← All changes | includes/seo/class-pattern-resolver.php +197 -8 2.9.0 → 2.11.0 View file →

		@@ -392,21 +392,148 @@
392	392	$categories = get_the_category($post_id);
393	393	$category = !empty($categories) ? $categories[0]->name : '';
394	394	}
395	395
	396	+ return array_merge(
	397	+ [
	398	+ '%title%' => get_the_title($post_id),
	399	+ '%sitename%' => get_bloginfo('name'),
	400	+ '%sep%' => self::separator(),
	401	+ '%excerpt%' => $excerpt,
	402	+ '%date%' => get_the_date('', $post_id),
	403	+ '%modified%' => get_the_modified_date('', $post_id),
	404	+ '%author%' => $author_id ? get_the_author_meta('display_name', $author_id) : '',
	405	+ '%category%' => $category,
	406	+ ],
	407	+ self::product_placeholders($post_id)
	408	+ );
	409	+ }
	410	+
	411	+ /**
	412	+ * Tokens that only mean anything on a WooCommerce product.
	413	+ *
	414	+ * Rank Math and Yoast WooCommerce SEO both let a product title or
	415	+ * description carry the price, the SKU and the stock status, and both of
	416	+ * our converters dropped every token they did not recognise — so
	417	+ * "Buy %title% for %wc_price%" imported as "Buy %title% for" and the
	418	+ * customer in support #171748 found four variables where they had had a
	419	+ * stock one (#715).
	420	+ *
	421	+ * Always present, never conditional on the post type. A token that exists
	422	+ * on a product and is an unknown token everywhere else would resolve on
	423	+ * one page and leak literally on another; resolving to an empty string off
	424	+ * a product is the same answer every other token gives when it has nothing
	425	+ * to say, and process() then collapses the separator it leaves behind.
	426	+ *
	427	+ * @since 2.10.1
	428	+ *
	429	+ * @param int $post_id Post ID.
	430	+ * @return array<string,string>
	431	+ */
	432	+ private static function product_placeholders(int $post_id): array {
	433	+ $empty = [
	434	+ '%price%' => '',
	435	+ '%sale_price%' => '',
	436	+ '%sku%' => '',
	437	+ '%stock_status%' => '',
	438	+ '%short_description%' => '',
	439	+ '%brand%' => '',
	440	+ ];
	441	+
	442	+ if (!function_exists('wc_get_product') || 'product' !== get_post_type($post_id)) {
	443	+ return $empty;
	444	+ }
	445	+
	446	+ $product = wc_get_product($post_id);
	447	+ if (!$product) {
	448	+ return $empty;
	449	+ }
	450	+
	451	+ // Read through wc_price()/get_price_html() rather than formatting the
	452	+ // number here: currency symbol, position, decimals and the "from X"
	453	+ // form on a variable product are all store settings, and a second
	454	+ // formatter is how a template starts disagreeing with the price shown
	455	+ // three lines below it on the same page.
	456	+ $price = (string) $product->get_price();
	457	+ $sale_price = (string) $product->get_sale_price();
	458	+
	459	+ $stock_status = (string) $product->get_stock_status();
	460	+ $stock_labels = [
	461	+ 'instock' => __('In stock', 'thinkrank'),
	462	+ 'outofstock' => __('Out of stock', 'thinkrank'),
	463	+ 'onbackorder' => __('On backorder', 'thinkrank'),
	464	+ ];
	465	+
396	466	return [
397		- '%title%' => get_the_title($post_id),
398		- '%sitename%' => get_bloginfo('name'),
399		- '%sep%' => self::separator(),
400		- '%excerpt%' => $excerpt,
401		- '%date%' => get_the_date('', $post_id),
402		- '%modified%' => get_the_modified_date('', $post_id),
403		- '%author%' => $author_id ? get_the_author_meta('display_name', $author_id) : '',
404		- '%category%' => $category,
	467	+ '%price%' => self::formatted_price($price),
	468	+ '%sale_price%' => self::formatted_price($sale_price),
	469	+ '%sku%' => (string) $product->get_sku(),
	470	+ '%stock_status%' => $stock_labels[$stock_status] ?? $stock_status,
	471	+ '%short_description%' => self::derive_excerpt((string) $product->get_short_description()),
	472	+ '%brand%' => self::product_brand($post_id),
405	473	];
406	474	}
407	475
408	476	/**
	477	+ * A price as text, in the store's own currency format.
	478	+ *
	479	+ * wc_price() returns markup, and stripping the tags off it leaves the
	480	+ * entities behind: a title read "11.05&#2547;&nbsp;" on the first live
	481	+ * run. Entities are decoded and the non-breaking space collapsed, because
	482	+ * this value ends up inside a `<title>` and a meta description, where
	483	+ * markup has no meaning and an entity is just noise a reader sees.
	484	+ *
	485	+ * @since 2.10.1
	486	+ *
	487	+ * @param string $price Raw price, or '' when the product has none.
	488	+ * @return string
	489	+ */
	490	+ private static function formatted_price(string $price): string {
	491	+ if ('' === $price) {
	492	+ return '';
	493	+ }
	494	+
	495	+ $text = wp_strip_all_tags((string) wc_price((float) $price));
	496	+ $text = html_entity_decode($text, ENT_QUOTES | ENT_HTML5, 'UTF-8');
	497	+
	498	+ // \xC2\xA0 is the non-breaking space wc_price() puts between the
	499	+ // amount and the symbol; a literal one in a title is invisible to a
	500	+ // reader and awkward for everything else.
