<?php
if (!defined('ABSPATH')) exit;

/**
 * ADL-023: validates AI-generated content immediately before it can be saved.
 */
class AIWEST_Content_Preservation_Gate {
    const VISUAL_ONLY = 'visual_only';
    const CONTENT_REWRITE = 'content_rewrite';
    const CONTENT_EXPANSION = 'content_expansion';

    public static function detect_intent($action, $instruction = '') {
        $action = sanitize_key((string)$action);
        $instruction = self::lower((string)$instruction);

        if ($action === 'rewrite' || ($action === 'custom' && preg_match('/(rewrite|replace|regenerate (?:the )?content|change (?:the )?text|rewrite copy|write anew|from scratch|przepisz|zastąp|wygeneruj nową treść|zmień tekst|napisz od nowa)/u', $instruction))) {
            return self::CONTENT_REWRITE;
        }
        if ($action === 'visual' || ($action === 'custom' && preg_match('/(visual|layout|landing page|design|presentation|look better|readability visually|improve cards|improve sections|cta layout|układ|wizualn|wygląd|prezentac|popraw wizualnie|dodaj karty|ładniejsze sekcje)/u', $instruction))) {
            return self::VISUAL_ONLY;
        }
        return self::CONTENT_EXPANSION;
    }

    public static function capture_snapshot($post_id, $action, $instruction, $intent) {
        $post = get_post(intval($post_id));
        if (!$post) return new WP_Error('aiwest_preservation_no_post', __('Content to protect was not found.', 'ai-webmaster-studio'));

        if ($intent === self::VISUAL_ONLY && class_exists('AIWEST_Undo')) {
            AIWEST_Undo::snapshot($post_id, 'visual_enhancement', __('Automatic backup before preserving-content visual enhancement.', 'ai-webmaster-studio'));
        }

        $protected_content = (string)$post->post_content;
        if ($intent === self::VISUAL_ONLY && class_exists('AIWEST_Visual_Layout_Engine')) {
            $protected_content = AIWEST_Visual_Layout_Engine::unwrap($protected_content);
            if (class_exists('AIWEST_Semantic_Layout_Composer')) {
                $protected_content = AIWEST_Semantic_Layout_Composer::decompose($protected_content);
            }
        }
        $heading_report = [];
        if (class_exists('AIWEST_Page_Title_Heading_Normalizer')) {
            $normalized = AIWEST_Page_Title_Heading_Normalizer::normalize($post_id, $protected_content, [
                'post_title' => (string)$post->post_title,
                'post_type' => (string)$post->post_type,
                'mode' => 'standard',
            ]);
            $protected_content = $normalized['content'];
            $heading_report = $normalized['report'];
        }

        $snapshot = self::snapshot_from_content($protected_content, [
            'post_id' => intval($post_id),
            'post_title' => (string)$post->post_title,
            'post_excerpt' => (string)$post->post_excerpt,
            'post_status' => (string)$post->post_status,
            'post_type' => (string)$post->post_type,
            'post_name' => isset($post->post_name) ? (string)$post->post_name : '',
            'action' => sanitize_key($action),
            'instruction' => sanitize_textarea_field($instruction),
            'intent' => $intent,
            'duplicate_title_heading' => $heading_report,
            'captured_at' => current_time('mysql'),
        ]);

        update_post_meta($post_id, '_aiwest_content_preservation_last_snapshot', $snapshot);
        return $snapshot;
    }

    public static function snapshot_from_content($content, $context = []) {
        $content = (string)$content;
        return array_merge($context, [
            'content' => $content,
            'visible_text' => self::visible_text($content),
            'word_count' => self::word_count($content),
            'headings' => self::headings($content),
            'links' => self::links($content),
            'section_count' => self::section_count($content),
            'critical_phrases' => self::critical_phrases($content),
            'content_hash' => hash('sha256', $content),
            'block_structure_hash' => hash('sha256', self::block_structure($content)),
        ]);
    }

    public static function validate($post_id, $snapshot, $candidate_html, $intent) {
        $post = get_post(intval($post_id));
        $report = self::validate_snapshot($snapshot, $candidate_html, $intent, $post ? [
            'post_id' => intval($post_id),
            'post_title' => (string)$post->post_title,
            'post_name' => isset($post->post_name) ? (string)$post->post_name : '',
        ] : []);

