← All changes | includes/seo/class-pattern-resolver.php +242 -17 2.4.0 → 2.14.3 View file →

		@@ -165,9 +165,14 @@
165	165	$text = excerpt_remove_blocks($content);
166	166	$text = strip_shortcodes($text);
167	167	$text = wp_strip_all_tags($text);
168	168
169		- return trim(wp_trim_words($text, $words, '...'));
	169	+ // $words is a WORD cap, but wp_trim_words() counts CHARACTERS on
	170	+ // th/ja/zh_*, where it would cut to ~25 characters instead of ~25
	171	+ // words — about six times too short (#687). trim_words() keeps the
	172	+ // word cap where words are the unit and falls back to a character
	173	+ // budget where they are not.
	174	+ return trim(\ThinkRank\Core\Seo_Text::trim_words($text, $words));
170	175	}
171	176
172	177	/**
173	178	* Resolve any variable-tag string against a term's values.
		@@ -215,10 +220,12 @@
215	220	'%title%' => $name,
216	221	'%term%' => $name,
217	222	'%sitename%' => get_bloginfo('name'),
218	223	'%sep%' => self::separator(),
	224	+ // Same locale trap as derive_excerpt(): a word cap here is a
	225	+ // ~25-character cap on th/ja/zh_* (#687).
219	226	'%excerpt%' => $description !== ''
220		- ? wp_trim_words(wp_strip_all_tags($description), 25, '...')
	227	+ ? self::derive_excerpt($description)
221	228	: '',
222	229	'%date%' => '',
223	230	'%modified%' => '',
224	231	'%author%' => '',
		@@ -253,11 +260,13 @@
253	260	public static function description(int $post_id): string {
254	261	$template = self::template_for($post_id, 'description', self::DEFAULT_DESCRIPTION);
255	262	$description = self::resolve_value($template, $post_id);
256	263
257		- if (strlen($description) > 160) {
258		- $description = wp_trim_words($description, 25, '...');
259		- }
	264	+ // Measure and cut in CHARACTERS. strlen() counts bytes, so a Thai or
	265	+ // CJK description tripped this limit at a third of its length, and
	266	+ // wp_trim_words() then cut by a unit the locale chooses — 25 words in
	267	+ // English, 25 characters in Thai (#687).
	268	+ $description = \ThinkRank\Core\Seo_Text::trim_to_length($description);
260	269
261	270	return $description;
262	271	}
263	272
		@@ -360,8 +369,27 @@
360	369	return is_string($template) && $template !== '' ? $template : $fallback;
361	370	}
362	371
363	372	/**
	373	+ * Whether the post's content is behind a password for this request.
	374	+ *
	375	+ * The same test SEO_Manager uses for the Global SEO path (#363), guarded
	376	+ * because this class is also exercised outside a front-end request.
	377	+ *
	378	+ * @since 2.15.0
	379	+ *
	380	+ * @param \WP_Post $post Post.
	381	+ * @return bool
	382	+ */
	383	+ private static function is_password_required(\WP_Post $post): bool {
	384	+ if ('' === (string) $post->post_password) {
	385	+ return false;
	386	+ }
	387	+
	388	+ return !function_exists('post_password_required') || post_password_required($post);
	389	+ }
	390	+
	391	+ /**
364	392	* Build placeholder values for an explicit post (no loop dependency).
365	393	*
366	394	* @param int $post_id Post ID.
367	395	* @return array<string,string> Placeholder map.
		@@ -368,13 +396,21 @@
368	396	*/
369	397	private static function placeholders_for(int $post_id): array {
370	398	$post = get_post($post_id);
371	399
	400	+ // An authored excerpt is written for public consumption, so it stays.
	401	+ // The body is not: deriving %excerpt% from a protected post's body put
	402	+ // the gated text into the title, meta, og/twitter tags and JSON-LD of
	403	+ // the page that asks for the password. The Global SEO path closed this
	404	+ // in #363; the per-post templates resolve here, so the same rule
	405	+ // applies here too (#973).
