long as an admin was logged in). Claiming the attempt up front makes the
	2022	+ * backoff hold even when nothing after this line gets to run, and leaves a
	2023	+ * `started` stamp the next attempt can recognise as an interrupted one.
	2024	+ *
	2025	+ * @since 2.10.1
	2026	+ * @return void
	2027	+ */
	2028	+ private function claim_regeneration_attempt(): void {
	2029	+ $pending = get_option(self::REGENERATION_PENDING_OPTION, null);
	2030	+
	2031	+ // Nothing outstanding (e.g. a manual generation already satisfied it):
	2032	+ // there is no marker to retry from, so nothing to claim.
	2033	+ if (!is_array($pending) || empty($pending['since'])) {
	2034	+ return;
	2035	+ }
	2036	+
	2037	+ if (!empty($pending['started'])) {
	2038	+ // The previous attempt claimed itself and never reported back.
	2039	+ update_option(
	2040	+ self::REGENERATION_ERROR_OPTION,
	2041	+ [
	2042	+ 'message' => __('The previous automatic sitemap rebuild stopped before it finished, most likely because PHP ran out of memory or time. It is retried with a growing delay. If this keeps happening, raise the PHP memory_limit or max_execution_time, or run WP-Cron from a system cron.', 'thinkrank'),
	2043	+ 'source' => isset($pending['source']) ? (string) $pending['source'] : 'content',
	2044	+ 'attempts' => !empty($pending['attempts']) ? (int) $pending['attempts'] : 1,
	2045	+ 'time' => (int) $pending['started'],
	2046	+ ],
	2047	+ false
	2048	+ );
	2049	+ }
	2050	+
	2051	+ $attempts = (!empty($pending['attempts']) ? (int) $pending['attempts'] : 0) + 1;
	2052	+
	2053	+ $pending['attempts'] = $attempts;
	2054	+ $pending['next_attempt'] = time() + $this->regeneration_backoff($attempts);
	2055	+ $pending['started'] = time();
	2056	+
	2057	+ update_option(self::REGENERATION_PENDING_OPTION, $pending, true);
	2058	+ }
	2059	+
	2060	+ /**
	2061	+ * Delay before the next takeover after the given number of attempts.
	2062	+ *
	2063	+ * @since 2.10.1
	2064	+ * @param int $attempts Attempts made so far (1 or more).
	2065	+ * @return int Seconds.
	2066	+ */
	2067	+ private function regeneration_backoff(int $attempts): int {
	2068	+ return (int) min(
	2069	+ self::REGENERATION_TAKEOVER_GRACE * (2 ** min(max($attempts, 1), 10)),
	2070	+ self::REGENERATION_MAX_BACKOFF
	2071	+ );
	2072	+ }
	2073	+
	2074	+ /**
	2075	+ * Record a failed regeneration instead of discarding it.
	2076	+ *
	2077	+ * Keeps the pending marker in place so the rebuild is retried, but backs the
	2078	+ * next attempt off exponentially (capped) so a persistently failing
	2079	+ * generation cannot run on every admin request.
	2080	+ *
	2081	+ * @since 2.2.1
	2082	+ * @since 2.10.1 Accepts the memory limit a rebuild had to stop short of, and
	2083	+ * does not count an attempt claim_regeneration_attempt()
	2084	+ * already counted.
	2085	+ * @param string $message Failure detail.
	2086	+ * @param string $source Either 'content' or 'settings'.
	2087	+ * @param int|null $memory_limit Memory limit (bytes) the rebuild stopped
	2088	+ * short of, when that was the failure.
	2089	+ * @return void
	2090	+ */
	2091	+ private function record_regeneration_failure(string $message, string $source, ?int $memory_limit = null): void {
	2092	+ $pending = get_option(self::REGENERATION_PENDING_OPTION, []);
	2093	+ $pending = is_array($pending) ? $pending : [];
	2094	+ $attempts = !empty($pending['attempts']) ? (int) $pending['attempts'] : 0;
	2095	+
	2096	+ // An attempt that claimed itself up front has already been counted.
	2097	+ if (empty($pending['started'])) {
	2098	+ $attempts++;
	2099	+ }
	2100	+ $attempts = max($attempts, 1);
	2101	+
	2102	+ $backoff = $this->regeneration_backoff($attempts);
	2103	+
	2104	+ // Same precedence mark_regeneration_pending() enforces: a settings
	2105	+ // rebuild outranks a content one and must not be downgraded by a failed
	2106	+ // attempt. Overwriting it routed the retry back through the content
	2107	+ // path, where should_auto_generate() can be false and the completion
	2108	+ // marker then discards the settings rebuild entirely. Only the
	2109	+ // outstanding rebuild is upgraded — the recorded error keeps reporting
	2110	+ // whichever attempt actually failed.
	2111	+ $current = isset($pending['source']) ? (string) $pending['source'] : '';
	2112	+ $pending_source = ($current === 'settings' || $source === 'settings') ? 'settings' : 'content';
	2113	+
	2114	+ $marker = [
	2115	+ 'since' => !empty($pending['since']) ? (int) $pending['since'] : time(),
	2116	+ 'source' => $pending_source,
	2117	+ 'attempts' => $attempts,
	2118	+ 'next_attempt' => time() + $backoff,
	2119	+ 'revision' => !empty($pending['revision']) ? (int) $pending['revision'] : 0,
	2120	+ ];
	2121	+
	2122	+ // Remembered so has_memory_for_retry() can keep requests with no more
	2123	+ // memory than this from repeating the same attempt.
	2124	+ if ($memory_limit !== null && $memory_limit > 0) {
	2125	+ $marker['memory_limit'] = $memory_limit;
	2126	+ }
	2127	+
	2128	+ update_option(self::REGENERATION_PENDING_OPTION, $marker, true);
	2129	+
	2130	+ update_option(
	2131	+ self::REGENERATION_ERROR_OPTION,
	2132	+ [
	2133	+ 'message' => $message,
	2134	+ 'source' => $source,
	2135	+ 'attempts' => $attempts,
	2136	+ 'time' => time(),
	2137	+ ],
	2138	+ false
	2139	+ );
	2140	+
	2141	+ if (defined('WP_DEBUG') && WP_DEBUG) {
	2142	+ // phpcs:ignore WordPress.PHP.DevelopmentFunctions.error_log_error_log
	2143	+ error_log(sprintf('ThinkRank: sitemap %s regeneration failed — %s', $source, $message));
	2144	+ }
	2145	+ }
	2146	+
	2147	+ /**
	2148	+ * Is a rebuild outstanding and past the point where WP-Cron should have run
	2149	+ * it?
	2150	+ *
	2151	+ * Deliberately cheap — one autoloaded option read — because it is consulted
	2152	+ * on every admin request to decide whether the takeover is needed.
	2153	+ *
	2154	+ * @since 2.2.1
	2155	+ * @return bool True when a request should rebuild the sitemap itself.
	2156	+ */
	2157	+ public static function has_overdue_regeneration(): bool {
	2158	+ $pending = get_option(self::REGENERATION_PENDING_OPTION, []);
	2159	+
	2160	+ if (!is_array($pending) || empty($pending['since'])) {
	2161	+ return false;
	2162	+ }
	2163	+
	2164	+ $due = !empty($pending['next_attempt'])
	2165	+ ? (int) $pending['next_attempt']
	2166	+ : (int) $pending['since'] + self::REGENERATION_TAKEOVER_GRACE;
	2167	+
	2168	+ return time() >= $due;
	2169	+ }
	2170	+
	2171	+ /**
	2172	+ * Rebuild the sitemap in-request when WP-Cron has not delivered.
	2173	+ *
	2174	+ * Hooked on `shutdown` for admin, REST and CLI requests only (see
	2175	+ * Plugin::register_sitemap_cron_listeners()), so the work happens after the
	2176	+ * response has been sent and never adds latency to a visitor page view.
	2177	+ *
	2178	+ * @since 2.2.1
	2179	+ * @return void
	2180	+ */
	2181	+ public function run_overdue_regeneration(): void {
	2182	+ if (!self::has_overdue_regeneration()) {
	2183	+ return;
	2184	+ }
	2185	+
	2186	+ $pending = get_option(self::REGENERATION_PENDING_OPTION, []);
	2187	+ $pending = is_array($pending) ? $pending : [];
	2188	+ $source = isset($pending['source']) ? (string) $pending['source'] : 'content';
	2189	+
	2190	+ // Cron is running and its own event for this rebuild is still queued
	2191	+ // and due: it is about to do exactly this work. Only then — an event
	2192	+ // that has already been consumed (e.g. it fired while another process
	2193	+ // held the generation lock) is never re-queued, and returning here
	2194	+ // unconditionally left the rebuild to requests that could not finish it.
	2195	+ if (wp_doing_cron()) {
	2196	+ $hook = $source === 'settings' ? 'thinkrank_regenerate_sitemap_settings' : 'thinkrank_regenerate_sitemap';
	2197	+ $next = wp_next_scheduled($hook);
	2198	+
	2199	+ if ($next !== false && $next <= time()) {
	2200	+ return;
	2201	+ }
	2202	+ }
	2203	+
	2204	+ // The last attempt had to stop short of this process's memory limit.
	2205	+ // Retrying at the same (or a lower) limit only repeats that, so leave
	2206	+ // the rebuild to a process with more room — WP-CLI, a system cron, or a
	2207	+ // host with a higher limit — instead of burning it on every request.
	2208	+ if (!$this->has_memory_for_retry($pending)) {
	2209	+ return;
	2210	+ }
	2211	+
	2212	+ if ($source === 'settings') {
	2213	+ $this->regenerate_sitemap_from_settings();
	2214	+ return;
	2215	+ }
	2216	+
	2217	+ $this->auto_regenerate_sitemap();
	2218	+ }
	2219	+
	2220	+ /**
	2221	+ * Acquire the shared generation lock.
	2222	+ *
	2223	+ * @since 2.2.1
	2224	+ * @return bool True when this process may generate.
	2225	+ */
	2226	+ private function acquire_generation_lock(): bool {
	2227	+ if (get_transient(self::GENERATION_LOCK_TRANSIENT)) {
	2228	+ return false;
	2229	+ }
	2230	+
	2231	+ set_transient(self::GENERATION_LOCK_TRANSIENT, time(), 5 * MINUTE_IN_SECONDS);
	2232	+
	2233	+ return true;
	2234	+ }
	2235	+
	2236	+ /**
	2237	+ * Release the shared generation lock.
	2238	+ *
	2239	+ * @since 2.2.1
	2240	+ * @return void
	2241	+ */
	2242	+ private function release_generation_lock(): void {
	2243	+ delete_transient(self::GENERATION_LOCK_TRANSIENT);
	2244	+ }
	2245	+
	2246	+ /**
	2247	+ * This process's PHP memory limit in bytes.
	2248	+ *
	2249	+ * @since 2.10.1
	2250	+ * @return int Bytes, or -1 when unlimited (or unreadable).
	2251	+ */
	2252	+ private function current_memory_limit(): int {
	2253	+ $limit = (string) ini_get('memory_limit');
	2254	+
	2255	+ if ($limit === '' || $limit === '-1') {
	2256	+ return -1;
	2257	+ }
	2258	+
	2259	+ $bytes = (int) wp_convert_hr_to_bytes($limit);
	2260	+
	2261	+ return $bytes > 0 ? $bytes : -1;
	2262	+ }
	2263	+
	2264	+ /**
	2265	+ * May this process retry a rebuild that last stopped at the memory limit?
	2266	+ *
	2267	+ * Raises the limit the way wp-admin does first, so a request that can get
	2268	+ * more room than the failed attempt had is still allowed to try.
	2269	+ *
	2270	+ * @since 2.10.1
	2271	+ * @param array $pending The pending marker.
	2272	+ * @return bool True when there is no recorded memory failure, or this
	2273	+ * process has more memory than the attempt that failed.
	2274	+ */
	2275	+ private function has_memory_for_retry(array $pending): bool {
	2276	+ if (empty($pending['memory_limit'])) {
	2277	+ return true;
	2278	+ }
	2279	+
	2280	+ wp_raise_memory_limit('admin');
	2281	+
	2282	+ $limit = $this->current_memory_limit();
	2283	+
	2284	+ return $limit === -1 || $limit > (int) $pending['memory_limit'];
	2285	+ }
	2286	+
	2287	+ /**
	2288	+ * Release what one walked chunk left behind.
	2289	+ *
	2290	+ * Hydrating a chunk through get_posts()/get_terms() also stores every
	2291	+ * object and its meta in the in-process object cache, which nothing
	2292	+ * empties until the request ends. Unsetting the chunk therefore freed
	2293	+ * nothing, and the walk grew with the size of the site instead of the size
	2294	+ * of a chunk — about 1.3 GB on a 45k-post site.
	2295	+ *
	2296	+ * A persistent object cache that supports it drops only its in-process
	2297	+ * copy (`flush_runtime`); the shared store keeps its data. WordPress's
	2298	+ * default cache has no shared store, and flushing it would empty every
	2299	+ * group for the rest of the request (options, the queried object, other
	2300	+ * plugins' data), so there only the chunk's own entries are deleted. A
	2301	+ * persistent cache without `flush_runtime` is left alone: deleting from it
	2302	+ * would evict the objects for every other request too.
	2303	+ *
	2304	+ * @since 2.10.1
	2305	+ * @param int $chunk_start memory_get_usage(true) before the chunk was hydrated.
	2306	+ * @param array $groups Cache groups keyed by the chunk's object IDs.
	2307	+ * @param int[] $ids The chunk's object IDs, plus any objects it
	2308	+ * hydrated under their own (featured images).
	2309	+ * @return void
	2310	+ * @throws \Error See assert_memory_headroom().
	2311	+ */
	2312	+ private function release_walk_memory(int $chunk_start, array $groups, array $ids): void {
	2313	+ // What one chunk costs before it is released: the margin the next one
	2314	+ // needs. Measured in the same real allocated size assert_memory_headroom()
	2315	+ // compares against the limit, so the two are the same unit.
	2316	+ $this->walk_chunk_cost = max($this->walk_chunk_cost, memory_get_usage(true) - $chunk_start);
	2317	+
	2318	+ if (wp_using_ext_object_cache()) {
	2319	+ if (
	2320	+ function_exists('wp_cache_supports')
	2321	+ && wp_cache_supports('flush_runtime')
	2322	+ && function_exists('wp_cache_flush_runtime')
	2323	+ ) {
	2324	+ wp_cache_flush_runtime();
	2325	+ }
	2326	+ } elseif (!empty($ids)) {
	2327	+ foreach ($groups as $group) {
	2328	+ wp_cache_delete_multiple($ids, $group);
	2329	+ }
	2330	+ }
	2331	+
	2332	+ $this->assert_memory_headroom();
	2333	+ }
	2334	+
	2335	+ /**
	2336	+ * Cache groups get_posts() fills per post for the given post types.
	2337	+ *
	2338	+ * The post, its meta, and one relationships group per taxonomy the post
	2339	+ * type uses (update_object_term_cache()). Term objects themselves are
	2340	+ * bounded by the number of terms, not posts, so they are left cached.
	2341	+ *
	2342	+ * @since 2.10.1
	2343	+ * @param string[] $post_types Post types being walked.
	2344	+ * @return string[] Cache groups keyed by post ID.
