← All changes | includes/ai/class-seo-score-calculator.php +210 -14 2.10.0 → 2.13.0 View file →

		@@ -140,8 +140,9 @@
140	140	'grade' => $result['grade'],
141	141	]];
142	142	$result['keyword_checks'] = $this->analyze_keyword_checks($content_data, $metadata, $keywords);
143	143	}
	144	+ $result['keywords'] = $this->keyword_placements($content_data, $metadata, $keywords);
144	145
145	146	return $result;
146	147	}
147	148
		@@ -251,8 +252,9 @@
251	252	$best['target_keyword'] = $best_keyword;
252	253	$best['target_keywords'] = $keywords;
253	254	$best['keyword_results'] = $per_keyword;
254	255	$best['keyword_checks'] = $this->analyze_keyword_checks($content_data, $metadata, $keywords);
	256	+ $best['keywords'] = $this->keyword_placements($content_data, $metadata, $keywords);
255	257
256	258	return $best;
257	259	}
258	260
		@@ -267,15 +269,15 @@
267	269	* @param string[] $keywords Target keywords.
268	270	* @return array<string,array{passed:bool,matched_keywords:string[]}>
269	271	*/
270	272	private function analyze_keyword_checks(array $content_data, array $metadata, array $keywords): array {
271		- $title = strtolower((string) ($metadata['title'] ?? $content_data['title'] ?? ''));
272		- $description = strtolower((string) ($metadata['description'] ?? ''));
273		- $content = strtolower(wp_strip_all_tags((string) ($content_data['content'] ?? '')));
	273	+ $title = self::lower((string) ($metadata['title'] ?? $content_data['title'] ?? ''));
	274	+ $description = self::lower((string) ($metadata['description'] ?? ''));
	275	+ $content = self::lower(self::plain_text((string) ($content_data['content'] ?? '')));
274	276
275	277	$alts = '';
276	278	foreach ((array) ($content_data['images'] ?? []) as $image) {
277		- $alts .= ' ' . strtolower((string) ($image['alt'] ?? ''));
	279	+ $alts .= ' ' . self::lower((string) ($image['alt'] ?? ''));
278	280	}
279	281
280	282	// Build a searchable slug haystack from the post's OWN slug — never the
281	283	// full URL path. The path carries ancestors, category bases and date
		@@ -284,17 +286,9 @@
284	286	// way: an unpublished post has no pretty permalink (get_permalink()
285	287	// returns ?p=123), so the path held no slug at all and every draft
286	288	// scored "no match" until it was published. Hyphens/underscores become
287	289	// spaces so multi-word keywords can match.
288		- $slug_source = (string) ($content_data['slug'] ?? '');
289		- if ($slug_source === '') {
290		- // Draft with no slug assigned yet: score what WordPress would
291		- // generate from the title, which is what the editor shows as the
292		- // proposed URL — so the check reads the same before and after
293		- // publishing instead of flipping.
294		- $slug_source = sanitize_title((string) ($content_data['title'] ?? ''));
295		- }
296		- $slug = strtolower(str_replace(['-', '_'], ' ', $slug_source));
	290	+ $slug = self::lower(self::slug_haystack($content_data));
297	291
298	292	$haystacks = [
299	293	'title' => trim($title),
300	294	'meta_description' => trim($description),
		@@ -306,9 +300,9 @@
306	300	$checks = [];
307	301	foreach ($haystacks as $location => $haystack) {
308	302	$matched = [];
309	303	foreach ($keywords as $keyword) {
310		- if ($this->keyword_matches($haystack, strtolower(trim($keyword)))) {
	304	+ if ($this->keyword_matches($haystack, self::lower(trim($keyword)))) {
311	305	$matched[] = $keyword;
312	306	}
313	307	}
314	308	$checks[$location] = [
		@@ -317,8 +311,210 @@
317	311	];
318	312	}
319	313
320	314	return $checks;
	315	+ }
	316	+
	317	+ /**
	318	+ * The keyword placements the editor draws one gauge segment for, in the
	319	+ * order a reader meets them (#729).
	320	+ *
	321	+ * @since 2.11.0
	322	+ * @var string[]
	323	+ */
	324	+ public const PLACEMENTS = ['title', 'meta_description', 'slug', 'first_paragraph', 'subheading', 'content', 'image_alt'];
	325	+
	326	+ /**
	327	+ * Characters of plain text read as the opening when the content has no
	328	+ * paragraph tag.