372	406	$excerpt = '';
373	407	if ($post) {
374		- $excerpt = !empty($post->post_excerpt)
375		- ? $post->post_excerpt
376		- : self::derive_excerpt(Builder_Content::visible_content($post));
	408	+ if (!empty($post->post_excerpt)) {
	409	+ $excerpt = $post->post_excerpt;
	410	+ } elseif (!self::is_password_required($post)) {
	411	+ $excerpt = self::derive_excerpt(Builder_Content::visible_content($post));
	412	+ }
377	413	}
378	414
379	415	$author_id = (int) get_post_field('post_author', $post_id);
380	416
		@@ -383,21 +419,148 @@
383	419	$categories = get_the_category($post_id);
384	420	$category = !empty($categories) ? $categories[0]->name : '';
385	421	}
386	422
	423	+ return array_merge(
	424	+ [
	425	+ '%title%' => get_the_title($post_id),
	426	+ '%sitename%' => get_bloginfo('name'),
	427	+ '%sep%' => self::separator(),
	428	+ '%excerpt%' => $excerpt,
	429	+ '%date%' => get_the_date('', $post_id),
	430	+ '%modified%' => get_the_modified_date('', $post_id),
	431	+ '%author%' => $author_id ? get_the_author_meta('display_name', $author_id) : '',
	432	+ '%category%' => $category,
	433	+ ],
	434	+ self::product_placeholders($post_id)
	435	+ );
	436	+ }
	437	+
	438	+ /**
	439	+ * Tokens that only mean anything on a WooCommerce product.
	440	+ *
	441	+ * Rank Math and Yoast WooCommerce SEO both let a product title or
	442	+ * description carry the price, the SKU and the stock status, and both of
	443	+ * our converters dropped every token they did not recognise — so
	444	+ * "Buy %title% for %wc_price%" imported as "Buy %title% for" and the
	445	+ * customer in support #171748 found four variables where they had had a
	446	+ * stock one (#715).
	447	+ *
	448	+ * Always present, never conditional on the post type. A token that exists
	449	+ * on a product and is an unknown token everywhere else would resolve on
	450	+ * one page and leak literally on another; resolving to an empty string off
	451	+ * a product is the same answer every other token gives when it has nothing
	452	+ * to say, and process() then collapses the separator it leaves behind.
	453	+ *
	454	+ * @since 2.10.1
	455	+ *
	456	+ * @param int $post_id Post ID.
	457	+ * @return array<string,string>
	458	+ */
	459	+ private static function product_placeholders(int $post_id): array {
	460	+ $empty = [
	461	+ '%price%' => '',
	462	+ '%sale_price%' => '',
	463	+ '%sku%' => '',
	464	+ '%stock_status%' => '',
	465	+ '%short_description%' => '',
	466	+ '%brand%' => '',
	467	+ ];
	468	+
	469	+ if (!function_exists('wc_get_product') || 'product' !== get_post_type($post_id)) {
	470	+ return $empty;
	471	+ }
	472	+
	473	+ $product = wc_get_product($post_id);
	474	+ if (!$product) {
	475	+ return $empty;
	476	+ }
	477	+
	478	+ // Read through wc_price()/get_price_html() rather than formatting the
	479	+ // number here: currency symbol, position, decimals and the "from X"
	480	+ // form on a variable product are all store settings, and a second
	481	+ // formatter is how a template starts disagreeing with the price shown
	482	+ // three lines below it on the same page.