	2345	+ */
	2346	+ private function chunk_post_cache_groups(array $post_types): array {
	2347	+ $groups = ['posts', 'post_meta'];
	2348	+
	2349	+ foreach (get_object_taxonomies($post_types) as $taxonomy) {
	2350	+ $groups[] = $taxonomy . '_relationships';
	2351	+ }
	2352	+
	2353	+ return array_values(array_unique($groups));
	2354	+ }
	2355	+
	2356	+ /**
	2357	+ * Stop an automatic rebuild before the memory limit rather than at it.
	2358	+ *
	2359	+ * A PHP memory fatal skips every catch and finally, so the lock, the
	2360	+ * failure record and the backoff are all lost with it, while stopping here
	2361	+ * is an ordinary, fully recorded failure. Checked before each chunk is
	2362	+ * hydrated, against a margin of at least the largest chunk seen so far.
	2363	+ *
	2364	+ * It throws an \Error, not an \Exception, on purpose: the per-segment
	2365	+ * catch (\Exception) blocks in generate_multiple_sitemaps() would otherwise
	2366	+ * swallow it and carry on — writing an index without the aborted segments
	2367	+ * and then pruning their files as orphans. Only the automatic rebuild's
	2368	+ * catch (\Throwable) is meant to see it.
	2369	+ *
	2370	+ * @since 2.10.1
	2371	+ * @return void
	2372	+ * @throws \Error When the automatic rebuild is close to the memory limit.
	2373	+ */
	2374	+ private function assert_memory_headroom(): void {
	2375	+ if (!$this->memory_guard) {
	2376	+ return;
	2377	+ }
	2378	+
	2379	+ $limit = $this->current_memory_limit();
	2380	+ if ($limit === -1) {
	2381	+ return;
	2382	+ }
	2383	+
	2384	+ // A fifth of the limit (at least 32 MB) for writing the files, or one
	2385	+ // and a half of the costliest chunk if that is more — and never more
	2386	+ // than half the limit either way. Without that outer cap a single
	2387	+ // anomalously expensive chunk (500 posts of serialised page-builder or
	2388	+ // ACF meta reaches hundreds of megabytes) puts the margin above the
	2389	+ // limit itself, so every later check aborts at any usage at all, the
	2390	+ // failure records this process's limit, and has_memory_for_retry()
	2391	+ // then refuses every process that has the same limit. A site that
	2392	+ // never actually ran out of memory would stop rebuilding until WP-CLI
	2393	+ // or a system cron happened to run.
	2394	+ $headroom = (int) min(
	2395	+ max(
	2396	+ min(max($limit * 0.2, 32 * MB_IN_BYTES), $limit * 0.5),
	2397	+ $this->walk_chunk_cost * 1.5
	2398	+ ),
	2399	+ $limit * 0.5
	2400	+ );
	2401	+
	2402	+ // The real allocated size, which is what PHP enforces memory_limit
	2403	+ // against; memory_get_usage(false) reports only what is handed out of
	2404	+ // those allocations and so understates the margin by the allocator's
	2405	+ // slack.
	2406	+ $usage = memory_get_usage(true);
	2407	+
	2408	+ if ($usage > $limit - $headroom) {
	2409	+ throw new \Error(
	2410	+ sprintf(
	2411	+ /* translators: 1: memory in use, 2: PHP memory limit. */
	2412	+ __('The sitemap rebuild was stopped at %1$s of the %2$s PHP memory limit, before PHP would have run out of memory. It will be retried by a process with more memory (WP-CLI or a system cron). To let it finish in the admin, raise the PHP memory_limit.', 'thinkrank'),
	2413	+ size_format($usage),
	2414	+ size_format($limit)
	2415	+ ),
	2416	+ self::MEMORY_ABORT_CODE
	2417	+ );
	2418	+ }
	2419	+ }
	2420	+
	2421	+ /**
	2422	+ * Run an automatic rebuild's generation with the fatal-safe bookkeeping.
	2423	+ *
	2424	+ * @since 2.10.1
	2425	+ * @param array $settings Sitemap settings.
	2426	+ * @return bool Whatever generate_and_save() returned.
	2427	+ * @throws \Throwable Whatever generation throws, after the memory guard is
	2428	+ * switched back off.
	2429	+ */
	2430	+ private function generate_for_regeneration(array $settings): bool {
	2431	+ // Same headroom wp-admin gives itself; a no-op when the limit is
	2432	+ // already higher or unlimited.
	2433	+ wp_raise_memory_limit('admin');
	2434	+
	2435	+ $this->claim_regeneration_attempt();
	2436	+ $this->memory_guard = true;
	2437	+ $this->walk_chunk_cost = 0;
	2438	+
	2439	+ try {
	2440	+ return $this->generate_and_save($settings);
	2441	+ } finally {
	2442	+ $this->memory_guard = false;
	2443	+ }
	2444	+ }
	2445	+
	2446	+ /**
	2447	+ * Record a failure thrown by an automatic rebuild.
	2448	+ *
	2449	+ * @since 2.10.1
	2450	+ * @param \Throwable $e What was thrown.
	2451	+ * @param string $source Either 'content' or 'settings'.
	2452	+ * @return void
	2453	+ */
	2454	+ private function record_thrown_regeneration_failure(\Throwable $e, string $source): void {
	2455	+ $memory_limit = null;
	2456	+
	2457	+ if ($e instanceof \Error && $e->getCode() === self::MEMORY_ABORT_CODE) {
	2458	+ $memory_limit = $this->current_memory_limit();
	2459	+ $memory_limit = $memory_limit > 0 ? $memory_limit : null;
	2460	+ }
	2461	+
	2462	+ $this->record_regeneration_failure($e->getMessage(), $source, $memory_limit);
	2463	+ }
	2464	+
	2465	+ /**
	2466	+ * Report how automatic regeneration is faring, for the admin UI.
	2467	+ *
	2468	+ * The feature used to fail invisibly: `last_generated` simply stopped
	2469	+ * advancing and nothing drew attention to it (#629).
	2470	+ *
	2471	+ * @since 2.2.1
	2472	+ * @return array Health payload.
	2473	+ */
	2474	+ public function get_regeneration_health(): array {
	2475	+ $settings = $this->get_settings('site');
	2476	+ $pending = get_option(self::REGENERATION_PENDING_OPTION, []);
	2477	+ $pending = is_array($pending) ? $pending : [];
	2478	+ $error = get_option(self::REGENERATION_ERROR_OPTION, []);
	2479	+ $error = is_array($error) ? $error : [];
	2480	+
	2481	+ $since = !empty($pending['since']) ? (int) $pending['since'] : 0;
	2482	+
	2483	+ $next_scheduled = wp_next_scheduled('thinkrank_regenerate_sitemap');
	2484	+ if ($next_scheduled === false) {
	2485	+ $next_scheduled = wp_next_scheduled('thinkrank_regenerate_sitemap_settings');
	2486	+ }
	2487	+
	2488	+ return [
	2489	+ 'auto_generate' => !empty($settings['auto_generate']),
	2490	+ 'last_generated' => $settings['last_generated'] ?? '',
	2491	+ 'pending_since' => $since ? gmdate('c', $since) : null,
	2492	+ 'pending_seconds' => $since ? max(0, time() - $since) : 0,
	2493	+ 'next_scheduled' => $next_scheduled ? gmdate('c', (int) $next_scheduled) : null,
	2494	+ 'cron_disabled' => defined('DISABLE_WP_CRON') && DISABLE_WP_CRON,
	2495	+ 'stale' => $this->is_sitemap_stale($settings),
	2496	+ 'last_error' => !empty($error['message'])
	2497	+ ? [
	2498	+ 'message' => (string) $error['message'],
	2499	+ 'source' => isset($error['source']) ? (string) $error['source'] : 'content',
	2500	+ 'time' => !empty($error['time']) ? gmdate('c', (int) $error['time']) : null,
	2501	+ ]
	2502	+ : null,
	2503	+ ];
	2504	+ }
	2505	+
	2506	+ /**
	2507	+ * Has published content changed since the served sitemap was last written?
	2508	+ *
	2509	+ * Uses core's cached last-modified lookup, which only considers published
	2510	+ * posts — the same content the sitemap covers.
	2511	+ *
	2512	+ * @since 2.2.1
	2513	+ * @param array $settings Sitemap settings.
	2514	+ * @return bool True when the sitemap is behind the content.
	2515	+ */
	2516	+ private function is_sitemap_stale(array $settings): bool {
	2517	+ if (empty($settings['enabled']) || empty($settings['last_generated'])) {
	2518	+ // Never generated is already reported separately by the UI.
	2519	+ return false;
	2520	+ }
	2521	+
	2522	+ $generated = strtotime((string) $settings['last_generated']);
	2523	+ if (!$generated) {
	2524	+ return false;
	2525	+ }
	2526	+
	2527	+ $modified = get_lastpostmodified('gmt');
	2528	+ if (!$modified) {
	2529	+ return false;
	2530	+ }
	2531	+
	2532	+ $modified = strtotime($modified . ' UTC');
	2533	+ if (!$modified) {
	2534	+ return false;
	2535	+ }
	2536	+
	2537	+ // A minute of slack keeps a rebuild that ran alongside the edit from
	2538	+ // reporting itself as stale.
	2539	+ return $modified > ($generated + MINUTE_IN_SECONDS);
	2540	+ }
	2541	+
	2542	+ /**
	2543	+ * Public entry point to debounce-rebuild the sitemap after a settings change
	2544	+ * (e.g. toggling inclusion rules via REST or the MCP ability), so the served
	2545	+ * file reflects the new settings instead of going stale until a content edit.
	2546	+ *
	2547	+ * @return void
	2548	+ */
	2549	+ public function schedule_regeneration(): void {
	2550	+ // Debounce against rapid successive saves, but use the settings-specific
	2551	+ // hook so the rebuild runs regardless of the auto_generate toggle (which
	2552	+ // only governs content-change-triggered regeneration).
	2553	+ $this->mark_regeneration_pending('settings');
	2554	+ $this->debounce_event('thinkrank_regenerate_sitemap_settings');
	2555	+ }
	2556	+
	2557	+ /**
	2558	+ * Rebuild the served sitemap after an explicit settings change.
	2559	+ *
	2560	+ * Unlike {@see auto_regenerate_sitemap()}, this is NOT gated on the
	2561	+ * auto_generate setting: the user deliberately changed inclusion rules and
	2562	+ * expects the served file to reflect them even if content-triggered
	2563	+ * auto-generation is turned off. Still respects the master `enabled` flag.
	2564	+ *
	2565	+ * @since 2.10.0 Reports whether the served sitemap was actually rebuilt, so
	2566	+ * a caller can say so rather than assume it (#764). Existing
	2567	+ * callers that ignore the return are unaffected.
	2568	+ *
	2569	+ * @return bool True when the served sitemap now reflects the settings.
	2570	+ */
	2571	+ public function regenerate_sitemap_from_settings(): bool {
	2572	+ if (!$this->acquire_generation_lock()) {
	2573	+ // A manual generation (or another request's takeover) is already
	2574	+ // writing the files; the pending marker survives so this rebuild is
	2575	+ // retried rather than lost.
	2576	+ return false;
	2577	+ }
	2578	+
	2579	+ try {
	2580	+ $settings = $this->get_settings('site');
	2581	+ if (empty($settings['enabled'])) {
	2582	+ // The sitemap was disabled: remove the previously generated static
	2583	+ // files so the web server stops serving a stale sitemap that
	2584	+ // crawlers would otherwise keep fetching.
	2585	+ //
	2586	+ // A file that could not be removed is still being served, so
	2587	+ // this is not a success. Reporting one here would tell a caller
	2588	+ // the sitemap was gone while the web server kept answering with
	2589	+ // it, which is the failure this return value exists to prevent
	2590	+ // (#764).
	2591	+ $removal = $this->delete_published_sitemaps($settings);
	2592	+ $stuck = is_array($removal['failed'] ?? null) ? $removal['failed'] : [];
	2593	+
	2594	+ if (!empty($stuck)) {
	2595	+ $this->record_regeneration_failure(
	2596	+ $this->stuck_files_message($stuck, true),
	2597	+ 'settings'
	2598	+ );
	2599	+
	2600	+ return false;
	2601	+ }
	2602	+
	2603	+ $this->mark_regeneration_complete();
	2604	+
	2605	+ return true;
	2606	+ }
	2607	+
	2608	+ $revision = $this->current_regeneration_revision();
	2609	+
	2610	+ if ('dynamic' === $this->resolve_delivery_mode($settings)) {
	2611	+ // Returns false when a static file is stuck in the web root:
	2612	+ // the server keeps serving that file in preference to WordPress,
	2613	+ // so the switch has not taken effect (#764).
	2614	+ return $this->switch_to_dynamic_delivery($settings, $revision, 'settings');
	2615	+ }
	2616	+
	2617	+ if ($this->generate_for_regeneration($settings)) {
	2618	+ $this->mark_regeneration_complete($revision);
	2619	+
	2620	+ return true;
	2621	+ }
	2622	+
	2623	+ $this->record_regeneration_failure(
	2624	+ $this->write_failure_message(),
	2625	+ 'settings'
	2626	+ );
	2627	+
	2628	+ return false;
	2629	+ } catch (\Throwable $e) {
	2630	+ $this->record_thrown_regeneration_failure($e, 'settings');
	2631	+
	2632	+ return false;
	2633	+ } finally {
	2634	+ $this->release_generation_lock();
	2635	+ }
	2636	+ }
	2637	+
	2638	+ /**
	2639	+ * When a rebuild has been outstanding since, or 0 when none is.
	2640	+ *
	2641	+ * Lets a caller report an honest "saved, but the served file has not caught
	2642	+ * up yet" instead of a bare success (#764).
	2643	+ *
	2644	+ * @since 2.10.0
	2645	+ * @return int Unix timestamp, or 0 when nothing is pending.
	2646	+ */
	2647	+ public static function regeneration_pending_since(): int {
	2648	+ $pending = get_option(self::REGENERATION_PENDING_OPTION, []);
	2649	+
	2650	+ if (!is_array($pending) || empty($pending['since'])) {
	2651	+ return 0;
	2652	+ }
	2653	+
	2654	+ return (int) $pending['since'];
	2655	+ }
	2656	+
	2657	+ /**
	2658	+ * Remove every static sitemap file ThinkRank publishes to the web root.
	2659	+ *
	2660	+ * Called when the sitemap feature is disabled, by the cleanup route, and by
	2661	+ * both removal paths, so /sitemap.xml, /sitemap_index.xml, the segmented
	2662	+ * children (incl. paginated -N pages), and /local-sitemap.xml stop being
	2663	+ * served. Only ThinkRank's own filenames are targeted; WordPress core's
	2664	+ * wp-sitemap.xml and any other plugin's sitemap in the web root are left
	2665	+ * untouched.
	2666	+ *
	2667	+ * @since 1.31.0 Returns the filenames removed, and accepts the settings to
	2668	+ * derive them from, so a caller that already read them (and
	2669	+ * needs to report what went) does not have to re-read or
	2670	+ * re-derive the name list.
	2671	+ * @since 2.1.0 Delegates to thinkrank_webroot_delete_sitemaps(). Uninstall
	2672	+ * needs the same removal but has no autoloader to reach this
	2673	+ * class, so the logic moved to includes/cleanup-webroot.php
	2674	+ * and this stays as the in-plugin entry point.
	2675	+ *
	2676	+ * @param array|null $settings Optional. Sitemap settings; defaults to the
	2677	+ * saved site settings.
