<?php

declare(strict_types=1);

namespace AetosSeo\Utils;

if (!defined('ABSPATH')) {
    exit;
}

/**
 * v2.8.3 — Shared URL classification utility.
 *
 * Replaces the per-check duplicated lists of "excluded URLs" and
 * "legitimately short URLs" that grew in v2.8.1/v2.8.2 across
 * CrawlCheck, TechnicalSeoCheck, OnPageCheck. Single source of truth
 * makes it easy to add a new pattern in one place + pass it through a
 * filter so site owners can extend per-install.
 *
 * Usage:
 *   if (UrlClassifier::is_excluded($url)) continue;          // crawl/SEO finds
 *   if (UrlClassifier::is_legitimately_short($url)) continue; // thin-content
 *   if (UrlClassifier::is_archive_index($url)) continue;     // listing pages
 */
class UrlClassifier
{
    /** @var array<int,string>|null */
    private static $excluded_cache = null;

    /** @var array<int,string>|null */
    private static $legit_short_cache = null;

    /** @var array<int,string>|null */
    private static $archive_cache = null;

    /**
     * Auto-redirect routes, admin/cart/login pages, REST endpoints, feeds.
     * Pages crawled but excluded from SEO findings.
     *
     * @return array<int,string>
     */
    public static function excluded_patterns(): array
    {
        if (self::$excluded_cache !== null) {
            return self::$excluded_cache;
        }
        $defaults = [
            '#^/\d{8,}/?$#',                                        // numeric-only paths (phone numbers, IDs)
            '#^/(redirect|go|qr|out|track|r|u)/.+#i',                // redirect prefixes
            '#^/(login|admin|cart|checkout|my-account|wp-admin|wp-login)#i',
            '#^/wp-json/#i',                                        // REST endpoints
            '#^/feed/?$#i',                                         // feeds
            '#^/(tag|author)/.+#i',                                 // tag/author archives
            '#\?(p|page_id|preview)=#',                             // raw query-string only URLs
        ];
        if (function_exists('apply_filters')) {
            self::$excluded_cache = (array) apply_filters('aetos_seo_excluded_url_patterns', $defaults);
        } else {
            self::$excluded_cache = $defaults;
        }
        return self::$excluded_cache;
    }

    /**
     * URLs whose short content is legitimate (legal pages, contact pages,
     * cart/checkout, archive indexes). Used by thin-content checks.
     *
     * @return array<int,string>
     */
    public static function legitimately_short_substrings(): array
    {
        if (self::$legit_short_cache !== null) {
            return self::$legit_short_cache;
        }
        $defaults = [
            // Trust / legal
            '/contact', '/التواصل', '/اتصل-بنا', '/اتصل', '/تواصل',
            '/disclaimer', '/إخلاء-المسؤولية', '/اخلاء-المسؤولية',
            '/privacy', '/سياسة-الخصوصية',
            '/terms', '/الشروط', '/شروط-الاستخدام',
            '/refund', '/الاسترجاع', '/سياسة-الاسترجاع',
            '/cookies', '/كوكيز',
            // Commerce flow
            '/cart', '/سلة', '/checkout', '/الدفع', '/login', '/تسجيل-الدخول',
            '/register', '/إنشاء-حساب', '/my-account', '/حسابي',
            '/thank-you', '/شكرا', '/404', '/not-found',
            // Site machinery
            '/sitemap', '/خريطة-الموقع',
            '/sla', '/service-level-agreement', '/intellectual-property',
            '/dmca', '/gdpr', '/accessibility',
            '/مركز-المساعدة', '/مركز-المساعده', '/help-center', '/support',
        ];
        if (function_exists('apply_filters')) {
            self::$legit_short_cache = (array) apply_filters('aetos_seo_short_path_substrings', $defaults);
        } else {
            self::$legit_short_cache = $defaults;
        }
        return self::$legit_short_cache;
    }

    /**
     * Substrings that identify archive/listing pages. AI-Overview-style
     * checks (DirectAnswerPattern, AiOverview, QuestionCoverage) should
     * skip these since they aggregate content rather than answer queries.
     *
     * @return array<int,string>
     */
    public static function archive_substrings(): array
    {
        if (self::$archive_cache !== null) {
            return self::$archive_cache;
        }
        $defaults = [
            '/blog', '/news', '/articles', '/posts',
            '/portfolio', '/gallery',
            '/category/', '/tag/', '/author/',
            '/page/', // paginated listing
            '/المدونة', '/مقالات', '/أخبار', '/الأخبار',
        ];
        if (function_exists('apply_filters')) {
            self::$archive_cache = (array) apply_filters('aetos_seo_archive_substrings', $defaults);
        } else {
            self::$archive_cache = $defaults;
        }
        return self::$archive_cache;
    }

    public static function is_excluded(string $url): bool
    {
        $path = wp_parse_url($url, PHP_URL_PATH);
        if (!is_string($path) || $path === '') {
            $path = '/';
        }
        foreach (self::excluded_patterns() as $pattern) {
            if (is_string($pattern) && $pattern !== '' && @preg_match($pattern, $path)) {
                return true;
            }
        }
        return false;
    }

    public static function is_legitimately_short(string $url): bool
    {
        if (self::is_archive_index($url)) {
            return true;
        }

        $u = Str::lower(rawurldecode($url));
        foreach (self::legitimately_short_substrings() as $sub) {
            if (is_string($sub) && $sub !== '' && strpos($u, $sub) !== false) {
                return true;
            }
        }
        return false;
    }

    public static function is_archive_index(string $url): bool
    {
        $path = wp_parse_url($url, PHP_URL_PATH);
        if (!is_string($path) || $path === '') {
            $path = '/';
        }
        $path = '/' . trim(Str::lower(rawurldecode($path)), '/');

        foreach (self::archive_substrings() as $sub) {
            if (!is_string($sub) || $sub === '') {
                continue;
            }

            $normalized = Str::lower(rawurldecode($sub));
            if (substr($normalized, -1) === '/' && strpos($path, $normalized) === 0) {
                return true;
            }

            $root = '/' . trim($normalized, '/');
            if ($path === $root || preg_match('~^' . preg_quote($root, '~') . '/page/\d+/?$~', $path)) {
                return true;
            }
        }
        return false;
    }

    /**
     * Reset internal caches. Call from tests or when filters change at runtime.
     */
    public static function reset_cache(): void
    {
        self::$excluded_cache = null;
        self::$legit_short_cache = null;
        self::$archive_cache = null;
    }
}