	483	+ $price = (string) $product->get_price();
	484	+ $sale_price = (string) $product->get_sale_price();
	485	+
	486	+ $stock_status = (string) $product->get_stock_status();
	487	+ $stock_labels = [
	488	+ 'instock' => __('In stock', 'thinkrank'),
	489	+ 'outofstock' => __('Out of stock', 'thinkrank'),
	490	+ 'onbackorder' => __('On backorder', 'thinkrank'),
	491	+ ];
	492	+
387	493	return [
388		- '%title%' => get_the_title($post_id),
389		- '%sitename%' => get_bloginfo('name'),
390		- '%sep%' => self::separator(),
391		- '%excerpt%' => $excerpt,
392		- '%date%' => get_the_date('', $post_id),
393		- '%modified%' => get_the_modified_date('', $post_id),
394		- '%author%' => $author_id ? get_the_author_meta('display_name', $author_id) : '',
395		- '%category%' => $category,
	494	+ '%price%' => self::formatted_price($price),
	495	+ '%sale_price%' => self::formatted_price($sale_price),
	496	+ '%sku%' => (string) $product->get_sku(),
	497	+ '%stock_status%' => $stock_labels[$stock_status] ?? $stock_status,
	498	+ '%short_description%' => self::derive_excerpt((string) $product->get_short_description()),
	499	+ '%brand%' => self::product_brand($post_id),
396	500	];
397	501	}
398	502
399	503	/**
	504	+ * A price as text, in the store's own currency format.
	505	+ *
	506	+ * wc_price() returns markup, and stripping the tags off it leaves the
	507	+ * entities behind: a title read "11.05&#2547;&nbsp;" on the first live
	508	+ * run. Entities are decoded and the non-breaking space collapsed, because
	509	+ * this value ends up inside a `<title>` and a meta description, where
	510	+ * markup has no meaning and an entity is just noise a reader sees.
	511	+ *
	512	+ * @since 2.10.1
	513	+ *
	514	+ * @param string $price Raw price, or '' when the product has none.
	515	+ * @return string
	516	+ */
	517	+ private static function formatted_price(string $price): string {
	518	+ if ('' === $price) {
	519	+ return '';
	520	+ }
	521	+
	522	+ $text = wp_strip_all_tags((string) wc_price((float) $price));
	523	+ $text = html_entity_decode($text, ENT_QUOTES | ENT_HTML5, 'UTF-8');
	524	+
	525	+ // \xC2\xA0 is the non-breaking space wc_price() puts between the
	526	+ // amount and the symbol; a literal one in a title is invisible to a
	527	+ // reader and awkward for everything else.
	528	+ $text = str_replace("\xC2\xA0", ' ', $text);
	529	+
	530	+ return trim((string) preg_replace('/\s+/u', ' ', $text));
	531	+ }
	532	+
	533	+ /**
	534	+ * A product's brand, from whichever taxonomy the store uses for one.
	535	+ *
	536	+ * WooCommerce core added `product_brand` in 9.4; before that every brand
	537	+ * plugin shipped its own taxonomy, and a store that migrated from one of
	538	+ * them still has the old terms. Asking each in turn costs one cached term
	539	+ * lookup and means the token is not empty on the stores most likely to
	540	+ * have used a brand token in the plugin they are leaving.
	541	+ *
	542	+ * @since 2.10.1
	543	+ *
	544	+ * @param int $post_id Product ID.
	545	+ * @return string
	546	+ */
	547	+ private static function product_brand(int $post_id): string {
	548	+ foreach (['product_brand', 'pwb-brand', 'yith_product_brand', 'berocket_brand'] as $taxonomy) {
	549	+ if (!taxonomy_exists($taxonomy)) {
	550	+ continue;
	551	+ }
	552	+
	553	+ $terms = get_the_terms($post_id, $taxonomy);
	554	+ if (is_array($terms) && !empty($terms)) {
	555	+ return (string) $terms[0]->name;
	556	+ }
	557	+ }
	558	+
	559	+ return '';
	560	+ }
	561	+
	562	+ /**
400	563	* Active title separator symbol.
401	564	*
402	565	* @return string Separator.
403	566	*/
		@@ -415,8 +578,17 @@
415	578	* @param array<string,string> $placeholders Placeholder map.
416	579	* @return string Resolved string.