	2678	+ * @return array{deleted: string[], failed: string[]} Basenames removed, and
	2679	+ * those that existed but could not be removed.
	2680	+ */
	2681	+ public function delete_published_sitemaps(?array $settings = null): array {
	2682	+ return thinkrank_webroot_delete_sitemaps($settings ?? $this->get_settings('site'));
	2683	+ }
	2684	+
	2685	+ /**
	2686	+ * Every child-sitemap filename this site could have published.
	2687	+ *
	2688	+ * @since 1.31.0
	2689	+ * @since 2.1.0 Delegates to thinkrank_webroot_segment_filenames().
	2690	+ *
	2691	+ * @param array $settings Sitemap settings (read for `custom_url_pattern`).
	2692	+ * @return string[] Basenames, e.g. ['sitemap-posts.xml', 'sitemap-pages.xml'].
	2693	+ */
	2694	+ private function publishable_segment_filenames(array $settings): array {
	2695	+ return thinkrank_webroot_segment_filenames($settings);
	2696	+ }
	2697	+
	2698	+ /**
	2699	+ * Can this web-root file be shown to be a sitemap ThinkRank wrote?
	2700	+ *
	2701	+ * The generator deletes as often as cleanup does — a segment that dropped
	2702	+ * out of the set, a pagination page beyond the new count, the local sitemap
	2703	+ * after the business identity was cleared — and until 2.1.1 it did all
	2704	+ * three by filename alone. That is the #515 bug on a far more frequent
	2705	+ * trigger: our names are the canonical ones, so an ordinary regeneration
	2706	+ * (post save, term change, settings save) destroyed RankMath's
	2707	+ * `sitemap-tags.xml` and `local-sitemap.xml` with no deactivation involved.
	2708	+ *
	2709	+ * Every name the generator derives comes from the current settings — the
	2710	+ * url pattern, the configured `sitemap_urls`, `local-sitemap.xml` — but the
	2711	+ * ownership test is still asked with `$name_derived = false`, which switches
	2712	+ * off the legacy fallback for the whole generator side.
	2713	+ *
	2714	+ * The fallback exists to recover a pre-2.1.1 file written with
	2715	+ * `enable_styling` off, which carries neither marker. That recovery belongs
	2716	+ * to the once-off cleanup paths. Here it can only do harm: this method runs
	2717	+ * on every post save, and everything this version writes carries
	2718	+ * THINKRANK_SITEMAP_MARKER, so after the site's first regeneration an
	2719	+ * unmarked file at one of our names is by definition somebody else's — and
	2720	+ * deleting it on an ordinary regeneration is #515 through the more common
	2721	+ * door. The cost is a stale unmarked segment left on disk until deactivation
	2722	+ * picks it up, which is the safe direction to fail in.
	2723	+ *
	2724	+ * @since 2.1.1
	2725	+ *
	2726	+ * @param string $path Absolute path to a file in the web root.
	2727	+ * @param array $settings Sitemap settings.
	2728	+ * @return bool True when the file may be deleted.
	2729	+ */
	2730	+ private function webroot_sitemap_is_ours(string $path, array $settings): bool {
	2731	+ return thinkrank_webroot_sitemap_is_ours($path, $settings, false);
	2732	+ }
	2733	+
	2734	+ /**
1112	2735	* Auto-regenerate sitemap (called by scheduled action)
1113	2736	*
1114	2737	* @since 1.0.0
1115	2738	* @return void
		@@ -1114,37 +2737,728 @@
1114	2737	* @since 1.0.0
1115	2738	* @return void
1116	2739	*/
1117	2740	public function auto_regenerate_sitemap(): void {
	2741	+ if (!$this->acquire_generation_lock()) {
	2742	+ // A manual generation (or another request's takeover) is already
	2743	+ // writing the files; the pending marker survives so this rebuild is
	2744	+ // retried rather than lost.
	2745	+ return;
	2746	+ }
	2747	+
1118	2748	try {
1119	2749	// Double-check that auto-generation is still enabled
1120	2750	if (!$this->should_auto_generate()) {
	2751	+ // Nothing outstanding can be delivered while the feature is off,
	2752	+ // so drop the marker rather than let the takeover retry forever.
	2753	+ $this->mark_regeneration_complete();
1121	2754	return;
1122	2755	}
1123	2756
1124		- // Generate sitemap with current settings
	2757	+ $revision = $this->current_regeneration_revision();
1125	2758	$settings = $this->get_settings('site');
1126		- $sitemap_xml = $this->generate_sitemap($settings);
1127	2759
1128		- // Save sitemap to file
1129		- $this->save_sitemap_to_file($sitemap_xml);
	2760	+ // See regenerate_sitemap_from_settings(): nothing to write.
	2761	+ if ('dynamic' === $this->resolve_delivery_mode($settings)) {
	2762	+ $this->switch_to_dynamic_delivery($settings, $revision, 'content');
	2763	+ return;
	2764	+ }
1130	2765
1131		- // Update last generated timestamp
1132		- $this->update_setting('last_generated', gmdate('c'), 'site');
	2766	+ if ($this->generate_for_regeneration($settings)) {
	2767	+ $this->mark_regeneration_complete($revision);
	2768	+ } else {
	2769	+ // Previously this returned quietly and last_generated simply
	2770	+ // stopped advancing, leaving the site owner with no way to learn
	2771	+ // the sitemap had stopped updating (#629).
	2772	+ $this->record_regeneration_failure(
	2773	+ $this->write_failure_message(),
	2774	+ 'content'
	2775	+ );
	2776	+ }
	2777	+ } catch (\Throwable $e) {
	2778	+ $this->record_thrown_regeneration_failure($e, 'content');
	2779	+ } finally {
	2780	+ $this->release_generation_lock();
	2781	+ }
	2782	+ }
1133	2783
1134		- // Ping search engines if enabled
1135		- if (!empty($settings['ping_search_engines'])) {
1136		- $this->ping_search_engines();
	2784	+ /**
	2785	+ * Build and write the sitemap files for the given settings.
	2786	+ *
	2787	+ * Segmented files plus an index when the settings configure one, a single
	2788	+ * file otherwise, and the standalone local business sitemap either way.
	2789	+ * Records `last_generated` so the UI's "View Generated Sitemaps" links
	2790	+ * unlock, mirroring the manual generate endpoint.
	2791	+ *
	2792	+ * @since 1.17.0
	2793	+ * @param array $settings Sitemap settings.
	2794	+ * @return bool True when the sitemap files were written.
	2795	+ */
	2796	+ public function generate_and_save(array $settings): bool {
	2797	+ // Dynamic delivery publishes no files, so writing them here would put a
	2798	+ // static copy back in the web root for the server to serve in place of
	2799	+ // the dynamic route. Guarding at each call site left gaps — the
	2800	+ // snapshot migrator's post-import regeneration had none — so the rule
	2801	+ // lives with the writing instead.
	2802	+ //
	2803	+ // `is_collecting()` is the exception that makes dynamic delivery work
	2804	+ // at all: render_document() and collect_documents() reach this same
	2805	+ // method with the writer swapped for a collector, and that is precisely
	2806	+ // the dynamic build. Only a real write is skipped.
	2807	+ if (!$this->is_collecting() && 'dynamic' === $this->resolve_delivery_mode($settings)) {
	2808	+ // Whatever prompted this call changed the sitemap's content, so the
	2809	+ // rendered copies must not outlive it.
	2810	+ $this->flush_dynamic_cache();
	2811	+
	2812	+ return true;
	2813	+ }
	2814	+
	2815	+ // Index mode is driven by the use_sitemap_index toggle (not merely by how
	2816	+ // many sitemap_urls happen to be configured). When the toggle is on but
	2817	+ // no child sitemaps are set up yet, synthesize the per-type segmented set
	2818	+ // so we emit a real <sitemapindex> with paginated children instead of a
	2819	+ // flat urlset misnamed sitemap_index.xml.
	2820	+ $settings = $this->maybe_promote_to_index($settings);
	2821	+
	2822	+ if (!empty($settings['use_sitemap_index']) || count((is_array($settings['sitemap_urls'] ?? null) ? $settings['sitemap_urls'] : [])) > 1) {
	2823	+ $results = $this->generate_multiple_sitemaps($settings);
	2824	+ $written = !empty($results['success']);
	2825	+ } else {
	2826	+ $xml = $this->generate_sitemap($settings);
	2827	+ $primary = $this->get_primary_sitemap_filename($settings);
	2828	+ $written = $this->save_sitemap_to_file($xml, $primary);
	2829	+
	2830	+ // Local business sitemap is a standalone file, regenerated on the
	2831	+ // single-sitemap path too (this is the default mode).
	2832	+ $this->regenerate_local_sitemap($settings);
	2833	+
	2834	+ // Switching out of index mode leaves sitemap_index.xml and every
	2835	+ // child on disk, still served and never refreshed again. The index
	2836	+ // path already prunes what it no longer owns; this path never did,
	2837	+ // so the site kept serving two sitemap trees (#563). Ownership is
	2838	+ // still tested per file, so another plugin's sitemap at one of our
	2839	+ // names is never touched (#515).
	2840	+ $this->prune_orphaned_segments($settings, [['filename' => basename($primary)]]);
	2841	+ }
	2842	+
	2843	+ // last_generated describes what is on disk. A dynamic render publishes
	2844	+ // nothing, so advancing it would report a static publication that never
	2845	+ // happened and would let primary_sitemap_file_exists() callers believe
	2846	+ // there is a file to serve.
	2847	+ if ($written && !$this->is_collecting()) {
	2848	+ $settings['last_generated'] = gmdate('c');
	2849	+ $this->save_settings('site', null, $settings);
	2850	+ }
	2851	+
	2852	+ return $written;
	2853	+ }
	2854	+
	2855	+ /**
	2856	+ * Whether this instance is rendering documents rather than publishing them.
	2857	+ *
	2858	+ * @since 2.9.0
	2859	+ *
	2860	+ * @return bool
	2861	+ */
	2862	+ private function is_collecting(): bool {
	2863	+ return $this->document_sink !== null;
	2864	+ }
	2865	+
	2866	+ /**
	2867	+ * Complete a regeneration that delivers dynamically, retiring stale files.
	2868	+ *
	2869	+ * Dynamic delivery renders nothing to disk, but that is only half the job.
	2870	+ * A web server hands back an existing `/sitemap.xml` without ever loading
	2871	+ * WordPress, so any file left over from a previous static generation goes on
	2872	+ * being served forever and {@see \ThinkRank\Frontend\SEO_Manager
	2873	+ * ::maybe_serve_sitemap()} is never reached. Switching to dynamic while
	2874	+ * leaving those files in place would therefore appear to do nothing at all.
	2875	+ *
	2876	+ * Both transitions matter and they differ:
	2877	+ *
	2878	+ * - An explicit switch to `dynamic` happens on a site whose root is usually
	2879	+ * still writable, so the files can simply be removed.
	2880	+ * - An `auto` site that becomes read-only cannot remove them, because
	2881	+ * deleting an entry needs write permission on the directory that holds
	2882	+ * it. There the stale sitemap really is stuck in front of us, and the
	2883	+ * honest outcome is a recorded failure naming it rather than a rebuild
	2884	+ * reported as complete (#754 review).
	2885	+ *
	2886	+ * Ownership is tested per file by the shared helper, so another plugin's
	2887	+ * sitemap at one of our names is never deleted (#515).
	2888	+ *
	2889	+ * @since 2.9.0
	2890	+ *
	2891	+ * @param array $settings Sitemap settings.
	2892	+ * @param int $revision Revision this rebuild is completing.
	2893	+ * @param string $source 'settings' or 'content', for the failure record.
	2894	+ * @return void
	2895	+ */
	2896	+ private function switch_to_dynamic_delivery(array $settings, int $revision, string $source): bool {
	2897	+ $this->flush_dynamic_cache();
	2898	+
	2899	+ $removal = $this->delete_published_sitemaps($settings);
	2900	+ $stuck = is_array($removal['failed'] ?? null) ? $removal['failed'] : [];
	2901	+
	2902	+ if (!empty($stuck)) {
	2903	+ $this->record_regeneration_failure($this->stuck_files_message($stuck), $source);
	2904	+
	2905	+ return false;
	2906	+ }
	2907	+
	2908	+ $this->mark_regeneration_complete($revision);
	2909	+
	2910	+ return true;
	2911	+ }
	2912	+
	2913	+ /**
	2914	+ * Why a stale file left in the web root means the change has not landed.
	2915	+ *
	2916	+ * Shared by every path that removes published files, so they cannot
	2917	+ * describe the same situation differently (#764).
	2918	+ *
	2919	+ * The two situations that reach it differ in what WordPress is doing, and
	2920	+ * the message has to say which. After a switch to dynamic delivery
	2921	+ * WordPress IS serving the sitemap and the files shadow it. After the
	2922	+ * sitemap is switched off WordPress serves nothing, so the one message
	2923	+ * used to tell a site owner who had just disabled the sitemap that it was
	2924	+ * "being served from WordPress", which is the opposite of what they did.
	2925	+ *
	2926	+ * @since 2.10.0
	2927	+ * @since 2.10.0 Public, so the REST endpoint uses it rather than a copy;
	2928	+ * takes $sitemap_disabled for the disabled path.
	2929	+ *
	2930	+ * @param string[] $stuck Basenames that could not be removed.
	2931	+ * @param bool $sitemap_disabled True when the files outlived disabling
	2932	+ * the sitemap rather than a switch to
	2933	+ * dynamic delivery.
	2934	+ * @return string
	2935	+ */
	2936	+ public function stuck_files_message(array $stuck, bool $sitemap_disabled = false): string {
	2937	+ if ($sitemap_disabled) {
	2938	+ return sprintf(
	2939	+ /* translators: 1: comma-separated file names, 2: absolute path to the WordPress root. */
	2940	+ __('The sitemap is disabled, but these files are still in the site root and your web server is still serving them: %1$s. They could not be removed because %2$s is not writable. Delete them, or ask your host to make the WordPress root writable.', 'thinkrank'),
	2941	+ implode(', ', $stuck),
	2942	+ untrailingslashit(ABSPATH)
	2943	+ );
	2944	+ }
	2945	+
	2946	+ return sprintf(
	2947	+ /* translators: 1: comma-separated file names, 2: absolute path to the WordPress root. */
	2948	+ __('The sitemap is being served from WordPress, but these files are still in the site root and your web server will keep serving them instead: %1$s. They could not be removed because %2$s is not writable. Delete them, or ask your host to make the WordPress root writable.', 'thinkrank'),
	2949	+ implode(', ', $stuck),
	2950	+ untrailingslashit(ABSPATH)
	2951	+ );
	2952	+ }
	2953	+
	2954	+ /**
	2955	+ * What to tell the site owner when publishing the files failed.
	2956	+ *
	2957	+ * The old wording stated the symptom and stopped there, so the reported
	2958	+ * cause was a guess and this reached support as a plugin fault rather than
	2959	+ * a folder permission (#752, #753). When the root is demonstrably
	2960	+ * unwritable, say that, and say what to do about it.
