← All changes | includes/seo/class-pattern-resolver.php +157 -4 1.30.0 → 2.7.0 View file →

		@@ -82,8 +82,159 @@
82	82	return self::process($value, self::placeholders_for($post_id));
83	83	}
84	84
85	85	/**
	86	+ * Sanitize a variable-tag template for storage.
	87	+ *
	88	+ * The write-side counterpart of resolve_value(): every template that
	89	+ * reaches this class has to survive the trip into the database first.
	90	+ *
	91	+ * sanitize_text_field() cannot be used for that. Core's
	92	+ * _sanitize_text_fields() strips percent-encoded characters, looping
	93	+ * `preg_replace( '/%[a-f0-9]{2}/i', ... )` until nothing matches, so any
	94	+ * token whose first two characters are hex digits is eaten on save:
	95	+ * %date% is stored as "te%" and %category% as "tegory%" (#521). They are
	96	+ * the only two tags in the language that collide, which is why the
	97	+ * corruption looked arbitrary — %title%, %sitename%, %sep%, %excerpt%,
	98	+ * %modified% and %author% all pass through core untouched.
	99	+ *
	100	+ * This mirrors what core does either side of that percent loop — invalid
	101	+ * UTF-8 dropped, tags stripped, control characters removed, whitespace
	102	+ * collapsed — and simply omits the loop itself.
	103	+ *
	104	+ * @since 2.1.1
	105	+ *
	106	+ * @param string $value Raw template as submitted.
	107	+ * @param bool $keep_newlines Preserve newlines, as sanitize_textarea_field() does.
	108	+ * @return string Sanitized template with its %tokens% intact.
	109	+ */
	110	+ public static function sanitize_template(string $value, bool $keep_newlines = false): string {
	111	+ $filtered = wp_check_invalid_utf8($value);
	112	+
	113	+ if (strpos($filtered, '<') !== false) {
	114	+ $filtered = wp_pre_kses_less_than($filtered);
	115	+ // Tags out, the text between them kept.
	116	+ $filtered = wp_strip_all_tags($filtered, false);
	117	+ $filtered = str_replace("<\n", "&lt;\n", $filtered);
	118	+ }
	119	+
	120	+ // C0 controls and DEL, less the tab/newline/carriage-return handled below.
	121	+ $filtered = (string) preg_replace('/[\x00-\x08\x0B\x0C\x0E-\x1F\x7F]/', '', $filtered);
	122	+
	123	+ if (!$keep_newlines) {
	124	+ $filtered = (string) preg_replace('/[\r\n\t ]+/', ' ', $filtered);
	125	+ }
	126	+
	127	+ return trim($filtered);
	128	+ }
	129	+
	130	+ /**
	131	+ * Sanitize a variable-tag template that may span multiple lines.
	132	+ *
	133	+ * The sanitize_textarea_field() counterpart of sanitize_template().
	134	+ *
	135	+ * @since 2.1.1
	136	+ *
	137	+ * @param string $value Raw template as submitted.
	138	+ * @return string Sanitized template with its %tokens% and newlines intact.
	139	+ */
	140	+ public static function sanitize_template_textarea(string $value): string {
	141	+ return self::sanitize_template($value, true);
	142	+ }
	143	+
	144	+ /**
	145	+ * Derive a description from raw post content.
	146	+ *
	147	+ * wp_strip_all_tags() removes HTML but not shortcodes, so a page built with
	148	+ * them published its shortcode source as the description — `[woocommerce_cart]`
	149	+ * as the meta description, og:description and twitter:description of the
	150	+ * cart page. Core's own wp_trim_excerpt() runs strip_shortcodes() and
	151	+ * excerpt_remove_blocks() first; this path did neither, which is why the
	152	+ * two disagreed about the same post (#387).
	153	+ *
	154	+ * @since 2.0.1
	155	+ *
	156	+ * @param string $content Raw post content.
	157	+ * @param int $words Word cap.
	158	+ * @return string Derived description, or '' when nothing survives.