	329	+ *
	330	+ * @since 2.11.0
	331	+ */
	332	+ private const OPENING_CHARS = 300;
	333	+
	334	+ /**
	335	+ * Where each focus keyword is placed, keyword by keyword (#729).
	336	+ *
	337	+ * analyze_keyword_checks() answers "does ANY keyword appear here" for
	338	+ * five places; this answers "where does THIS keyword appear" for seven,
	339	+ * so the editor can show each keyword's own gauge. Same matcher, so a
	340	+ * keyword counts as a word (not a fragment) and a keyword in a script
	341	+ * written without spaces (Thai, Chinese, Japanese) still matches.
	342	+ *
	343	+ * `where` names the heading or alt text that matched, so the editor can
	344	+ * say which one.
	345	+ *
	346	+ * @since 2.11.0
	347	+ *
	348	+ * @param array $content_data Content analysis data.
	349	+ * @param array $metadata Post metadata (title, description).
	350	+ * @param string[] $keywords Focus keywords.
	351	+ * @return array<int, array{keyword: string, passed: int, total: int, placements: array<string, array{passed: bool, where: string}>}>
	352	+ */
	353	+ public function keyword_placements(array $content_data, array $metadata, array $keywords): array {
	354	+ $html = (string) ($content_data['content'] ?? '');
	355	+ $plain = self::lower(self::plain_text($html));
	356	+
	357	+ $headings = [];
	358	+ $source = isset($content_data['headings']) && is_array($content_data['headings']) ? $content_data['headings'] : $this->extract_headings($html);
	359	+ foreach ($source as $heading) {
	360	+ $text = trim((string) ($heading['text'] ?? ''));
	361	+ if ((int) ($heading['level'] ?? 0) >= 2 && '' !== $text) {
	362	+ $headings[] = $text;
	363	+ }
	364	+ }
	365	+
	366	+ $alts = [];
	367	+ foreach ((array) ($content_data['images'] ?? []) as $image) {
	368	+ $alt = trim((string) ($image['alt'] ?? ''));
	369	+ if ('' !== $alt) {
	370	+ $alts[] = $alt;
	371	+ }
	372	+ }
	373	+
	374	+ $single = [
	375	+ 'title' => self::lower((string) ($metadata['title'] ?? $content_data['title'] ?? '')),
	376	+ 'meta_description' => self::lower((string) ($metadata['description'] ?? '')),
	377	+ 'slug' => self::lower(self::slug_haystack($content_data)),
	378	+ 'first_paragraph' => self::lower(self::opening($html)),
	379	+ 'content' => $plain,
	380	+ ];
	381	+
	382	+ $out = [];
	383	+ foreach ($keywords as $keyword) {
	384	+ $needle = self::lower(trim((string) $keyword));
	385	+ $placements = [];
	386	+
	387	+ foreach (self::PLACEMENTS as $placement) {
	388	+ if ('subheading' === $placement || 'image_alt' === $placement) {
	389	+ $where = '';
	390	+ foreach ('subheading' === $placement ? $headings : $alts as $text) {
	391	+ if ($this->keyword_matches(self::lower($text), $needle)) {
	392	+ $where = $text;
	393	+ break;
	394	+ }
	395	+ }
	396	+ $placements[$placement] = ['passed' => '' !== $where, 'where' => $where];
	397	+ continue;
	398	+ }
	399	+
	400	+ $placements[$placement] = ['passed' => $this->keyword_matches($single[$placement], $needle), 'where' => ''];
	401	+ }
	402	+
	403	+ $out[] = [
	404	+ 'keyword' => (string) $keyword,
	405	+ 'passed' => count(array_filter(array_column($placements, 'passed'))),
	406	+ 'total' => count(self::PLACEMENTS),
	407	+ 'placements' => $placements,
	408	+ ];
	409	+ }
	410	+
	411	+ return $out;
	412	+ }
	413	+
	414	+ /**
	415	+ * The post's own slug as searchable text: hyphens and underscores become
	416	+ * spaces so a multi-word keyword can match, and a slug WordPress
	417	+ * percent-encoded (Thai, Cyrillic, Chinese) is decoded, or it could never
	418	+ * match a keyword typed in that script.
	419	+ *
	420	+ * Never the full URL path: the path carries ancestors, category bases and
	421	+ * date segments, so a child of /clinical-trials/ reported "keyword in slug"
	422	+ * for a page actually slugged `contact-us`. An unpublished post has no
	423	+ * pretty permalink either, so the path held no slug at all.