	2961	+ *
	2962	+ * @since 2.9.0
	2963	+ *
	2964	+ * @return string
	2965	+ */
	2966	+ private function write_failure_message(): string {
	2967	+ if (!wp_is_writable(ABSPATH)) {
	2968	+ return sprintf(
	2969	+ /* translators: %s: absolute path to the WordPress root. */
	2970	+ __('The sitemap could not be written because the folder %s is not writable by PHP. Ask your host to make the WordPress root writable, or set Sitemap Delivery to Dynamic to serve the sitemap without writing files.', 'thinkrank'),
	2971	+ untrailingslashit(ABSPATH)
	2972	+ );
	2973	+ }
	2974	+
	2975	+ return __('The sitemap files could not be written to the site root.', 'thinkrank');
	2976	+ }
	2977	+
	2978	+ /**
	2979	+ * How this site delivers its sitemap.
	2980	+ *
	2981	+ * `auto` is resolved on whether the web root can be written. That is the
	2982	+ * right signal here (unlike llms.txt, where the question is whether the
	2983	+ * server applies the .htaccess charset block): a site whose root is
	2984	+ * read-only cannot publish a sitemap file at all, and before this existed
	2985	+ * the feature simply failed with "The sitemap files could not be written to
	2986	+ * the site root." and served nothing (#752).
	2987	+ *
	2988	+ * @since 2.9.0
	2989	+ *
	2990	+ * @param array|null $settings Sitemap settings (falls back to saved ones).
	2991	+ * @return string One of 'static' or 'dynamic'. Never 'auto'.
	2992	+ */
	2993	+ public function resolve_delivery_mode(?array $settings = null): string {
	2994	+ $settings = $settings ?? $this->get_settings('site');
	2995	+ $mode = (string) ($settings['delivery_mode'] ?? 'auto');
	2996	+
	2997	+ if ('static' === $mode || 'dynamic' === $mode) {
	2998	+ return $mode;
	2999	+ }
	3000	+
	3001	+ return wp_is_writable(ABSPATH) ? 'static' : 'dynamic';
	3002	+ }
	3003	+
	3004	+ /**
	3005	+ * Render one published sitemap document without touching the filesystem.
	3006	+ *
	3007	+ * Runs the ordinary build pipeline with the writer swapped for a collector,
	3008	+ * so the bytes returned here are the bytes the static path would have
	3009	+ * written. `SitemapDeliveryParityTest` asserts that equivalence rather than
	3010	+ * trusting it.
	3011	+ *
	3012	+ * The whole set is built to answer for one file, because the index can only
	3013	+ * be assembled from the children that were actually produced. The result is
	3014	+ * cached per document, so that cost is paid once per change and not once
	3015	+ * per crawler request.
	3016	+ *
	3017	+ * @since 2.9.0
	3018	+ *
	3019	+ * @param string $filename Published file name, e.g. 'sitemap.xml'.
	3020	+ * @param array|null $settings Sitemap settings (falls back to saved ones).
	3021	+ * @return string|null XML, or null when this site does not publish that name.
	3022	+ */
	3023	+ public function render_document(string $filename, ?array $settings = null): ?string {
	3024	+ $filename = basename($filename);
	3025	+ $settings = $settings ?? $this->get_settings('site');
	3026	+
	3027	+ if (empty($settings['enabled'])) {
	3028	+ return null;
	3029	+ }
	3030	+
	3031	+ $cached = get_transient($this->dynamic_cache_key($filename));
	3032	+ if (self::ABSENT_MARKER === $cached) {
	3033	+ return null;
	3034	+ }
	3035	+ if (is_string($cached) && '' !== $cached) {
	3036	+ return $cached;
	3037	+ }
	3038	+
	3039	+ // A miss builds the whole set, because the index can only be assembled
	3040	+ // from the children that were actually produced. Caching only the
	3041	+ // requested document therefore made a crawler walking the index and its
	3042	+ // children rebuild the entire site's sitemap once per file — every post
	3043	+ // and taxonomy query repeated N times on a public endpoint (#754
	3044	+ // review). The set is built once and stored in full.
	3045	+ return $this->stream_documents($settings, $filename);
	3046	+ }
	3047	+
	3048	+ /**
	3049	+ * Build every document, caching each as it is produced, keeping one.
	3050	+ *
	3051	+ * A miss has to build the whole set, because the index can only be
	3052	+ * assembled from the children that were actually produced. It does not have
	3053	+ * to *hold* the whole set: the static path never keeps more than one page
	3054	+ * in memory, writing each to disk as it goes, and buffering every
	3055	+ * document's XML to return one of them undid that on the request path,
	3056	+ * where a large site's entire sitemap corpus would sit in a single PHP
	3057	+ * process (#754 review).
	3058	+ *
	3059	+ * So the sink writes each document straight to its cache entry and lets it
	3060	+ * go, retaining only the one this request is answering. Peak retention is
	3061	+ * one document, whatever the site's size.
	3062	+ *
	3063	+ * Concurrency: the first request through takes a short lock and does the
	3064	+ * work. One that finds the lock held waits a bounded moment for the winner
	3065	+ * to publish, then builds anyway, because serving a correct sitemap late
	3066	+ * beats serving none.
	3067	+ *
	3068	+ * @since 2.9.0
	3069	+ *
	3070	+ * @param array $settings Sitemap settings.
	3071	+ * @param string $wanted Document this request is answering.
	3072	+ * @return string|null XML for $wanted, or null when the site does not publish it.
	3073	+ */
	3074	+ private function stream_documents(array $settings, string $wanted): ?string {
	3075	+ $lock = self::DYNAMIC_CACHE_PREFIX . 'lock';
	3076	+
	3077	+ if (!$this->acquire_render_lock($lock)) {
	3078	+ for ($attempt = 0; $attempt < self::RENDER_LOCK_WAIT_ATTEMPTS; $attempt++) {
	3079	+ usleep(self::RENDER_LOCK_WAIT_MICROSECONDS);
	3080	+
	3081	+ $cached = get_transient($this->dynamic_cache_key($wanted));
	3082	+ if (self::ABSENT_MARKER === $cached) {
	3083	+ return null;
	3084	+ }
	3085	+ if (is_string($cached) && '' !== $cached) {
	3086	+ return $cached;
	3087	+ }
1137	3088	}
	3089	+ }
1138	3090
1139		- // Sitemap auto-regenerated successfully
	3091	+ $kept = null;
	3092	+ // Names only. Keeping the bodies here would be the very retention this
	3093	+ // method exists to avoid.
	3094	+ $produced = [];
1140	3095
1141		- } catch (\Exception $e) {
1142		- // Sitemap auto-regeneration failed - error details available in exception
	3096	+ $previous = $this->document_sink;
	3097	+ $this->document_sink = function (string $name, string $xml) use (&$kept, &$produced, $wanted): void {
	3098	+ $produced[$name] = true;
	3099	+ set_transient($this->dynamic_cache_key($name), $xml, self::DYNAMIC_CACHE_TTL);
	3100	+
	3101	+ if ($name === $wanted) {
	3102	+ $kept = $xml;
	3103	+ }
	3104	+ };
	3105	+
	3106	+ try {
	3107	+ $this->generate_and_save($settings);
	3108	+
	3109	+ // Names the configuration lists but this build did not produce get
	3110	+ // a negative entry, so asking for one again is a cache hit rather
	3111	+ // than another full rebuild.
	3112	+ $absent = $this->published_document_names($settings);
	3113	+
	3114	+ // Also the exact name this request asked for: a paginated page past
	3115	+ // the end of a stem is a legitimate request shape that the base
	3116	+ // list cannot enumerate, and without an entry it would rebuild on
	3117	+ // every hit.
	3118	+ $absent[] = $wanted;
	3119	+
	3120	+ foreach (array_unique($absent) as $name) {
	3121	+ if (!isset($produced[$name])) {
	3122	+ set_transient($this->dynamic_cache_key($name), self::ABSENT_MARKER, self::DYNAMIC_CACHE_TTL);
	3123	+ }
	3124	+ }
	3125	+ } finally {
	3126	+ $this->document_sink = $previous;
	3127	+ delete_transient($lock);
1143	3128	}
	3129	+
	3130	+ return $kept;
1144	3131	}
1145	3132
1146	3133	/**
	3134	+ * Take the render lock, if it is free.
	3135	+ *
	3136	+ * Not atomic across processes, and deliberately so: the fallback for losing
	3137	+ * a race is duplicated work, never a wrong or missing sitemap, so a
	3138	+ * heavier primitive would buy nothing here.
	3139	+ *
	3140	+ * @since 2.9.0
	3141	+ *
	3142	+ * @param string $lock Lock transient name.
	3143	+ * @return bool True when this request holds the lock.
	3144	+ */
	3145	+ private function acquire_render_lock(string $lock): bool {
	3146	+ if (false !== get_transient($lock)) {
	3147	+ return false;
	3148	+ }
	3149	+
	3150	+ set_transient($lock, time(), self::RENDER_LOCK_TTL);
	3151	+
	3152	+ return true;
	3153	+ }
	3154	+
	3155	+ /**
	3156	+ * Build every document this site publishes and return them all.
	3157	+ *
	3158	+ * Verification and tooling only. This retains the whole set in memory, so
	3159	+ * it must never be used to answer a request: {@see self::stream_documents()}
	3160	+ * is the serving path and keeps one document at a time regardless of site
	3161	+ * size (#754 review). `SitemapDeliveryParityTest` enforces that separation
	3162	+ * by failing if the request path routes back through here.
	3163	+ *
	3164	+ * @since 2.9.0
	3165	+ *
	3166	+ * @param array $settings Sitemap settings.
	3167	+ * @return array<string,string> Filename => XML.
	3168	+ */
	3169	+ public function collect_documents(array $settings): array {
	3170	+ $documents = [];
	3171	+
	3172	+ $previous = $this->document_sink;
	3173	+ $this->document_sink = static function (string $name, string $xml) use (&$documents): void {
	3174	+ $documents[$name] = $xml;
	3175	+ };
	3176	+
	3177	+ try {
	3178	+ $this->generate_and_save($settings);
	3179	+ } finally {
	3180	+ $this->document_sink = $previous;
	3181	+ }
	3182	+
	3183	+ return $documents;
	3184	+ }
	3185	+
	3186	+ /**
	3187	+ * The file names this site publishes, without building their contents.
	3188	+ *
	3189	+ * Used by the request router to decide whether a URL is ours before doing
	3190	+ * any work. Cheap: it reads the configured child list rather than querying
	3191	+ * for entries.
	3192	+ *
	3193	+ * @since 2.9.0
	3194	+ *
	3195	+ * @param array|null $settings Sitemap settings (falls back to saved ones).
	3196	+ * @return string[] File names, including paginated pages that may exist.
	3197	+ */
	3198	+ public function published_document_names(?array $settings = null): array {
	3199	+ $settings = $settings ?? $this->get_settings('site');
	3200	+ $resolved = $this->maybe_promote_to_index($settings);
	3201	+
	3202	+ $names = [$this->get_primary_sitemap_filename($settings), 'local-sitemap.xml'];
	3203	+
	3204	+ foreach ((array) ($resolved['sitemap_urls'] ?? []) as $child) {
	3205	+ if (!is_array($child) || empty($child['enabled'])) {
	3206	+ continue;
	3207	+ }
	3208	+
	3209	+ $path = (string) wp_parse_url((string) ($child['url'] ?? ''), PHP_URL_PATH);
	3210	+ if ('' !== $path) {
	3211	+ $names[] = basename($path);
	3212	+ }
	3213	+ }
	3214	+
	3215	+ return array_values(array_unique(array_filter($names)));
	3216	+ }
	3217	+
	3218	+ /**
	3219	+ * Does this site publish a document under that name?
	3220	+ *
	3221	+ * Not a plain membership test against {@see self::published_document_names()}:
	3222	+ * that lists the configured children, and a child over the per-file URL cap
	3223	+ * is split into `<stem>-2.xml`, `<stem>-3.xml` and so on, with every page
	3224	+ * listed in the index. Gating the request router on the base list alone
	3225	+ * therefore 404'd exactly the pages the index points at, which is worse than
	3226	+ * not serving them at all.
	3227	+ *
	3228	+ * Page counts are not knowable without building, so the stem is what is
	3229	+ * matched; a page that does not exist is answered by the build finding
	3230	+ * nothing for it, and is then cached as absent.
	3231	+ *
	3232	+ * @since 2.9.0
	3233	+ *
	3234	+ * @param string $name Requested file name.
	3235	+ * @param array|null $settings Sitemap settings (falls back to saved ones).
	3236	+ * @return bool
	3237	+ */
	3238	+ public function publishes_document_name(string $name, ?array $settings = null): bool {
	3239	+ $names = $this->published_document_names($settings);
	3240	+
	3241	+ if (in_array($name, $names, true)) {
	3242	+ return true;
	3243	+ }
	3244	+
	3245	+ if (!preg_match('/^(.*)-\d+\.xml$/i', $name, $m)) {
	3246	+ return false;
	3247	+ }
	3248	+
	3249	+ return in_array($m[1] . '.xml', $names, true);
	3250	+ }
	3251	+
	3252	+ /**
	3253	+ * Transient key for a rendered document.
	3254	+ *
	3255	+ * @since 2.9.0
	3256	+ *
	3257	+ * @param string $filename Published file name.
	3258	+ * @return string
	3259	+ */
	3260	+ private function dynamic_cache_key(string $filename): string {
	3261	+ return self::DYNAMIC_CACHE_PREFIX . md5($filename);
	3262	+ }
	3263	+
	3264	+ /**
	3265	+ * Drop every cached dynamic document.
	3266	+ *
	3267	+ * Called from the same places that mark the static files stale, so the two
	3268	+ * delivery modes invalidate on identical triggers.
	3269	+ *
	3270	+ * @since 2.9.0
	3271	+ *
	3272	+ * @return void
	3273	+ */
	3274	+ public function flush_dynamic_cache(): void {
	3275	+ foreach ($this->published_document_names() as $name) {
	3276	+ delete_transient($this->dynamic_cache_key($name));
	3277	+ }
	3278	+ }
	3279	+
	3280	+ /**
	3281	+ * Resolve index-vs-single mode, synthesizing child sitemaps when needed.
	3282	+ *
	3283	+ * - When use_sitemap_index is on but no child sitemaps are configured, build
	3284	+ * the per-type segmented set so a real <sitemapindex> is produced (#127).
	3285	+ * - When the toggle is off but a single flat file would exceed the per-file
	3286	+ * URL cap, auto-promote to a paginated index instead of one oversized file
	3287	+ * that can cross Google's 50k-URL/50MB limits (#129).
	3288	+ *
	3289	+ * @param array $settings Sitemap settings.
	3290	+ * @return array Possibly-updated settings.
	3291	+ */
	3292	+ public function maybe_promote_to_index(array $settings): array {
	3293	+ $saved = $this->get_settings('site');
	3294	+
	3295	+ // Read from the *payload*, before the merge below folds the saved values
	3296	+ // in: "the caller named this" and "this has a value" are different
	3297	+ // questions, and the mode resolution turns on the former.
	3298	+ $mode_supplied = array_key_exists('use_sitemap_index', $settings);
	3299	+ $urls_supplied = is_array($settings['sitemap_urls'] ?? null);
	3300	+ $has_children = count($urls_supplied ? $settings['sitemap_urls'] : []) > 1;
	3301	+
	3302	+ // Which inclusion flags did the caller actually name? The child list is
	3303	+ // the only thing that reads them, and inheriting a saved one skipped
	3304	+ // that — so on an index-mode site the include_* flags were enforced
	3305	+ // nowhere but in the browser, where SitemapGeneration.js recomputes
	3306	+ // sitemap_urls itself. Every non-UI client, the shipped
	3307	+ // `update-sitemap-settings` ability included, saved the flag and changed
	3308	+ // nothing (#398). Read from the payload for the same reason as above:
	3309	+ // after the merge every saved flag would look like one the caller named.
