GDPR-Content-Blocker/gdpr-content-blocker/includes/class-autodetect.php

<?php
defined( 'ABSPATH' ) || exit;

/**
 * Auto-detects iframes in the rendered HTML and replaces them with consent
 * placeholders based on configured match patterns.
 *
 * Strategy (safe by design):
 *   1. A regex LOCATES each <iframe>…</iframe> block (it does NOT parse attributes).
 *   2. Each block is parsed individually with DOMDocument to read the src reliably.
 *   3. Only the matched iframe block is replaced in the original HTML via callback.
 *
 * The rest of the page HTML is never re-serialized, so themes, page builders,
 * inline JS and JSON-LD stay byte-for-byte intact.
 *
 * Known limitation: Only iframes present in the initial server-rendered HTML are
 * covered here. For iframes injected later by JavaScript, use the manual shortcode
 * [content_blocker id="…"] around the embed.
 */
class CB_Autodetect {

	public static function init(): void {
		add_action( 'template_redirect', [ __CLASS__, 'start_buffer' ] );
	}

	public static function start_buffer(): void {
		if ( is_admin() || wp_doing_ajax() || wp_doing_cron() ) {
			return;
		}
		if ( defined( 'REST_REQUEST' ) && REST_REQUEST ) {
			return;
		}
		if ( defined( 'XMLRPC_REQUEST' ) && XMLRPC_REQUEST ) {
			return;
		}

		$services = CB_Settings::get_services();
		$active   = array_values( array_filter(
			$services,
			fn( $s ) => ! empty( $s['match_pattern'] ) && ( $s['enabled'] ?? true )
		) );
		if ( empty( $active ) ) {
			return;
		}

		ob_start( fn( string $html ) => self::process( $html, $active ) );
	}

	public static function process( string $html, array $services ): string {
		if ( $html === '' || stripos( $html, '<iframe' ) === false ) {
			return $html;
		}

		// Locate iframe blocks only. Attribute parsing happens via DOMDocument below.
		$out = preg_replace_callback(
			'#<iframe\b[^>]*>.*?</iframe>#is',
			function ( array $m ) use ( $services ): string {
				return self::maybe_replace_iframe( $m[0], $services );
			},
			$html
		);

		// On a PCRE error (e.g. backtrack/recursion limit on a huge page),
		// preg_replace_callback returns null. Never blank the page — fall back to
		// the original, unmodified HTML.
		return $out === null ? $html : $out;
	}

	/**
	 * Parse a single iframe block with DOMDocument, read its src, and return either
	 * the consent placeholder (on match) or the unchanged original block.
	 */
	private static function maybe_replace_iframe( string $iframe_html, array $services ): string {
		$src = self::get_src( $iframe_html );
		if ( $src === '' ) {
			return $iframe_html;
		}

		foreach ( $services as $svc ) {
			$pattern = $svc['match_pattern'] ?? '';
			if ( $pattern !== '' && str_contains( $src, $pattern ) ) {
				$attrs = CB_Renderer::extract_iframe_attrs( $iframe_html );
				$dims  = [ 'width' => $attrs['width'], 'height' => $attrs['height'] ];
				return CB_Renderer::render_placeholder( $svc, $src, $dims );
			}
		}

		return $iframe_html;
	}

	/** Reliably extract the src attribute of a single iframe via DOMDocument. */
	private static function get_src( string $iframe_html ): string {
		$dom = new DOMDocument();
		libxml_use_internal_errors( true );
		// Wrap so loadHTML has a clean context; force UTF-8 so umlauts survive.
		$dom->loadHTML(
			'<?xml encoding="utf-8"?><div>' . $iframe_html . '</div>',
			LIBXML_HTML_NOIMPLIED | LIBXML_HTML_NODEFDTD
		);
		libxml_clear_errors();

		$node = $dom->getElementsByTagName( 'iframe' )->item( 0 );
		if ( ! $node instanceof DOMElement ) {
			return '';
		}

		// getAttribute() already returns the entity-decoded value (e.g. &amp; → &),
		// so no further unescaping is needed here.
		return trim( $node->getAttribute( 'src' ) );
	}
}