	424	+ *
	425	+ * @param array $content_data Content analysis data.
	426	+ * @return string
	427	+ */
	428	+ private static function slug_haystack(array $content_data): string {
	429	+ $slug = (string) ($content_data['slug'] ?? '');
	430	+ if ('' === $slug) {
	431	+ // Draft with no slug assigned yet: score what WordPress would
	432	+ // generate from the title, which is what the editor shows as the
	433	+ // proposed URL — so the check reads the same before and after
	434	+ // publishing instead of flipping.
	435	+ $slug = sanitize_title((string) ($content_data['title'] ?? ''));
	436	+ }
	437	+
	438	+ return trim(str_replace(['-', '_'], ' ', rawurldecode($slug)));
	439	+ }
	440	+
	441	+ /**
	442	+ * The opening of the content: its first paragraph with text, or the first
	443	+ * few hundred characters when it has none. Counted in characters, not
	444	+ * words, so a language written without spaces is not read as one word.
	445	+ *
	446	+ * @param string $html Content HTML.
	447	+ * @return string Plain text.
	448	+ */
	449	+ private static function opening(string $html): string {
	450	+ if (preg_match_all('/<p\b[^>]*>(.*?)<\/p>/isu', $html, $matches)) {
	451	+ foreach ($matches[1] as $paragraph) {
	452	+ $text = self::collapse_whitespace(wp_strip_all_tags($paragraph));
	453	+ if ('' !== $text) {
	454	+ return $text;
	455	+ }
	456	+ }
	457	+ }
	458	+
	459	+ $text = self::plain_text($html);
	460	+
	461	+ return function_exists('mb_substr') ? mb_substr($text, 0, self::OPENING_CHARS) : substr($text, 0, self::OPENING_CHARS);
	462	+ }
	463	+
	464	+ /**
	465	+ * Content as plain text, with a space where each tag was. wp_strip_all_tags()
	466	+ * alone joins neighbouring blocks — "…coffee grinder</h3><p>A good…" became
	467	+ * "coffee grinderA good" — so a keyword at the end of a heading or a
	468	+ * paragraph was no longer a word and did not match.
	469	+ *
	470	+ * @param string $html Content HTML.
	471	+ * @return string
	472	+ */
	473	+ private static function plain_text(string $html): string {
	474	+ $spaced = preg_replace('/<[^>]+>/', ' $0 ', $html);
	475	+
	476	+ return self::collapse_whitespace(wp_strip_all_tags(null === $spaced ? $html : (string) $spaced));
	477	+ }
	478	+
	479	+ /**
	480	+ * Runs of whitespace down to one space.
	481	+ *
	482	+ * The `/u` pass is the one that understands a multibyte space, but
	483	+ * preg_replace() answers null on bytes that are not valid UTF-8 rather than
	484	+ * throwing — and casting that null to a string blanked the haystack, so a
	485	+ * post carrying one mojibake byte (a Latin-1 paste, an old import) reported
	486	+ * every keyword as missing from its body, its opening, and every
	487	+ * subheading. The gauge said 0/7 and told the author to add a keyword that
	488	+ * was already there.
	489	+ *
	490	+ * Falls back to the byte-wise collapse, which is what this did before the
	491	+ * multibyte work added the modifier. Same reasoning keyword_matches()
	492	+ * already records for its own PCRE failure: a pattern PCRE refuses must not
	493	+ * be reported as a confident "no match".
	494	+ *
	495	+ * @param string $text Text to collapse.
	496	+ * @return string
	497	+ */
	498	+ private static function collapse_whitespace(string $text): string {
	499	+ $collapsed = preg_replace('/\s+/u', ' ', $text);
	500	+
	501	+ if (null === $collapsed) {
	502	+ $collapsed = preg_replace('/\s+/', ' ', $text);
	503	+ }
	504	+
	505	+ return trim(null === $collapsed ? $text : (string) $collapsed);
	506	+ }
	507	+
	508	+ /**
	509	+ * Lowercase in any script. strtolower() only folds ASCII, so "Кофе" never
	510	+ * matched "кофе".
	511	+ *
	512	+ * @param string $text Text.
	513	+ * @return string
	514	+ */
	515	+ private static function lower(string $text): string {
	516	+ return function_exists('mb_strtolower') ? mb_strtolower($text, 'UTF-8') : strtolower($text);
321	517	}
322	518
323	519	/**
324	520	* Scripts written without spaces between words.