	3310	+ $named_inclusions = array_intersect(
	3311	+ array_keys(self::INCLUSION_CHILD_TYPES),
	3312	+ array_keys($settings)
	3313	+ );
	3314	+ $inclusions_supplied = (bool) $named_inclusions;
	3315	+
	3316	+ // Inclusion flags may be absent from a partial payload (e.g. the manual
	3317	+ // generate endpoint) — fall back to saved settings so synthesized child
	3318	+ // sitemaps reflect the real include_posts/pages/categories choices.
	3319	+ $inclusions = array_merge($saved, $settings);
	3320	+
	3321	+ // Hand the generators a *complete* settings array. Only the mode was
	3322	+ // resolved before, so every other unnamed key reached them missing: a
	3323	+ // bare `{}` from a REST/MCP client republished the sitemap with
	3324	+ // enable_styling and include_images read as off, overwriting the live
	3325	+ // files with output that had lost its XSL stylesheet, its image
	3326	+ // namespace and its image entries. Presentation and inclusion settings
	3327	+ // are not something a generate call opts into — they are the site's
	3328	+ // configuration, and only a value actually present in the payload
	3329	+ // overrides them.
	3330	+ $settings = $inclusions;
	3331	+
	3332	+ // The two keys that drive mode keep their own resolution rules below,
	3333	+ // so they must go back to "not specified" when the caller omitted them.
	3334	+ if (!$mode_supplied) {
	3335	+ unset($settings['use_sitemap_index']);
	3336	+ }
	3337	+ if (!$urls_supplied) {
	3338	+ unset($settings['sitemap_urls']);
	3339	+ }
	3340	+
	3341	+ // An absent use_sitemap_index means "not specified", which is not the
	3342	+ // same as "single file". Reading it as the latter meant a partial payload
	3343	+ // — `{}` from a REST/MCP client, or anything short of the full settings
	3344	+ // object the admin bundle sends — republished one flat sitemap.xml on an
	3345	+ // index-mode site and left sitemap_index.xml and its children stale or
	3346	+ // missing. Inherit the saved mode instead; only a value actually present
	3347	+ // in the payload decides the mode.
	3348	+ if (!array_key_exists('use_sitemap_index', $settings)) {
	3349	+ $settings['use_sitemap_index'] = $saved['use_sitemap_index'] ?? '';
	3350	+
	3351	+ // Inheriting the mode means inheriting its children too, unless the
	3352	+ // caller named its own set.
	3353	+ $saved_children = (is_array($saved['sitemap_urls'] ?? null) ? $saved['sitemap_urls'] : []);
	3354	+ if (!empty($settings['use_sitemap_index'])
	3355	+ && !$urls_supplied
	3356	+ && count($saved_children) > 1) {
	3357	+ // A named inclusion flag is applied *to* the inherited list, not
	3358	+ // used to regenerate it. build_segmented_sitemap_urls() also adds
	3359	+ // a child for every public custom post type, so rebuilding here
	3360	+ // would make `{include_pages: false}` — one thing off — silently
	3361	+ // switch on children the saved list never had (an Elementor
	3362	+ // internal CPT, a WooCommerce product feed). Only the flags the
	3363	+ // caller actually named change anything.
	3364	+ $settings['sitemap_urls'] = $inclusions_supplied
	3365	+ ? $this->apply_inclusion_flags_to_children($saved_children, $named_inclusions, $inclusions)
	3366	+ : $saved_children;
	3367	+
	3368	+ // The children are resolved either way — including when the
	3369	+ // caller switched the last one off, which leaves a bare index and
	3370	+ // is what they asked for.
	3371	+ $has_children = true;
	3372	+ }
	3373	+ }
	3374	+
	3375	+ if (!empty($settings['use_sitemap_index'])) {
	3376	+ if (!$has_children) {
	3377	+ $settings['sitemap_urls'] = $this->build_segmented_sitemap_urls($inclusions);
	3378	+ }
	3379	+ return $settings;
	3380	+ }
	3381	+
	3382	+ if (!$has_children) {
	3383	+ $limit = $this->get_links_per_sitemap($settings);
	3384	+ if ($limit > 0 && $this->estimate_total_sitemap_urls($inclusions) > $limit) {
	3385	+ $settings['use_sitemap_index'] = true;
	3386	+ $settings['sitemap_urls'] = $this->build_segmented_sitemap_urls($inclusions);
	3387	+ }
	3388	+ }
	3389	+
	3390	+ return $settings;
	3391	+ }
	3392	+
	3393	+ /**
	3394	+ * Rough count of URLs a single-file sitemap would contain, used only to
	3395	+ * decide whether to auto-promote to a paginated index. Cheap COUNT queries;
	3396	+ * intentionally approximate (homepage + published posts of enabled types +
	3397	+ * terms of enabled taxonomies).
	3398	+ *
	3399	+ * @param array $settings Sitemap settings.
	3400	+ * @return int
	3401	+ */
	3402	+ private function estimate_total_sitemap_urls(array $settings): int {
	3403	+ $total = 1; // homepage
	3404	+
	3405	+ foreach ($this->get_enabled_post_types($settings) as $post_type) {
	3406	+ $counts = wp_count_posts($post_type);
	3407	+ $total += isset($counts->publish) ? (int) $counts->publish : 0;
	3408	+ }
	3409	+
	3410	+ foreach ($this->get_enabled_taxonomies($settings) as $taxonomy) {
	3411	+ $count = wp_count_terms(['taxonomy' => $taxonomy, 'hide_empty' => true]);
	3412	+ if (!is_wp_error($count)) {
	3413	+ $total += (int) $count;
	3414	+ }
	3415	+ }
	3416	+
	3417	+ return $total;
	3418	+ }
	3419	+
	3420	+ /**
	3421	+ * Resolve the sitemap file the site publishes for the given settings.
	3422	+ *
	3423	+ * Index mode serves the index (`sitemap_index.xml`); the default single-file
	3424	+ * mode serves `sitemap.xml`. Callers use this to link to — or check for —
	3425	+ * the file the site actually serves, since ThinkRank's sitemap is a static
	3426	+ * file in the web root rather than a route.
	3427	+ *
	3428	+ * @since 1.17.0
	3429	+ * @param array $settings Sitemap settings.
	3430	+ * @return string Sitemap filename.
	3431	+ */
	3432	+ public function get_primary_sitemap_filename(array $settings): string {
	3433	+ return thinkrank_webroot_primary_sitemap_filename($settings);
	3434	+ }
	3435	+
	3436	+ /**
	3437	+ * Public URL of the sitemap the site serves.
	3438	+ *
	3439	+ * @since 1.17.0
	3440	+ * @param array|null $settings Sitemap settings (falls back to saved site settings).
	3441	+ * @return string Absolute sitemap URL.
	3442	+ */
	3443	+ public function get_primary_sitemap_url(?array $settings = null): string {
	3444	+ $settings = $settings ?? $this->get_settings('site');
	3445	+
	3446	+ return home_url('/' . $this->get_primary_sitemap_filename($settings));
	3447	+ }
	3448	+
	3449	+ /**
	3450	+ * Whether the sitemap file this site publishes exists on disk.
	3451	+ *
	3452	+ * @since 1.17.0
	3453	+ * @param array $settings Sitemap settings.
	3454	+ * @return bool True when the file is present.
	3455	+ */
	3456	+ public function primary_sitemap_file_exists(array $settings): bool {
	3457	+ return file_exists(ABSPATH . $this->get_primary_sitemap_filename($settings));
	3458	+ }
	3459	+
	3460	+ /**
1147	3461	* Save sitemap XML to file
1148	3462	*
1149	3463	* @since 1.0.0
1150	3464	* @param string $sitemap_xml Sitemap XML content
		@@ -1159,8 +3473,18 @@
1159	3473	// File validation failed - error details available in exception
1160	3474	return false;
1161	3475	}
1162	3476
	3477	+ // Dynamic delivery: hand the document to the collector instead of the
	3478	+ // filesystem. Reported as published, because for this run it is — the
	3479	+ // caller's success/failure bookkeeping and the index assembly both key
	3480	+ // off this return value.
	3481	+ if ($this->document_sink !== null) {
	3482	+ ($this->document_sink)($filename, $sitemap_xml);
	3483	+
	3484	+ return true;
	3485	+ }
	3486	+
1163	3487	$sitemap_path = ABSPATH . $filename;
1164	3488
1165	3489	// Use WordPress filesystem API for better security
1166	3490	global $wp_filesystem;
		@@ -1169,16 +3493,202 @@
1169	3493	WP_Filesystem();
1170	3494	}
1171	3495
1172	3496	if ($wp_filesystem) {
1173		- return $wp_filesystem->put_contents($sitemap_path, $sitemap_xml, FS_CHMOD_FILE);
	3497	+ $written = $wp_filesystem->put_contents($sitemap_path, $sitemap_xml, FS_CHMOD_FILE);
	3498	+
	3499	+ if ($written) {
	3500	+ // Every sitemap this version writes carries the ownership
	3501	+ // marker, so once one has been written an unmarked file at one
	3502	+ // of our names cannot be ours. Recording that retires the
	3503	+ // legacy fallback for this install — see
	3504	+ // thinkrank_webroot_sitemap_is_ours().
	3505	+ if (get_option(THINKRANK_SITEMAP_MARKED_WRITE_OPTION) !== '1') {
	3506	+ update_option(THINKRANK_SITEMAP_MARKED_WRITE_OPTION, '1', false);
	3507	+ }
	3508	+ }
	3509	+
	3510	+ return $written;
1174	3511	}
1175	3512
1176		- // Fallback to file_put_contents if WP_Filesystem fails
1177		- return file_put_contents($sitemap_path, $sitemap_xml) !== false;
	3513	+ // WP_Filesystem initialization failed
	3514	+ return false;
1178	3515	}
1179	3516
1180	3517	/**
	3518	+ * Build a segmented, index-based sitemap_urls list from inclusion settings.
	3519	+ *
	3520	+ * Produces the index entry plus one child sitemap per enabled content type
	3521	+ * (posts/pages/categories) and one per public custom post type — the shape
	3522	+ * the "complete" preset creates and that generate_multiple_sitemaps() expects.
	3523	+ * Used when enabling the index during import so the index has real children
	3524	+ * instead of being empty.
	3525	+ *
	3526	+ * @since 1.14.0
	3527	+ *
	3528	+ * @param array $inclusions Inclusion flags (include_posts/pages/categories)
	3529	+ * and optionally custom_url_pattern.
	3530	+ * @return array Sitemap URL configs
	3531	+ */
	3532	+ public function build_segmented_sitemap_urls(array $inclusions): array {
	3533	+ $pattern = $inclusions['custom_url_pattern'] ?? 'sitemap-{type}.xml';
	3534	+
	3535	+ $urls = [$this->sitemap_child_entry('/sitemap_index.xml', 'index')];
	3536	+
	3537	+ foreach (self::INCLUSION_CHILD_TYPES as $flag => $type) {
	3538	+ if (!empty($inclusions[$flag])) {
	3539	+ $urls[] = $this->build_child_sitemap_entry($type, $pattern);
	3540	+ }
	3541	+ }
	3542	+
	3543	+ // Public custom post types each get a child sitemap (parity with the
	3544	+ // "complete" preset and with Rank Math, which lists every public CPT).
	3545	+ foreach (get_post_types(['public' => true, '_builtin' => false], 'names') as $cpt) {
	3546	+ if (!$this->should_include_post_type($cpt)) {
	3547	+ continue;
	3548	+ }
	3549	+
	3550	+ // ...and the per-content-type sitemap switch (#660).
	3551	+ // get_enabled_post_types() already honours it, but this list is what
	3552	+ // index mode builds its children from — so without the same test a
	3553	+ // CPT the user had switched off still got its own child sitemap,
	3554	+ // created and streamed in full. The flags live in $inclusions, which
	3555	+ // is the settings array these children are derived from.
	3556	+ if (!\ThinkRank\SEO\Content_Type_Settings::is_included_in_sitemap('post_type', $cpt, $inclusions)) {
	3557	+ continue;
	3558	+ }
	3559	+
	3560	+ $urls[] = $this->build_child_sitemap_entry($cpt, $pattern);
	3561	+ }
	3562	+
	3563	+ // Public custom taxonomies get the same treatment (#690). Flat mode has
	3564	+ // always walked them through get_enabled_taxonomies(); index mode built
	3565	+ // its children from the list above and never consulted a taxonomy at
	3566	+ // all, so every custom-taxonomy archive silently vanished from the
	3567	+ // sitemap the moment a site switched modes — and the per-taxonomy switch
	3568	+ // the matrix writes had nothing to act on. Same two tests the post-type
	3569	+ // walk applies, in the same order.
	3570	+ $taken = array_column($urls, 'type');
	3571	+
	3572	+ foreach (get_taxonomies(['public' => true, '_builtin' => false], 'names') as $taxonomy) {
	3573	+ if (!$this->should_include_taxonomy($taxonomy)) {
	3574	+ continue;
	3575	+ }
	3576	+
	3577	+ if (!\ThinkRank\SEO\Content_Type_Settings::is_included_in_sitemap('taxonomy', $taxonomy, $inclusions)) {
	3578	+ continue;
	3579	+ }
	3580	+
	3581	+ // Post types and taxonomies are separate registries, so a site can
	3582	+ // hold both a `foo` post type and a `foo` taxonomy. They would
	3583	+ // resolve to one filename, and stream_type_entries() answers post
	3584	+ // types first, so the second child would list the first one's file
	3585	+ // twice in the index rather than adding anything.
	3586	+ if (in_array($taxonomy, $taken, true)) {
	3587	+ continue;
	3588	+ }
	3589	+
	3590	+ $urls[] = $this->build_child_sitemap_entry($taxonomy, $pattern);
	3591	+ }
	3592	+
	3593	+ return $urls;
	3594	+ }
	3595	+
	3596	+ /**
	3597	+ * Apply only the inclusion flags the caller named to an existing child list.
	3598	+ *
	3599	+ * The narrow counterpart to build_segmented_sitemap_urls(): that one
	3600	+ * regenerates the whole set from scratch, which is right when there is no set
	3601	+ * yet and wrong when there is. Rebuilding an existing list would add a child
	3602	+ * for every public custom post type it had never contained, so a payload that
	3603	+ * switches one thing off would switch others on. Here a flag adds or removes
	3604	+ * exactly its own child and leaves every other entry — custom post types,
	3605	+ * hand-added URLs, per-child enabled/status state — untouched (#398).
	3606	+ *
	3607	+ * @since 1.31.0
	3608	+ *
	3609	+ * @param array $children Existing child sitemap entries.
	3610	+ * @param string[] $named_inclusions Inclusion flag keys present in the payload.
	3611	+ * @param array $inclusions Merged settings, for the resolved flag
	3612	+ * values and custom_url_pattern.
	3613	+ * @return array Updated child sitemap entries.