417	580	*/
418	581	private static function process(string $template, array $placeholders): string {
	582	+ // Both forms go BEFORE substitution, and the strip is decided against
	583	+ // the placeholder map rather than by what is left over afterwards.
	584	+ // Running it on the substituted string scanned the resolved VALUES too,
	585	+ // so a post whose own title read "Using %name% placeholders in
	586	+ // %%mustache%% templates" published "Using placeholders in %%
	587	+ // templates" — its title edited, and a stray double percent where the
	588	+ // inner token had been eaten out of the middle of one.
	589	+ $template = self::strip_unresolved_tokens($template, $placeholders);
	590	+
419	591	$value = str_replace(array_keys($placeholders), array_values($placeholders), $template);
420	592
421	593	// Collapse whitespace.
422	594	$value = preg_replace('/\s+/', ' ', $value);
		@@ -431,7 +603,60 @@
431	603	$value
432	604	);
433	605
434	606	// Strip leading/trailing separators and whitespace.
435		- return trim($value, " \t\n\r\0\x0B" . $separator);
	607	+ return \ThinkRank\Core\Seo_Text::trim_separator($value, $separator);
	608	+ }
	609	+
	610	+ /**
	611	+ * Remove any token the placeholder map did not resolve.
	612	+ *
	613	+ * The last line of defence, and the reason it exists is that every layer
	614	+ * above it is a list someone has to remember to extend. A converter that
	615	+ * misses a token, a template typed by hand, a value written straight into
	616	+ * postmeta by an importer we have not met: each one ends with `%%title%%`
	617	+ * or `%some_token%` rendering literally in a `<title>` on a live site, and
	618	+ * that is precisely what was reported on 14 September (#715).
	619	+ *
	620	+ * Both syntaxes, because a migrated site carries both: Yoast's `%%x%%`
	621	+ * (which includes Rank Math tokens Yoast's own importer wrapped in double
	622	+ * percent signs without translating them) and the single-percent form
	623	+ * ThinkRank and Rank Math share. `%%x%%` is matched as a whole so the pass
	624	+ * cannot eat the inner `%x%` and leave a stray percent sign at each end.
	625	+ *
	626	+ * Applied to the TEMPLATE, and a token is kept only when the placeholder
	627	+ * map has it. Stripping whatever still looked like a token after
	628	+ * substitution read the resolved values as well, and content is not a
	629	+ * template: a post titled "Using %name% placeholders in %%mustache%%
	630	+ * templates" published "Using placeholders in %% templates", and an excerpt
	631	+ * of "Learn %name% and %fabric% placeholders." published "Learn and
	632	+ * placeholders." Deciding against the map also means a token this resolver
	633	+ * knows about is never at risk, whatever a value happens to contain.
	634	+ *
	635	+ * A bare percent is left alone: "50% off" is ordinary copy, and a guard
	636	+ * that ate it would be a worse bug than the one it prevents.
	637	+ *
	638	+ * @since 2.10.1
	639	+ *
	640	+ * @param string $template Raw template.
	641	+ * @param array<string, string> $placeholders Tokens this resolver can resolve.
	642	+ * @return string
	643	+ */
	644	+ private static function strip_unresolved_tokens(string $template, array $placeholders): string {
	645	+ if (false === strpos($template, '%')) {
	646	+ return $template;
	647	+ }
	648	+
	649	+ $stripped = preg_replace_callback(
	650	+ '/%%[a-z0-9_-]+%%|%[a-z0-9_-]+%/i',
	651	+ static function (array $found) use ($placeholders): string {
	652	+ return array_key_exists($found[0], $placeholders) ? $found[0] : '';
	653	+ },
	654	+ $template
	655	+ );
	656	+
	657	+ // preg_replace_callback() answers null on content that is not valid
	658	+ // UTF-8 rather than throwing, and returning null here would blank a
	659	+ // title outright.
	660	+ return null === $stripped ? $template : (string) $stripped;
436	661	}
437	662	}