	501	+ $text = str_replace("\xC2\xA0", ' ', $text);
	502	+
	503	+ return trim((string) preg_replace('/\s+/u', ' ', $text));
	504	+ }
	505	+
	506	+ /**
	507	+ * A product's brand, from whichever taxonomy the store uses for one.
	508	+ *
	509	+ * WooCommerce core added `product_brand` in 9.4; before that every brand
	510	+ * plugin shipped its own taxonomy, and a store that migrated from one of
	511	+ * them still has the old terms. Asking each in turn costs one cached term
	512	+ * lookup and means the token is not empty on the stores most likely to
	513	+ * have used a brand token in the plugin they are leaving.
	514	+ *
	515	+ * @since 2.10.1
	516	+ *
	517	+ * @param int $post_id Product ID.
	518	+ * @return string
	519	+ */
	520	+ private static function product_brand(int $post_id): string {
	521	+ foreach (['product_brand', 'pwb-brand', 'yith_product_brand', 'berocket_brand'] as $taxonomy) {
	522	+ if (!taxonomy_exists($taxonomy)) {
	523	+ continue;
	524	+ }
	525	+
	526	+ $terms = get_the_terms($post_id, $taxonomy);
	527	+ if (is_array($terms) && !empty($terms)) {
	528	+ return (string) $terms[0]->name;
	529	+ }
	530	+ }
	531	+
	532	+ return '';
	533	+ }
	534	+
	535	+ /**
409	536	* Active title separator symbol.
410	537	*
411	538	* @return string Separator.
412	539	*/
		@@ -424,8 +551,17 @@
424	551	* @param array<string,string> $placeholders Placeholder map.
425	552	* @return string Resolved string.
426	553	*/
427	554	private static function process(string $template, array $placeholders): string {
	555	+ // Both forms go BEFORE substitution, and the strip is decided against
	556	+ // the placeholder map rather than by what is left over afterwards.
	557	+ // Running it on the substituted string scanned the resolved VALUES too,
	558	+ // so a post whose own title read "Using %name% placeholders in
	559	+ // %%mustache%% templates" published "Using placeholders in %%
	560	+ // templates" — its title edited, and a stray double percent where the
	561	+ // inner token had been eaten out of the middle of one.
	562	+ $template = self::strip_unresolved_tokens($template, $placeholders);
	563	+
428	564	$value = str_replace(array_keys($placeholders), array_values($placeholders), $template);
429	565
430	566	// Collapse whitespace.
431	567	$value = preg_replace('/\s+/', ' ', $value);
		@@ -441,6 +577,59 @@
441	577	);
442	578
443	579	// Strip leading/trailing separators and whitespace.
444	580	return trim($value, " \t\n\r\0\x0B" . $separator);
	581	+ }
	582	+
	583	+ /**
	584	+ * Remove any token the placeholder map did not resolve.
	585	+ *
	586	+ * The last line of defence, and the reason it exists is that every layer
	587	+ * above it is a list someone has to remember to extend. A converter that
	588	+ * misses a token, a template typed by hand, a value written straight into
	589	+ * postmeta by an importer we have not met: each one ends with `%%title%%`
	590	+ * or `%some_token%` rendering literally in a `<title>` on a live site, and
	591	+ * that is precisely what was reported on 14 September (#715).
	592	+ *
	593	+ * Both syntaxes, because a migrated site carries both: Yoast's `%%x%%`
	594	+ * (which includes Rank Math tokens Yoast's own importer wrapped in double
	595	+ * percent signs without translating them) and the single-percent form
	596	+ * ThinkRank and Rank Math share. `%%x%%` is matched as a whole so the pass
	597	+ * cannot eat the inner `%x%` and leave a stray percent sign at each end.
	598	+ *
	599	+ * Applied to the TEMPLATE, and a token is kept only when the placeholder
	600	+ * map has it. Stripping whatever still looked like a token after
	601	+ * substitution read the resolved values as well, and content is not a
	602	+ * template: a post titled "Using %name% placeholders in %%mustache%%
	603	+ * templates" published "Using placeholders in %% templates", and an excerpt
	604	+ * of "Learn %name% and %fabric% placeholders." published "Learn and
	605	+ * placeholders." Deciding against the map also means a token this resolver
	606	+ * knows about is never at risk, whatever a value happens to contain.
	607	+ *
	608	+ * A bare percent is left alone: "50% off" is ordinary copy, and a guard
	609	+ * that ate it would be a worse bug than the one it prevents.
	610	+ *
	611	+ * @since 2.10.1
	612	+ *
	613	+ * @param string $template Raw template.
	614	+ * @param array<string, string> $placeholders Tokens this resolver can resolve.
	615	+ * @return string
	616	+ */
	617	+ private static function strip_unresolved_tokens(string $template, array $placeholders): string {
	618	+ if (false === strpos($template, '%')) {
	619	+ return $template;
	620	+ }
	621	+
	622	+ $stripped = preg_replace_callback(
	623	+ '/%%[a-z0-9_-]+%%|%[a-z0-9_-]+%/i',
	624	+ static function (array $found) use ($placeholders): string {
	625	+ return array_key_exists($found[0], $placeholders) ? $found[0] : '';
	626	+ },
	627	+ $template
	628	+ );
	629	+
	630	+ // preg_replace_callback() answers null on content that is not valid
	631	+ // UTF-8 rather than throwing, and returning null here would blank a
	632	+ // title outright.
	633	+ return null === $stripped ? $template : (string) $stripped;
445	634	}
446	635	}