	3614	+ */
	3615	+ private function apply_inclusion_flags_to_children(
	3616	+ array $children,
	3617	+ array $named_inclusions,
	3618	+ array $inclusions
	3619	+ ): array {
	3620	+ $pattern = $inclusions['custom_url_pattern'] ?? 'sitemap-{type}.xml';
	3621	+
	3622	+ foreach ($named_inclusions as $flag) {
	3623	+ $type = self::INCLUSION_CHILD_TYPES[$flag];
	3624	+
	3625	+ $present = false;
	3626	+ foreach ($children as $entry) {
	3627	+ if (($entry['type'] ?? '') === $type) {
	3628	+ $present = true;
	3629	+ break;
	3630	+ }
	3631	+ }
	3632	+
	3633	+ if (empty($inclusions[$flag])) {
	3634	+ if ($present) {
	3635	+ $children = array_values(array_filter(
	3636	+ $children,
	3637	+ static function ($entry) use ($type): bool {
	3638	+ return (is_array($entry) ? ($entry['type'] ?? '') : '') !== $type;
	3639	+ }
	3640	+ ));
	3641	+ }
	3642	+ continue;
	3643	+ }
	3644	+
	3645	+ if (!$present) {
	3646	+ $children[] = $this->build_child_sitemap_entry($type, $pattern);
	3647	+ }
	3648	+ }
	3649	+
	3650	+ return $children;
	3651	+ }
	3652	+
	3653	+ /**
	3654	+ * Build one child sitemap entry, resolving its filename from the url pattern.
	3655	+ *
	3656	+ * @since 1.31.0
	3657	+ *
	3658	+ * @param string $type Child sitemap type (posts, pages, a post type name).
	3659	+ * @param string $pattern Filename pattern containing {type}.
	3660	+ * @return array Sitemap URL config.
	3661	+ */
	3662	+ private function build_child_sitemap_entry(string $type, string $pattern): array {
	3663	+ $file = str_replace('{type}', $type, $pattern);
	3664	+ if (strpos($file, '/') !== 0) {
	3665	+ $file = '/' . $file;
	3666	+ }
	3667	+
	3668	+ return $this->sitemap_child_entry($file, $type);
	3669	+ }
	3670	+
	3671	+ /**
	3672	+ * The shape generate_multiple_sitemaps() expects of a sitemap_urls entry.
	3673	+ *
	3674	+ * @since 1.31.0
	3675	+ *
	3676	+ * @param string $url Sitemap path.
	3677	+ * @param string $type Entry type.
	3678	+ * @return array Sitemap URL config.
	3679	+ */
	3680	+ private function sitemap_child_entry(string $url, string $type): array {
	3681	+ return [
	3682	+ 'url' => $url,
	3683	+ 'type' => $type,
	3684	+ 'enabled' => true,
	3685	+ 'last_checked' => null,
	3686	+ 'status' => 'unknown',
	3687	+ ];
	3688	+ }
	3689	+
	3690	+ /**
1181	3691	* Generate multiple sitemaps based on settings
1182	3692	*
1183	3693	* @since 1.0.0
1184	3694	* @param array $settings Sitemap settings
		@@ -1191,9 +3701,9 @@
1191	3701	'errors' => [],
1192	3702	'total_urls' => 0
1193	3703	];
1194	3704
1195		- $sitemap_urls = $settings['sitemap_urls'] ?? [];
	3705	+ $sitemap_urls = (is_array($settings['sitemap_urls'] ?? null) ? $settings['sitemap_urls'] : []);
1196	3706
1197	3707	if (empty($sitemap_urls)) {
1198	3708	$results['success'] = false;
1199	3709	$results['errors'][] = 'No sitemap URLs configured';
		@@ -1199,215 +3709,663 @@
1199	3709	$results['errors'][] = 'No sitemap URLs configured';
1200	3710	return $results;
1201	3711	}
1202	3712
	3713	+ $limit = $this->get_links_per_sitemap($settings);
	3714	+
	3715	+ // Defer the index until every child sitemap has been generated, so it can
	3716	+ // list the actual files produced (including pagination pages).
	3717	+ $index_config = null;
	3718	+ $index_children = [];
	3719	+
1203	3720	foreach ($sitemap_urls as $sitemap_config) {
1204		- if (!$sitemap_config['enabled']) {
	3721	+ if (empty($sitemap_config['enabled'])) {
1205	3722	continue;
1206	3723	}
1207	3724
	3725	+ $type = $sitemap_config['type'];
	3726	+
	3727	+ if ($type === 'index') {
	3728	+ $index_config = $sitemap_config;
	3729	+ continue;
	3730	+ }
	3731	+
	3732	+ // Skip CPT children that no longer qualify (e.g. an internal store
	3733	+ // like Templately's `templately_library`) even when a previously
	3734	+ // saved config still lists them. Built-in aggregate types ('posts',
	3735	+ // 'pages', 'general', etc.) are not post type names, so this only
	3736	+ // affects real custom post types.
	3737	+ if (post_type_exists($type) && !$this->should_include_post_type($type)) {
	3738	+ continue;
	3739	+ }
	3740	+
	3741	+ // Same for the matrix switch: a child list saved before the user
	3742	+ // excluded this content type still names it, and regenerating from
	3743	+ // that list would rewrite the file they asked not to have. Built-in
	3744	+ // aggregates ('posts', 'pages', ...) are not post type names, so
	3745	+ // post_type_exists() keeps this to real custom post types.
	3746	+ if (post_type_exists($type)
	3747	+ && !\ThinkRank\SEO\Content_Type_Settings::is_included_in_sitemap('post_type', $type, $settings)) {
	3748	+ continue;
	3749	+ }
	3750	+
	3751	+ // The taxonomy counterpart of the two guards above (#690). A child
	3752	+ // list saved while a taxonomy was still included keeps naming it, so
	3753	+ // without this, excluding one in the matrix would still rewrite and
	3754	+ // re-list the file the user asked not to have. The built-in
	3755	+ // aggregates are named 'categories'/'tags' rather than
	3756	+ // 'category'/'post_tag', so taxonomy_exists() leaves them to the
	3757	+ // inclusion-flag check below.
	3758	+ $child_taxonomy = self::CHILD_TYPE_ALIASES[$type] ?? $type;
	3759	+ if (taxonomy_exists($child_taxonomy)
	3760	+ && (!$this->should_include_taxonomy($child_taxonomy)
	3761	+ || !\ThinkRank\SEO\Content_Type_Settings::is_included_in_sitemap('taxonomy', $child_taxonomy, $settings))) {
	3762	+ continue;
	3763	+ }
	3764	+
	3765	+ // The built-in aggregates carry their switch in the inclusion flag
	3766	+ // rather than under a post type name, and they are the four most
	3767	+ // people actually use. build_segmented_sitemap_urls() drops a child
	3768	+ // whose flag is empty; regenerating from a list saved while it was
	3769	+ // still on has to make the same decision, or turning Posts, Pages,
	3770	+ // Categories or Tags off in the matrix rewrites and re-lists the
	3771	+ // very file it was asked to remove.
	3772	+ // An absent flag means "not configured", which every other reader
	3773	+ // treats as included; only a flag that is present and off excludes.
	3774	+ $aggregate_flag = array_search($type, self::INCLUSION_CHILD_TYPES, true);
	3775	+ if ($aggregate_flag !== false
	3776	+ && array_key_exists($aggregate_flag, $settings)
	3777	+ && empty($settings[$aggregate_flag])) {
	3778	+ continue;
	3779	+ }
	3780	+
1208	3781	try {
1209		- $sitemap_xml = '';
1210		- $url_count = 0;
	3782	+ // The single "general"/"WordPress" sitemap is one un-paginated file
	3783	+ // (there is no index to reference extra pages); it no longer drops
	3784	+ // overflow URLs.
	3785	+ if ($type === 'general' || $type === 'wordpress') { // phpcs:ignore WordPress.WP.CapitalPDangit.MisspelledInText -- lowercase on purpose: this is the stored type slug.
	3786	+ $xml = $this->generate_sitemap($settings);
	3787	+ $this->write_sitemap_page($sitemap_config['url'], $xml, $type, $this->count_urls_in_xml($xml), $results, $index_children);
	3788	+ continue;
	3789	+ }
1211	3790
1212		- // Generate sitemap based on type
1213		- switch ($sitemap_config['type']) {
1214		- case 'index':
1215		- $sitemap_xml = $this->generate_sitemap_index($sitemap_urls);
1216		- $url_count = count(array_filter($sitemap_urls, fn($s) => $s['enabled'] && $s['type'] !== 'index'));
1217		- break;
1218		- case 'general':
1219		- case 'wordpress':
1220		- $sitemap_xml = $this->generate_sitemap($settings);
1221		- $url_count = $this->count_urls_in_xml($sitemap_xml);
1222		- break;
1223		- case 'posts':
1224		- $sitemap_xml = $this->generate_post_sitemap('post', $settings);
1225		- $url_count = $this->count_urls_in_xml($sitemap_xml);
1226		- break;
1227		- case 'pages':
1228		- $sitemap_xml = $this->generate_post_sitemap('page', $settings);
1229		- $url_count = $this->count_urls_in_xml($sitemap_xml);
1230		- break;
1231		- case 'categories':
1232		- $sitemap_xml = $this->generate_taxonomy_sitemap('category', $settings);
1233		- $url_count = $this->count_urls_in_xml($sitemap_xml);
1234		- break;
1235		- case 'products':
1236		- // Generate product-specific sitemap
1237		- if (post_type_exists('product')) {
1238		- $sitemap_xml = $this->generate_post_sitemap('product', $settings);
1239		- } else {
1240		- // Create empty sitemap if WooCommerce not installed
1241		- $sitemap_xml = '<?xml version="1.0" encoding="UTF-8"?>' . "\n";
	3791	+ $source = $this->stream_type_entries($type, $settings);
1242	3792
1243		- // Add XSL stylesheet only if styling is enabled
1244		- if (!empty($settings['enable_styling'])) {
1245		- $sitemap_xml .= '<?xml-stylesheet type="text/xsl" href="' . home_url('/wp-content/plugins/thinkrank/assets/sitemap.xsl') . '"?>' . "\n";
1246		- }
	3793	+ // Unknown type — fall back to a single general sitemap file.
	3794	+ if ($source === null) {
	3795	+ $xml = $this->generate_sitemap($settings);
	3796	+ $this->write_sitemap_page($sitemap_config['url'], $xml, $type, $this->count_urls_in_xml($xml), $results, $index_children);
	3797	+ continue;
	3798	+ }
1247	3799
1248		- $sitemap_xml .= '<urlset xmlns="http://www.sitemaps.org/schemas/sitemap/0.9">' . "\n";
1249		- $sitemap_xml .= '</urlset>';
1250		- }
1251		- $url_count = $this->count_urls_in_xml($sitemap_xml);
1252		- break;
1253		- case 'product_categories':
1254		- // Generate product category sitemap
1255		- if (taxonomy_exists('product_cat')) {
1256		- $sitemap_xml = $this->generate_taxonomy_sitemap('product_cat', $settings);
1257		- } else {
1258		- // Create empty sitemap if WooCommerce not installed
1259		- $sitemap_xml = '<?xml version="1.0" encoding="UTF-8"?>' . "\n";
	3800	+ // Stream the entries into files of at most $limit URLs, matching
	3801	+ // Rank Math: page 1 keeps the base filename, pages 2+ get a -N
	3802	+ // suffix, and every page is listed in the index. Buffering only one
	3803	+ // page at a time keeps peak memory bounded to $limit entries rather
	3804	+ // than every URL of the type.
	3805	+ $image_ns = $source['image_ns'];
	3806	+ $buffer = [];
	3807	+ $page = 0;
1260	3808
1261		- // Add XSL stylesheet only if styling is enabled
1262		- if (!empty($settings['enable_styling'])) {
1263		- $sitemap_xml .= '<?xml-stylesheet type="text/xsl" href="' . home_url('/wp-content/plugins/thinkrank/assets/sitemap.xsl') . '"?>' . "\n";
1264		- }
	3809	+ foreach ($source['entries'] as $entry) {
	3810	+ $buffer[] = $entry;
	3811	+ if (count($buffer) >= $limit) {
	3812	+ $page++;
	3813	+ $this->write_sitemap_page($this->paginate_url($sitemap_config['url'], $page), $this->wrap_urlset($buffer, $settings, $image_ns), $type, count($buffer), $results, $index_children);
	3814	+ $buffer = [];
	3815	+ }
	3816	+ }
1265	3817
1266		- $sitemap_xml .= '<urlset xmlns="http://www.sitemaps.org/schemas/sitemap/0.9">' . "\n";
1267		- $sitemap_xml .= '</urlset>';
1268		- }
1269		- $url_count = $this->count_urls_in_xml($sitemap_xml);
1270		- break;
1271		- default:
1272		- // Check if it's a custom post type
1273		- $post_types = get_post_types(['public' => true, '_builtin' => false], 'names');
1274		- if (in_array($sitemap_config['type'], $post_types)) {
1275		- $sitemap_xml = $this->generate_post_sitemap($sitemap_config['type'], $settings);
1276		- $url_count = $this->count_urls_in_xml($sitemap_xml);
1277		- } else {
1278		- // Fallback to general sitemap
1279		- $sitemap_xml = $this->generate_sitemap($settings);
1280		- $url_count = $this->count_urls_in_xml($sitemap_xml);
1281		- }
1282		- break;
	3818	+ // Flush the trailing partial page, or a single empty page when the
	3819	+ // type had no entries at all (parity with the previous behavior of
	3820	+ // always writing at least one page per configured child).
	3821	+ if (!empty($buffer) || $page === 0) {
	3822	+ $page++;
	3823	+ $this->write_sitemap_page($this->paginate_url($sitemap_config['url'], $page), $this->wrap_urlset($buffer, $settings, $image_ns), $type, count($buffer), $results, $index_children);
1283	3824	}
1284	3825
1285		- // Extract filename from URL
1286		- $filename = basename(wp_parse_url($sitemap_config['url'], PHP_URL_PATH));
	3826	+ // Remove pages left over from a previous, larger generation.
	3827	+ $this->cleanup_stale_pages($sitemap_config['url'], $page, $settings);
1287	3828
1288		- // Save sitemap to file
1289		- if ($this->save_sitemap_to_file($sitemap_xml, $filename)) {
	3829	+ } catch (\Exception $e) {
	3830	+ $results['errors'][] = "Error generating {$type} sitemap: " . $e->getMessage();
	3831	+ $results['success'] = false;
	3832	+ }
	3833	+ }
	3834	+
	3835	+ // Local business sitemap (parity with Rank Math's local-sitemap.xml).
	3836	+ // The physical file is written whenever a business identity is
	3837	+ // configured — independent of segmented vs single mode — and is also
	3838	+ // listed in the sitemap index when one exists.
	3839	+ try {
	3840	+ if ($this->regenerate_local_sitemap($settings) && $index_config !== null) {
	3841	+ $index_children[] = ['url' => '/local-sitemap.xml'];
	3842	+ }
	3843	+ } catch (\Exception $e) {
	3844	+ $results['errors'][] = 'Error generating local sitemap: ' . $e->getMessage();
	3845	+ $results['success'] = false;
	3846	+ }
	3847	+
	3848	+ // Build the index last, from the child files actually generated, plus the
	3849	+ // sitemaps other plugins own: those serve their own URLs and write no
	3850	+ // file here, so they are appended to the index only (#104).
