WP-LIT-MCP/src/Integrations/Divi/Library/UsageMatcher.php

<?php

declare(strict_types=1);

namespace LIT\MCP\Integrations\Divi\Library;

defined('ABSPATH') || exit;

/**
 * Whether post content shows library entry $id (rulings P23, P24), fail closed:
 * - Divi 5: the textual "globalModule":"<digits>" Divi's render takes (leading zeros = intval), and
 *   every block opener whose JSON – in any spelling, escaped keys or values included – decodes to a
 *   globalModule with that ID at any depth;
 * - Divi 4 shortcodes – in the content outside block openers and in every decoded string value of an
 *   opener (Divi runs them there too): attributes parsed like shortcode_parse_atts(), global_module
 *   cast with (int) like Divi; a non-empty value other than "0" that casts to 0 makes Divi render the
 *   newest layout (et_pb_load_global_module, WP_Query p=0), so it counts for EVERY entry;
 * - an opener with JSON escapes (or a globalModule/global_module) that cannot be decoded, or content
 *   the regex engine cannot scan, counts as a possible reference.
 * referencesAny() asks the same for every entry at once (ruling Q7): content that becomes a global entry
 * must not render another entry – nor, through a value that casts to 0, the newest layout, which may be
 * that entry itself.
 * Pure PHP.
 */
final class UsageMatcher
{
    /** A block opener and its attribute JSON (the Parser's delimiter grammar). */
    private const OPENER = '/<!--\s+wp:(?:[a-z][a-z0-9_-]*\/)?[a-z][a-z0-9_-]*\s+'
        . '({(?:(?:[^}]+|}+(?=})|(?!}\s+\/?-->).)*+)?})\s+\/?-->/s';

    /** An opening shortcode tag and its attribute text (get_shortcode_regex(), any tag name). */
    private const TAG = '/\[\[?([\w-]+)(?![\w-])([^\]\/]*(?:\/(?!\])[^\]\/]*)*?)\/?\]/';

    /** WordPress' get_shortcode_atts_regex(). */
    private const ATTS = '/([\w-]+)\s*=\s*"([^"]*)"(?:\s|$)|([\w-]+)\s*=\s*\'([^\']*)\'(?:\s|$)|([\w-]+)\s*=\s*'
        . '([^\s\'"]+)(?:\s|$)|"([^"]*)"(?:\s|$)|\'([^\']*)\'(?:\s|$)|(\S+)(?:\s|$)/';

    public static function shows(string $content, int $id): bool
    {
        return self::matches($content, $id);
    }

    /**
     * Whether $content renders any library entry, in every spelling shows() knows (ruling Q7): a
     * globalModule key with any value, a global_module attribute Divi does not skip; fail closed.
     */
    public static function referencesAny(string $content): bool
    {
        return self::matches($content, null);
    }

    /** shows() for entry $id, or for any entry when $id is null. */
    private static function matches(string $content, ?int $id): bool
    {
        $textual = preg_match(
            $id === null ? '/"globalModule"\s*:/' : '/"globalModule":"0*' . $id . '"/',
            $content
        );
        if ($textual !== 0) {
            return true; // A match, or a regex failure (fail closed).
        }

        $found = preg_match_all(self::OPENER, $content, $matches);
        if ($found === false) {
            return true;
        }

        // Shortcodes outside the openers (Divi 4 pages, free text between blocks).
        $outside = $found === 0 ? $content : str_replace($matches[1], '', $content);
        if (self::shortcodeShows($outside, $id)) {
            return true;
        }

        $escape = chr(92) . 'u';
        foreach ($matches[1] as $json) {
            if (
                ! str_contains($json, $escape)
                && ! str_contains($json, 'globalModule')
                && stripos($json, 'global_module') === false
            ) {
                continue;
            }
            $attrs = json_decode($json, true);
            if (! is_array($attrs) || self::names($attrs, $id)) {
                return true;
            }
            foreach (self::strings($attrs) as $value) {
                if (self::shortcodeShows($value, $id)) {
                    return true;
                }
            }
        }

        return false;
    }

    /** Whether a Divi 4 shortcode in $text renders entry $id (null: any entry); a value casting to 0 renders any. */
    private static function shortcodeShows(string $text, ?int $id): bool
    {
        if (stripos($text, 'global_module') === false) {
            return false;
        }

        $found = preg_match_all(self::TAG, $text, $tags);
        if ($found === false) {
            return true;
        }
        foreach ($tags[2] as $attributes) {
            $atts = self::parseAtts($attributes);
            if ($atts === null) {
                return true;
            }
            $value = $atts['global_module'] ?? null;
            // Divi: ! empty( $global_module_id ), then WP_Query p => (int) $global_module_id.
            if (! is_string($value) || $value === '' || $value === '0') {
                continue;
            }
            $cast = (int) $value;
            if ($id === null || $cast === 0 || $cast === $id) {
                return true;
            }
        }

        return false;
    }

    /**
     * shortcode_parse_atts() of WordPress 6.9 (named attributes only); null when PCRE fails.
     *
     * @return array<string, string>|null
     */
    private static function parseAtts(string $text): ?array
    {
        $text = preg_replace('/[\x{00a0}\x{200b}]+/u', ' ', $text);
        if (! is_string($text)) {
            return null;
        }

        $found = preg_match_all(self::ATTS, $text, $matches, PREG_SET_ORDER);
        if ($found === false) {
            return null;
        }

        $atts = [];
        foreach ($matches as $m) {
            foreach ([1, 3, 5] as $key) {
                if (! empty($m[$key])) {
                    $atts[strtolower($m[$key])] = stripcslashes($m[$key + 1]);
                    break;
                }
            }
        }

        // WordPress rejects values with unclosed HTML elements.
        foreach ($atts as $name => $value) {
            if (str_contains($value, '<') && preg_match('/^[^<]*+(?:<[^>]*+>[^<]*+)*+$/', $value) !== 1) {
                $atts[$name] = '';
            }
        }

        return $atts;
    }

    /**
     * @param array<array-key, mixed> $attrs Whether a globalModule key at any depth names $id (null: any
     *        globalModule key, whatever its value – fail closed).
     */
    private static function names(array $attrs, ?int $id): bool
    {
        foreach ($attrs as $key => $value) {
            if ($id === null && $key === 'globalModule') {
                return true;
            }
            if (
                $key === 'globalModule'
                && (is_int($value) || (is_string($value) && preg_match('/^\d+$/D', $value) === 1))
                && (int) $value === $id
            ) {
                return true;
            }
            if (is_array($value) && self::names($value, $id)) {
                return true;
            }
        }

        return false;
    }

    /**
     * @param array<array-key, mixed> $attrs
     * @return list<string> Every string value at any depth.
     */
    private static function strings(array $attrs): array
    {
        $strings = [];
        array_walk_recursive($attrs, static function (mixed $value) use (&$strings): void {
            if (is_string($value)) {
                $strings[] = $value;
            }
        });

        return $strings;
    }
}