← All changes | includes/seo/class-pattern-resolver.php +74 -6 2.0.1 → 2.7.0 View file →

		@@ -82,8 +82,67 @@
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	+ /**
86	145	* Derive a description from raw post content.
87	146	*
88	147	* wp_strip_all_tags() removes HTML but not shortcodes, so a page built with
89	148	* them published its shortcode source as the description — `[woocommerce_cart]`
		@@ -106,9 +165,14 @@
106	165	$text = excerpt_remove_blocks($content);
107	166	$text = strip_shortcodes($text);
108	167	$text = wp_strip_all_tags($text);
109	168
110		- 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));
111	175	}
112	176
113	177	/**
114	178	* Resolve any variable-tag string against a term's values.
		@@ -156,10 +220,12 @@
156	220	'%title%' => $name,
157	221	'%term%' => $name,
158	222	'%sitename%' => get_bloginfo('name'),
159	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).
160	226	'%excerpt%' => $description !== ''
161		- ? wp_trim_words(wp_strip_all_tags($description), 25, '...')
	227	+ ? self::derive_excerpt($description)
162	228	: '',
163	229	'%date%' => '',
164	230	'%modified%' => '',
165	231	'%author%' => '',
		@@ -194,11 +260,13 @@
194	260	public static function description(int $post_id): string {
195	261	$template = self::template_for($post_id, 'description', self::DEFAULT_DESCRIPTION);
196	262	$description = self::resolve_value($template, $post_id);
197	263
198		- if (strlen($description) > 160) {
199		- $description = wp_trim_words($description, 25, '...');
200		- }
	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);
201	269
202	270	return $description;
203	271	}
204	272
		@@ -313,9 +381,9 @@
313	381	$excerpt = '';
314	382	if ($post) {
315	383	$excerpt = !empty($post->post_excerpt)
316	384	? $post->post_excerpt
317		- : self::derive_excerpt((string) $post->post_content);
	385	+ : self::derive_excerpt(Builder_Content::visible_content($post));
318	386	}
319	387
320	388	$author_id = (int) get_post_field('post_author', $post_id);
321	389