	3851	+ if ($index_config !== null) {
	3852	+ foreach (self::additional_sitemaps() as $extra) {
	3853	+ $index_children[] = ['url' => $extra];
	3854	+ }
	3855	+
	3856	+ try {
	3857	+ $index_xml = $this->generate_sitemap_index($index_children, $settings);
	3858	+ $filename = basename(wp_parse_url($index_config['url'], PHP_URL_PATH));
	3859	+
	3860	+ if ($this->save_sitemap_to_file($index_xml, $filename)) {
1290	3861	$results['sitemaps_generated'][] = [
1291		- 'url' => $sitemap_config['url'],
1292		- 'type' => $sitemap_config['type'],
1293		- 'filename' => $filename,
1294		- 'url_count' => $url_count
	3862	+ 'url' => $index_config['url'],
	3863	+ 'type' => 'index',
	3864	+ 'filename' => $filename,
	3865	+ 'url_count' => count($index_children),
1295	3866	];
1296		- $results['total_urls'] += $url_count;
1297	3867	} else {
1298		- $error_message = "Failed to save sitemap: {$filename}";
1299		- $results['errors'][] = $error_message;
	3868	+ $results['errors'][] = "Failed to save sitemap: {$filename}";
1300	3869	$results['success'] = false;
1301		-
1302		- // File save error details available in results array
1303	3870	}
1304		-
1305	3871	} catch (\Exception $e) {
1306		- $results['errors'][] = "Error generating {$sitemap_config['type']} sitemap: " . $e->getMessage();
	3872	+ $results['errors'][] = 'Error generating index sitemap: ' . $e->getMessage();
1307	3873	$results['success'] = false;
1308	3874	}
1309	3875	}
1310	3876
	3877	+ // Remove segments that are no longer part of the set. Publishing was
	3878	+ // purely additive: a type that dropped out (Categories unticked, a CPT
	3879	+ // that stopped qualifying) simply stopped being overwritten, so its file
	3880	+ // kept serving and — until the index happened to be rebuilt — kept being
	3881	+ // listed in it. The index above is built from the children actually
	3882	+ // generated, so pruning here leaves disk and index agreeing.
	3883	+ $this->prune_orphaned_segments($settings, $results['sitemaps_generated']);
	3884	+
1311	3885	return $results;
1312	3886	}
1313	3887
1314	3888	/**
1315		- * Generate sitemap index XML
	3889	+ * Delete published segment files that this run did not write.
1316	3890	*
1317		- * @since 1.0.0
1318		- * @param array $sitemap_urls Array of sitemap configurations
1319		- * @return string Sitemap index XML
	3891	+ * Only filenames this site could have published under its own url pattern
	3892	+ * are considered, so another plugin's or core's sitemap in the web root is
	3893	+ * never a candidate — the same reason cleanup does not glob 'sitemap-*.xml'.
	3894	+ *
	3895	+ * @since 1.31.0
	3896	+ *
	3897	+ * @param array $settings Sitemap settings (read for `custom_url_pattern`).
	3898	+ * @param array $generated Entries from $results['sitemaps_generated'].
	3899	+ * @return string[] Basenames removed.
1320	3900	*/
1321		- private function generate_sitemap_index(array $sitemap_urls): string {
1322		- $settings = $this->get_settings('site');
1323		- $xml = '<?xml version="1.0" encoding="UTF-8"?>' . "\n";
	3901	+ private function prune_orphaned_segments(array $settings, array $generated): array {
	3902	+ // Rendering for a request, not publishing: there is nothing on disk
	3903	+ // this run owns, and a dynamic render must never delete the files a
	3904	+ // site's previous static mode left behind.
	3905	+ if ($this->is_collecting()) {
	3906	+ return [];
	3907	+ }
1324	3908
1325		- // Add XSL stylesheet only if styling is enabled
1326		- if (!empty($settings['enable_styling'])) {
1327		- $xml .= '<?xml-stylesheet type="text/xsl" href="' . home_url('/wp-content/plugins/thinkrank/assets/sitemap-index.xsl') . '"?>' . "\n";
	3909	+ $kept = [];
	3910	+ foreach ($generated as $entry) {
	3911	+ if (!empty($entry['filename'])) {
	3912	+ $kept[strtolower((string) $entry['filename'])] = true;
	3913	+ }
1328	3914	}
1329		- $xml .= '<sitemapindex xmlns="http://www.sitemaps.org/schemas/sitemap/0.9">' . "\n";
1330	3915
1331		- $site_url = home_url();
	3916	+ // The current mode's primary and the local business sitemap are written
	3917	+ // by their own paths and are never orphans here.
	3918	+ $primary = strtolower(basename($this->get_primary_sitemap_filename($settings)));
	3919	+ $kept[$primary] = true;
	3920	+ $kept['local-sitemap.xml'] = true;
1332	3921
1333		- foreach ($sitemap_urls as $sitemap_config) {
1334		- if (!$sitemap_config['enabled'] || $sitemap_config['type'] === 'index') {
	3922	+ // The OTHER mode's primary is an orphan the moment the mode changes:
	3923	+ // index mode leaves sitemap.xml behind, flat mode leaves
	3924	+ // sitemap_index.xml and its children. Both used to be kept
	3925	+ // unconditionally, so the site served two sitemap trees and only ever
	3926	+ // refreshed one (#563). The children are already covered by the segment
	3927	+ // sweep below, which now sees them because the index is no longer kept.
	3928	+ $stale_primaries = array_diff(['sitemap.xml', 'sitemap_index.xml'], [$primary]);
	3929	+
	3930	+ $removed = [];
	3931	+
	3932	+ global $wp_filesystem;
	3933	+ if (!$wp_filesystem) {
	3934	+ require_once ABSPATH . 'wp-admin/includes/file.php';
	3935	+ WP_Filesystem();
	3936	+ }
	3937	+ if (!$wp_filesystem) {
	3938	+ return $removed;
	3939	+ }
	3940	+
	3941	+ $candidates = array_merge($this->publishable_segment_filenames($settings), $stale_primaries);
	3942	+
	3943	+ foreach ($candidates as $candidate) {
	3944	+ if (isset($kept[strtolower($candidate)])) {
1335	3945	continue;
1336	3946	}
1337	3947
1338		- $sitemap_url = $sitemap_config['url'];
1339		- if (!str_starts_with($sitemap_url, 'http')) {
1340		- $sitemap_url = $site_url . $sitemap_url;
	3948	+ if (!preg_match('/^(.*)\.xml$/i', $candidate, $m)) {
	3949	+ continue;
1341	3950	}
1342	3951
1343		- $xml .= " <sitemap>\n";
1344		- $xml .= " <loc>" . esc_url($sitemap_url) . "</loc>\n";
1345		- $xml .= " <lastmod>" . gmdate('c') . "</lastmod>\n";
1346		- $xml .= " </sitemap>\n";
	3952	+ // The base file plus its numeric pagination pages.
	3953	+ $paths = [ABSPATH . $candidate];
	3954	+ foreach (glob(ABSPATH . $m[1] . '-*.xml') ?: [] as $paged) {
	3955	+ if (preg_match('/^' . preg_quote($m[1], '/') . '-\d+\.xml$/i', basename($paged))) {
	3956	+ $paths[] = $paged;
	3957	+ }
	3958	+ }
	3959	+
	3960	+ foreach ($paths as $path) {
	3961	+ if (!file_exists($path)) {
	3962	+ continue;
	3963	+ }
	3964	+ // A name we could have published is not proof we published
	3965	+ // this file: RankMath and core write at the same paths (#515).
	3966	+ if (!$this->webroot_sitemap_is_ours($path, $settings)) {
	3967	+ continue;
	3968	+ }
	3969	+ if ($wp_filesystem->delete($path)) {
	3970	+ $removed[] = basename($path);
	3971	+ }
	3972	+ }
1347	3973	}
1348	3974
1349		- $xml .= '</sitemapindex>';
	3975	+ return $removed;
	3976	+ }
1350	3977
1351		- return $xml;
	3978	+
	3979	+ /**
	3980	+ * Collect the full (un-paginated) entry list for a content sitemap type.
	3981	+ *
	3982	+ * @since 1.14.0
	3983	+ *
	3984	+ * @param string $type Sitemap config type (posts, pages, categories, …)
	3985	+ * @param array $settings Sitemap settings
	3986	+ * @return array{entries: array<string>, image_ns: bool}|null Null for an
	3987	+ * unknown type (caller falls back to a general sitemap).
	3988	+ */
	3989	+ /**
	3990	+ * Write (or remove) the physical local-sitemap.xml file.
	3991	+ *
	3992	+ * Called from every sitemap regeneration path — segmented generation, the
	3993	+ * single-sitemap auto-regeneration, and the manual generate endpoint — so
	3994	+ * the local sitemap works regardless of whether the site uses a sitemap
	3995	+ * index. When no business identity is configured, any stale file is removed.
	3996	+ *
	3997	+ * @since 1.15.x
	3998	+ * @param array|null $settings Sitemap settings (falls back to saved site settings).
	3999	+ * @return bool True when the file was written, false when nothing was written.
	4000	+ */
	4001	+ public function regenerate_local_sitemap(?array $settings = null): bool {
	4002	+ $settings = $settings ?? $this->get_settings('site');
	4003	+ $entries = $this->collect_local_entries();
	4004	+
	4005	+ if (empty($entries)) {
	4006	+ // Business identity was cleared — drop the file we left from
	4007	+ // before, but only ours. `local-sitemap.xml` is the name Rank Math
	4008	+ // publishes under too (this method mirrors it deliberately), so on
	4009	+ // a migrated site the file at that path may never have been ours
	4010	+ // to delete (#515).
	4011	+ $path = ABSPATH . 'local-sitemap.xml';
	4012	+ if (!$this->is_collecting() && file_exists($path) && $this->webroot_sitemap_is_ours($path, $settings)) {
	4013	+ wp_delete_file($path);
	4014	+ }
	4015	+ return false;
	4016	+ }
	4017	+
	4018	+ $xml = $this->wrap_urlset($entries, $settings, false);
	4019	+ return $this->save_sitemap_to_file($xml, 'local-sitemap.xml');
1352	4020	}
1353	4021
1354	4022	/**
1355		- * Generate sitemap for specific post type
	4023	+ * Build the local business sitemap entries.
1356	4024	*
1357		- * @since 1.0.0
1358		- * @param string $post_type Post type to generate sitemap for
1359		- * @param array $settings Sitemap settings
1360		- * @return string Sitemap XML
	4025	+ * Returns a single URL entry pointing at the on-site page that carries the
	4026	+ * business's LocalBusiness/Organization schema (the configured business URL
	4027	+ * when it's on this site, otherwise the homepage). Gated — like Rank Math's
	4028	+ * `local-sitemap.xml` — on a business (non-person) type being configured
	4029	+ * with an actual location (address or geo coordinates). Returns an empty
	4030	+ * array when no business location is set, so no empty local sitemap is
	4031	+ * written or added to the index.
	4032	+ *
	4033	+ * @since 1.15.x
	4034	+ * @return array Zero or one URL entry
1361	4035	*/
1362		- private function generate_post_sitemap(string $post_type, array $settings): string {
1363		- $xml = '<?xml version="1.0" encoding="UTF-8"?>' . "\n";
	4036	+ /**
	4037	+ * Does this site publish a local business sitemap right now?
	4038	+ *
	4039	+ * The same gate {@see self::regenerate_local_sitemap()} applies, asked
	4040	+ * without writing anything. Callers that need to know whether the document
	4041	+ * exists must not test the filesystem: under dynamic delivery it is served
	4042	+ * from PHP and there is no file, which is how `local-sitemap.xml` came to be
	4043	+ * dropped from robots.txt on exactly those sites (#752).
	4044	+ *
	4045	+ * @since 2.9.0
	4046	+ *
	4047	+ * @return bool True when the local sitemap has content to publish.
	4048	+ */
	4049	+ public function publishes_local_sitemap(): bool {
	4050	+ return !empty($this->collect_local_entries());
	4051	+ }
1364	4052
1365		- // Add XSL stylesheet only if styling is enabled
1366		- if (!empty($settings['enable_styling'])) {
1367		- $xml .= '<?xml-stylesheet type="text/xsl" href="' . home_url('/wp-content/plugins/thinkrank/assets/sitemap.xsl') . '"?>' . "\n";
	4053	+ private function collect_local_entries(): array {
	4054	+ if (!class_exists('ThinkRank\\SEO\\Site_Identity_Manager')) {
	4055	+ require_once THINKRANK_PLUGIN_DIR . 'includes/seo/class-site-identity-manager.php';
1368	4056	}
1369	4057
1370		- // Add image namespace if images are enabled
1371		- if (!empty($settings['include_images'])) {
1372		- $xml .= '<urlset xmlns="http://www.sitemaps.org/schemas/sitemap/0.9" xmlns:image="http://www.google.com/schemas/sitemap-image/1.1">' . "\n";
	4058	+ $identity = (new \ThinkRank\SEO\Site_Identity_Manager())->get_settings('site');
	4059	+
	4060	+ // Gate like Rank Math's local sitemap: a business (non-person) identity
	4061	+ // is configured. We only emit a single URL (the location page), so a
	4062	+ // full postal address is NOT required — requiring one wrongly excluded
	4063	+ // migrated sites, since the Rank Math importer carries over business
	4064	+ // type/name/phone but not the address. Personal sites, and sites with no
	4065	+ // business identity at all, get no local sitemap.
	4066	+ $business_type = strtolower((string) ($identity['business_type'] ?? ''));
	4067	+ if ($business_type === 'person') {
	4068	+ return [];
	4069	+ }
	4070	+ if ($business_type === '' && empty($identity['business_name'])) {
	4071	+ return [];
	4072	+ }
	4073	+
	4074	+ // Prefer the configured business URL when it points at this site;
	4075	+ // otherwise fall back to the homepage (which outputs the schema).
	4076	+ $location_url = home_url('/');
	4077	+ if (!empty($identity['business_url'])) {
	4078	+ $candidate = esc_url_raw((string) $identity['business_url']);
	4079	+ if ($candidate && wp_parse_url($candidate, PHP_URL_HOST) === wp_parse_url(home_url(), PHP_URL_HOST)) {
	4080	+ $location_url = $candidate;
	4081	+ }
	4082	+ }
	4083	+
	4084	+ // Omit lastmod — no real modification source for the local business URL.
	4085	+ return [$this->generate_url_entry($location_url, '', 0.8, 'weekly')];
	4086	+ }
	4087	+
	4088	+ /**
	4089	+ * Resolve a streaming entry source for a sitemap type.
	4090	+ *
	4091	+ * Returns an iterable that yields the type's <url> entries one at a time
	4092	+ * (so the paginator never materializes the whole type), the image-namespace
	4093	+ * flag, or null when the type isn't one we build (caller falls back to a
	4094	+ * single general sitemap). Entry order matches the previous array-based
	4095	+ * collect_type_entries() exactly.
	4096	+ *
	4097	+ * @param string $type Sitemap type.
	4098	+ * @param array $settings Sitemap settings.
	4099	+ * @return array{entries: iterable<string>, image_ns: bool}|null
	4100	+ */
	4101	+ private function stream_type_entries(string $type, array $settings): ?array {
	4102	+ switch ($type) {
	4103	+ case 'posts':
	4104	+ return ['entries' => $this->collect_post_entries_iter(['post'], $settings), 'image_ns' => true];
	4105	+
	4106	+ case 'pages':
	4107	+ // The homepage lives in the pages sitemap (parity with prior
	4108	+ // output) and is emitted first; collect_post_entries_iter()
	4109	+ // already excludes the static front page, so there's no duplicate.