	159	+ */
	160	+ public static function derive_excerpt(string $content, int $words = 25): string {
	161	+ if ('' === trim($content)) {
	162	+ return '';
	163	+ }
	164	+
	165	+ $text = excerpt_remove_blocks($content);
	166	+ $text = strip_shortcodes($text);
	167	+ $text = wp_strip_all_tags($text);
	168	+
	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));
	175	+ }
	176	+
	177	+ /**
	178	+ * Resolve any variable-tag string against a term's values.
	179	+ *
	180	+ * The term counterpart of resolve_value(). Term SEO fields reach the
	181	+ * frontend from three writers — the term UI, the abilities API and the
	182	+ * Yoast/RankMath/AIOSEO/SEOPress importer — and the importers already
	183	+ * substitute their own term tokens (%%term_title%%, %term%) with the term
	184	+ * name at export time, so what lands here is either literal text or
	185	+ * ThinkRank's own tags.
	186	+ *
	187	+ * @since 2.0.1
	188	+ *
	189	+ * @param string $value Raw string, possibly containing variable tags.
	190	+ * @param int $term_id Term ID.
	191	+ * @return string Resolved string.
	192	+ */
	193	+ public static function resolve_term_value(string $value, int $term_id): string {
	194	+ if (strpos($value, '%') === false) {
	195	+ return $value;
	196	+ }
	197	+ return self::process($value, self::placeholders_for_term($term_id));
	198	+ }
	199	+
	200	+ /**
	201	+ * Token => value map for a term.
	202	+ *
	203	+ * The post-only tokens resolve to an empty string rather than being left
	204	+ * unreplaced: they have no meaning on an archive, and process() collapses
	205	+ * the separators an empty token leaves behind. A raw "%author%" in the
	206	+ * rendered title would be worse than nothing.
	207	+ *
	208	+ * @since 2.0.1
	209	+ *
	210	+ * @param int $term_id Term ID.
	211	+ * @return array<string,string> Placeholder map.
	212	+ */
	213	+ private static function placeholders_for_term(int $term_id): array {
	214	+ $term = get_term($term_id);
	215	+
	216	+ $name = ($term && !is_wp_error($term)) ? $term->name : '';
	217	+ $description = ($term && !is_wp_error($term)) ? (string) $term->description : '';
	218	+
	219	+ return [
	220	+ '%title%' => $name,
	221	+ '%term%' => $name,
	222	+ '%sitename%' => get_bloginfo('name'),
	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).
	226	+ '%excerpt%' => $description !== ''
	227	+ ? self::derive_excerpt($description)
	228	+ : '',
	229	+ '%date%' => '',
	230	+ '%modified%' => '',
	231	+ '%author%' => '',
	232	+ '%category%' => '',
	233	+ ];
	234	+ }
	235	+
	236	+ /**
86	237	* Token => value map for a post, keyed WITHOUT the surrounding percents
87	238	* (e.g. 'title' => 'My Post'). Used by the editor for live client-side
88	239	* preview of a pattern as the user types.
89	240	*
		@@ -109,11 +260,13 @@
109	260	public static function description(int $post_id): string {
110	261	$template = self::template_for($post_id, 'description', self::DEFAULT_DESCRIPTION);
111	262	$description = self::resolve_value($template, $post_id);
112	263
113		- if (strlen($description) > 160) {
114		- $description = wp_trim_words($description, 25, '...');
115		- }
	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);
116	269
117	270	return $description;
118	271	}
119	272
		@@ -228,9 +381,9 @@
228	381	$excerpt = '';
229	382	if ($post) {
230	383	$excerpt = !empty($post->post_excerpt)
231	384	? $post->post_excerpt
232		- : wp_trim_words(wp_strip_all_tags($post->post_content), 25, '...');
	385	+ : self::derive_excerpt(Builder_Content::visible_content($post));
233	386	}
234	387
235	388	$author_id = (int) get_post_field('post_author', $post_id);
236	389