orascheandClaude Fable 5
5c640d02f7
17 SEO-Artikel tragen ein rohes FAQPage-JSON-Objekt in einem abschliessenden Absatz ihres Bodytexts - das Frontend renderte es als sichtbaren Text, Google sah es nie als strukturierte Daten. Neue Klasse Seo\EmbeddedJsonLd trennt solche schema.org-Bloecke aus dem RTE-HTML (fail-safe: was nicht sauber dekodiert, bleibt stehen): - NewsJsonRenderer: Detail-bodytext und Listen-Exzerpt ohne die Bloecke - PageJsonLdRenderer: die dekodierten Nodes wandern in den @graph der Seite (FAQPage landet damit als <script type=application/ld+json> im Head, den das Frontend bereits injiziert) - VitecMetaHandler: Description-Fallback aus dem Bodytext ohne JSON-Muell Verifiziert an 3 Artikeln: @graph mit NewsArticle + FAQPage (8-9 Fragen), kein FAQPage-Rest im Textkoerper, kein sichtbarer JSON-Absatz im HTML. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
57 lines
2.0 KiB
PHP
57 lines
2.0 KiB
PHP
<?php
|
|
|
|
declare(strict_types=1);
|
|
|
|
namespace Evomedien\Vitec\Seo;
|
|
|
|
/**
|
|
* Splits editor-pasted schema.org JSON blocks out of RTE HTML.
|
|
*
|
|
* The SEO-maintained news articles carry a raw FAQPage JSON object in a
|
|
* trailing <p>…</p> of their bodytext (found 2026-11-10: 17 records). Left
|
|
* in place, the frontend renders the JSON as visible text while Google
|
|
* never sees it as structured data. This helper removes such paragraphs
|
|
* from the HTML and returns the decoded nodes, ready to join the page's
|
|
* JSON-LD @graph (their own @context is dropped - the graph carries one).
|
|
*
|
|
* Fail-safe: a paragraph that merely looks like JSON but does not decode
|
|
* to a schema.org node stays in the HTML untouched.
|
|
*/
|
|
final class EmbeddedJsonLd
|
|
{
|
|
/**
|
|
* @return array{html: string, nodes: list<array<string,mixed>>}
|
|
*/
|
|
public static function extract(string $html): array
|
|
{
|
|
if ($html === '' || !str_contains($html, 'schema.org')) {
|
|
return ['html' => $html, 'nodes' => []];
|
|
}
|
|
|
|
$nodes = [];
|
|
$clean = preg_replace_callback(
|
|
'#<p[^>]*>\s*(\{.+?\})\s*</p>#s',
|
|
static function (array $match) use (&$nodes): string {
|
|
$candidate = $match[1];
|
|
if (!str_contains($candidate, 'schema.org')) {
|
|
return $match[0];
|
|
}
|
|
$decoded = json_decode($candidate, true);
|
|
if (!is_array($decoded)) {
|
|
// RTE processing may have entity-encoded the quotes.
|
|
$decoded = json_decode(html_entity_decode($candidate, ENT_QUOTES | ENT_HTML5), true);
|
|
}
|
|
if (!is_array($decoded) || empty($decoded['@type'])) {
|
|
return $match[0];
|
|
}
|
|
unset($decoded['@context']);
|
|
$nodes[] = $decoded;
|
|
return '';
|
|
},
|
|
$html
|
|
) ?? $html;
|
|
|
|
return ['html' => trim($clean), 'nodes' => $nodes];
|
|
}
|
|
}
|