	4110	+ $entries = (function () use ($settings) {
	4111	+ yield $this->generate_url_entry(home_url('/'), $this->get_homepage_lastmod(), 1.0, 'daily');
	4112	+ yield from $this->collect_post_entries_iter(['page'], $settings);
	4113	+ })();
	4114	+ return ['entries' => $entries, 'image_ns' => true];
	4115	+
	4116	+ case 'categories':
	4117	+ return ['entries' => $this->collect_taxonomy_entries_iter('category', $settings), 'image_ns' => false];
	4118	+
	4119	+ case 'tags':
	4120	+ return ['entries' => $this->collect_taxonomy_entries_iter('post_tag', $settings), 'image_ns' => false];
	4121	+
	4122	+ case 'products':
	4123	+ $entries = post_type_exists('product') ? $this->collect_post_entries_iter(['product'], $settings) : [];
	4124	+ return ['entries' => $entries, 'image_ns' => true];
	4125	+
	4126	+ default:
	4127	+ // Resolve a preset's display name to the object it streams, so
	4128	+ // 'product_categories' is an ordinary taxonomy child rather than
	4129	+ // a case of its own (#690).
	4130	+ $alias = self::CHILD_TYPE_ALIASES[$type] ?? null;
	4131	+ $object = $alias ?? $type;
	4132	+
	4133	+ $custom_post_types = get_post_types(['public' => true, '_builtin' => false], 'names');
	4134	+ if (in_array($object, $custom_post_types, true)) {
	4135	+ return ['entries' => $this->collect_post_entries_iter([$object], $settings), 'image_ns' => true];
	4136	+ }
	4137	+
	4138	+ // Custom taxonomies reach index mode here, streamed through the
	4139	+ // same iterator flat mode uses so the two modes emit identical
	4140	+ // URLs for the same settings.
	4141	+ if (taxonomy_exists($object) && $this->should_include_taxonomy($object)) {
	4142	+ return ['entries' => $this->collect_taxonomy_entries_iter($object, $settings), 'image_ns' => false];
	4143	+ }
	4144	+
	4145	+ // An aliased child whose object is gone (WooCommerce deactivated)
	4146	+ // keeps writing the empty file it always wrote. Returning null
	4147	+ // here would hand it the whole-site fallback instead, dumping
	4148	+ // every URL on the site into a file named for products.
	4149	+ //
	4150	+ // A registered taxonomy this generator will not emit
	4151	+ // (`post_format`, `nav_menu`, a non-public one) needs the same
	4152	+ // answer for the same reason. generate_multiple_sitemaps()
	4153	+ // skips those before they reach here, so nothing takes this
	4154	+ // path today — but it is the one branch where falling through
	4155	+ // to null is silently catastrophic rather than merely wrong,
	4156	+ // and the guard keeping it unreachable lives in another method.
	4157	+ if ($alias !== null || taxonomy_exists($object)) {
	4158	+ return ['entries' => [], 'image_ns' => false];
	4159	+ }
	4160	+
	4161	+ return null;
	4162	+ }
	4163	+ }
	4164	+
	4165	+ /**
	4166	+ * Write one sitemap page file and record it in the results + index child list.
	4167	+ *
	4168	+ * @since 1.14.0
	4169	+ *
	4170	+ * @param string $url Sitemap page URL
	4171	+ * @param string $xml Sitemap XML
	4172	+ * @param string $type Sitemap type (for reporting)
	4173	+ * @param int $url_count Number of URLs in this page
	4174	+ * @param array $results Results accumulator (by reference)
	4175	+ * @param array $index_children Index child list (by reference)
	4176	+ * @return void
	4177	+ */
	4178	+ private function write_sitemap_page(string $url, string $xml, string $type, int $url_count, array &$results, array &$index_children): void {
	4179	+ $filename = basename(wp_parse_url($url, PHP_URL_PATH));
	4180	+
	4181	+ if ($this->save_sitemap_to_file($xml, $filename)) {
	4182	+ $results['sitemaps_generated'][] = [
	4183	+ 'url' => $url,
	4184	+ 'type' => $type,
	4185	+ 'filename' => $filename,
	4186	+ 'url_count' => $url_count,
	4187	+ ];
	4188	+ $results['total_urls'] += $url_count;
	4189	+ $index_children[] = ['url' => $url];
1373	4190	} else {
1374		- $xml .= '<urlset xmlns="http://www.sitemaps.org/schemas/sitemap/0.9">' . "\n";
	4191	+ $results['errors'][] = "Failed to save sitemap: {$filename}";
	4192	+ $results['success'] = false;
1375	4193	}
	4194	+ }
1376	4195
1377		- // Add homepage only for general sitemap
1378		- if ($post_type === 'page') {
1379		- $xml .= $this->generate_url_entry(home_url('/'), gmdate('c'), 1.0, 'daily');
	4196	+ /**
	4197	+ * Delete pagination files left over when a type shrinks to fewer pages.
	4198	+ *
	4199	+ * For a base URL like /sitemap-posts.xml, removes sitemap-posts-N.xml files
	4200	+ * whose page number N exceeds the current page count. Page 1 (the base file,
	4201	+ * which has no -N suffix) is never touched.
	4202	+ *
	4203	+ * @since 1.14.0
	4204	+ *
	4205	+ * @since 2.1.1 Each candidate must pass the content ownership test — a
	4206	+ * `-N.xml` page of another plugin's sitemap paginates our
	4207	+ * stem exactly as ours does (#515).
	4208	+ *
	4209	+ * @param string $base_url Base (page 1) sitemap URL
	4210	+ * @param int $current_pages Number of pages generated this run
	4211	+ * @param array $settings Sitemap settings, for the ownership test.
	4212	+ * @return void
	4213	+ */
	4214	+ private function cleanup_stale_pages(string $base_url, int $current_pages, array $settings): void {
	4215	+ // See prune_orphaned_segments(): a dynamic render deletes nothing.
	4216	+ if ($this->is_collecting()) {
	4217	+ return;
1380	4218	}
1381	4219
1382		- // Add posts/pages of specific type
1383		- $xml .= $this->generate_post_entries($post_type, $settings);
	4220	+ $filename = basename(wp_parse_url($base_url, PHP_URL_PATH));
	4221	+ if (!preg_match('/^(.*)\.xml$/i', $filename, $m)) {
	4222	+ return;
	4223	+ }
	4224	+ $stem = $m[1];
1384	4225
1385		- $xml .= '</urlset>';
1386		- return $xml;
	4226	+ global $wp_filesystem;
	4227	+ if (!$wp_filesystem) {
	4228	+ require_once ABSPATH . 'wp-admin/includes/file.php';
	4229	+ WP_Filesystem();
	4230	+ }
	4231	+ if (!$wp_filesystem) {
	4232	+ return;
	4233	+ }
	4234	+
	4235	+ $candidates = glob(ABSPATH . $stem . '-*.xml') ?: [];
	4236	+ foreach ($candidates as $path) {
	4237	+ // Only delete numeric-suffixed pages beyond the current count.
	4238	+ if (preg_match('/-(\d+)\.xml$/', basename($path), $mm)
	4239	+ && (int) $mm[1] > $current_pages
	4240	+ && $this->webroot_sitemap_is_ours($path, $settings)) {
	4241	+ $wp_filesystem->delete($path);
	4242	+ }
	4243	+ }
1387	4244	}
1388	4245
1389	4246	/**
1390		- * Generate sitemap for specific taxonomy
	4247	+ * Sitemap URLs contributed by other plugins.
1391	4248	*
1392		- * @since 1.0.0
1393		- * @param string $taxonomy Taxonomy to generate sitemap for
	4249	+ * ThinkRank owns the sitemap index and the robots.txt `Sitemap:` lines, so a
	4250	+ * companion plugin that serves its own sitemap — Pro's news and video
	4251	+ * sitemaps, for instance — had no way to be discovered: it appeared in
	4252	+ * neither, leaving manual Search Console submission as the only route in
	4253	+ * (#104). Registering here puts a sitemap in the index when one exists, and
	4254	+ * in robots.txt when it does not.
	4255	+ *
	4256	+ * Callers get root-relative paths. Entries are normalised to a leading
	4257	+ * slash, de-duplicated, and anything that is not a non-empty string is
	4258	+ * dropped, so one badly-behaved callback cannot produce a malformed index.
	4259	+ *
	4260	+ * @since 2.3.1
	4261	+ *
	4262	+ * @return string[] Root-relative sitemap paths, e.g. ['/news-sitemap.xml'].
	4263	+ */
	4264	+ public static function additional_sitemaps(): array {
	4265	+ /**
	4266	+ * Filters the sitemaps contributed by other plugins.
	4267	+ *
	4268	+ * @since 2.3.1
	4269	+ *
	4270	+ * @param string[] $sitemaps Root-relative sitemap paths.
	4271	+ */
	4272	+ $sitemaps = apply_filters('thinkrank_additional_sitemaps', []);
	4273	+
	4274	+ if (!is_array($sitemaps)) {
	4275	+ return [];
	4276	+ }
	4277	+
	4278	+ // Both consumers resolve an entry with home_url(), which prefixes the
	4279	+ // install's own directory. Everything below is measured against that so
	4280	+ // an absolute URL is reduced to what home_url() will put back.
	4281	+ $home = wp_parse_url(home_url('/'));
	4282	+ $home_host = strtolower((string) ($home['host'] ?? ''));
	4283	+ $home_path = '/' . trim((string) ($home['path'] ?? ''), '/');
	4284	+
	4285	+ $clean = [];
	4286	+ foreach ($sitemaps as $sitemap) {
	4287	+ if (!is_string($sitemap)) {
	4288	+ continue;
	4289	+ }
	4290	+
	4291	+ $sitemap = trim($sitemap);
	4292	+ if ('' === $sitemap) {
	4293	+ continue;
	4294	+ }
	4295	+
	4296	+ // A full URL on this site is accepted and reduced to the part
	4297	+ // home_url() does not already supply, so a caller that reached for
	4298	+ // home_url() still lands in the right place — including on a
	4299	+ // subdirectory install, where keeping the whole path would repeat
	4300	+ // the directory. A URL on another host is dropped rather than
	4301	+ // rewritten: the sitemaps protocol will not accept a cross-host
	4302	+ // child anyway, and reusing its path would advertise a URL on this
	4303	+ // site that does not exist.
	4304	+ if (preg_match('#^(https?:)?//#i', $sitemap)) {
	4305	+ $parts = wp_parse_url('//' === substr($sitemap, 0, 2) ? 'https:' . $sitemap : $sitemap);
	4306	+ if (!is_array($parts)) {
	4307	+ continue;
	4308	+ }
	4309	+
	4310	+ if (strtolower((string) ($parts['host'] ?? '')) !== $home_host) {
	4311	+ continue;
	4312	+ }
	4313	+
	4314	+ $path = (string) ($parts['path'] ?? '');
	4315	+ if ('' === $path) {
	4316	+ continue;
	4317	+ }
	4318	+
	4319	+ if ('/' !== $home_path && ($path === $home_path || 0 === strpos($path, $home_path . '/'))) {
	4320	+ $path = substr($path, strlen($home_path));
	4321	+ }
	4322	+
	4323	+ // A sitemap served from a query string keeps it; dropping the
	4324	+ // query would point at a different document.
	4325	+ $query = (string) ($parts['query'] ?? '');
	4326	+ $sitemap = $path . ('' !== $query ? '?' . $query : '');
	4327	+ }
	4328	+
	4329	+ $clean[] = '/' . ltrim($sitemap, '/');
	4330	+ }
	4331	+
	4332	+ return array_values(array_unique($clean));
	4333	+ }
	4334	+
	4335	+ /**
	4336	+ * Generate sitemap index XML from the list of child sitemap files produced
	4337	+ * during generation (each already resolved to its final, possibly paginated,
	4338	+ * URL).
	4339	+ *
	4340	+ * @since 1.0.0 (signature updated 1.14.0)
	4341	+ * @param array $children Array of ['url' => string] child sitemap entries
1394	4342	* @param array $settings Sitemap settings
1395		- * @return string Sitemap XML
	4343	+ * @return string Sitemap index XML
1396	4344	*/
1397		- private function generate_taxonomy_sitemap(string $taxonomy, array $settings): string {
1398		- $xml = '<?xml version="1.0" encoding="UTF-8"?>' . "\n";
	4345	+ private function generate_sitemap_index(array $children, array $settings): string {
	4346	+ $xml = $this->xml_prolog($settings, 'index');
	4347	+ $xml .= '<sitemapindex xmlns="http://www.sitemaps.org/schemas/sitemap/0.9">' . "\n";
1399	4348
1400		- // Add XSL stylesheet only if styling is enabled
1401		- if (!empty($settings['enable_styling'])) {
1402		- $xml .= '<?xml-stylesheet type="text/xsl" href="' . home_url('/wp-content/plugins/thinkrank/assets/sitemap.xsl') . '"?>' . "\n";
	4349	+ $site_url = home_url();
	4350	+
	4351	+ foreach ($children as $child) {
	4352	+ $sitemap_url = $child['url'] ?? '';
	4353	+ if ($sitemap_url === '') {
	4354	+ continue;
	4355	+ }
	4356	+ if (!str_starts_with($sitemap_url, 'http')) {
	4357	+ $sitemap_url = $site_url . $sitemap_url;
	4358	+ }
	4359	+
	4360	+ $xml .= " <sitemap>\n";
	4361	+ $xml .= " <loc>" . esc_url(Url_Scheme::apply($sitemap_url)) . "</loc>\n";
	4362	+ $xml .= " <lastmod>" . gmdate('c') . "</lastmod>\n";
	4363	+ $xml .= " </sitemap>\n";
1403	4364	}
1404		- $xml .= '<urlset xmlns="http://www.sitemaps.org/schemas/sitemap/0.9">' . "\n";
1405	4365
1406		- // Add taxonomy terms
1407		- $xml .= $this->generate_taxonomy_entries($taxonomy, $settings);
	4366	+ $xml .= '</sitemapindex>';
1408	4367
1409		- $xml .= '</urlset>';
1410	4368	return $xml;
1411	4369	}
1412	4370
1413	4371	/**
		@@ -1418,35 +4376,8 @@
1418	4376	* @return int Number of URLs
1419	4377	*/
1420	4378	private function count_urls_in_xml(string $sitemap_xml): int {
1421	4379	return substr_count($sitemap_xml, '<url>');
1422		- }
1423		-
1424		- /**
1425		- * Ping search engines about sitemap update
1426		- *
1427		- * @since 1.0.0
1428		- * @return void
1429		- */
1430		- private function ping_search_engines(): void {
1431		- $sitemap_url = home_url('/sitemap.xml');
1432		-
1433		- $ping_urls = [
1434		- 'google' => 'https://www.google.com/ping?sitemap=' . urlencode($sitemap_url),
1435		- 'bing' => 'https://www.bing.com/ping?sitemap=' . urlencode($sitemap_url)
1436		- ];
1437		-
1438		- foreach ($ping_urls as $engine => $ping_url) {
1439		- // Use wp_remote_get with timeout to prevent hanging
1440		- $response = wp_remote_get($ping_url, [
1441		- 'timeout' => 10,
1442		- 'blocking' => false // Non-blocking to prevent delays
1443		- ]);
1444		-
1445		- if (is_wp_error($response)) {
1446		- // Search engine ping failed - error details available in WP_Error response
1447		- }
1448		- }
1449	4380	}
1450	4381
1451	4382	/**
1452	4383	* Validate and sanitize exclude posts input