        update_post_meta($post_id, '_aiwest_content_preservation_last_report', $report);
        if (!$report['passed']) {
            update_post_meta($post_id, '_aiwest_content_preservation_review_required', current_time('mysql'));
            if (class_exists('AIWEST_Logger')) {
                AIWEST_Logger::warning('Content Preservation Gate blocked an unsafe AI update.', [
                    'post_id' => intval($post_id),
                    'intent' => $intent,
                    'reasons' => $report['reasons'],
                ]);
            }
            $message = $intent === self::VISUAL_ONLY
                ? __('The visual layout could not be applied safely. No rewrite was attempted, and the original content was preserved.', 'ai-webmaster-studio')
                : __('The AI update was stopped by the Content Preservation Gate. The original content was preserved.', 'ai-webmaster-studio');
            return new WP_Error(
                'aiwest_content_preservation_blocked',
                $message,
                $report
            );
        }
        delete_post_meta($post_id, '_aiwest_content_preservation_review_required');
        return $report;
    }

    public static function validate_snapshot($snapshot, $candidate_html, $intent, $context = []) {
        $candidate_html = (string)$candidate_html;
        $candidate = self::snapshot_from_content($candidate_html, $context);
        $original_word_count = intval($snapshot['word_count'] ?? 0);
        $original_words = max(1, $original_word_count);
        $word_ratio = $original_word_count === 0 ? 1.0 : ($candidate['word_count'] / $original_words);
        $strategic = self::is_strategic($snapshot + $context);
        $minimum_ratio = $strategic ? 0.97 : 0.95;
        $reasons = [];
        $missing_headings = self::missing_values($candidate['headings'], $snapshot['headings'] ?? []);
        $missing_links = self::missing_values($candidate['links'], $snapshot['links'] ?? []);

        if ($candidate_html === '') $reasons[] = 'empty_content';
        if (!self::valid_html_structure($candidate_html)) $reasons[] = 'invalid_html_structure';
        if (!self::valid_block_comments($candidate_html)) $reasons[] = 'invalid_wordpress_blocks';

        if ($intent !== self::CONTENT_REWRITE) {
            if ($word_ratio < $minimum_ratio) $reasons[] = 'word_count_reduced';
            if ($missing_headings) $reasons[] = 'headings_removed_or_changed';
            if ($missing_links) $reasons[] = 'links_removed_or_changed';
            if (!self::phrases_present($candidate['visible_text'], $snapshot['critical_phrases'] ?? [])) $reasons[] = 'critical_text_removed';
            $original_sections = intval($snapshot['section_count'] ?? 0);
            if ($original_sections > 0 && $candidate['section_count'] < max(1, intval(ceil($original_sections * $minimum_ratio)))) $reasons[] = 'sections_reduced';
        }

        if ($intent === self::VISUAL_ONLY && $candidate['visible_text'] !== (string)($snapshot['visible_text'] ?? '')) {
            $reasons[] = 'visible_text_changed';
        }

        $semantic_identity = null;
        if ($intent === self::VISUAL_ONLY && class_exists('AIWEST_Semantic_Layout_Composer')) {
            $original_inventory = AIWEST_Semantic_Layout_Composer::inventory((string)($snapshot['content'] ?? ''));
            $candidate_inventory = AIWEST_Semantic_Layout_Composer::inventory($candidate_html);
            $semantic_identity = ($original_inventory === $candidate_inventory);
            if (!$semantic_identity) $reasons[] = 'semantic_content_identity_changed';
        }

        return [
            'passed' => empty($reasons),
            'intent' => $intent,
            'strategic' => $strategic,
            'minimum_word_ratio' => $minimum_ratio,
            'word_ratio' => round($word_ratio, 4),
            'original_words' => intval($snapshot['word_count'] ?? 0),
            'candidate_words' => $candidate['word_count'],
            'original_sections' => intval($snapshot['section_count'] ?? 0),
            'candidate_sections' => $candidate['section_count'],
            'missing_headings' => $missing_headings,
            'missing_links' => $missing_links,
            'original_block_structure_hash' => (string)($snapshot['block_structure_hash'] ?? ''),
            'candidate_block_structure_hash' => $candidate['block_structure_hash'],
            'semantic_content_identity_preserved' => $semantic_identity,
            'reasons' => array_values(array_unique($reasons)),
            'checked_at' => function_exists('current_time') ? current_time('mysql') : gmdate('Y-m-d H:i:s'),
        ];
    }

    protected static function visible_text($html) {
        $html = preg_replace('/<!--.*?-->/s', ' ', (string)$html);
        $html = preg_replace('/<(script|style|noscript)\b[^>]*>.*?<\/\1>/is', ' ', $html);
        $html = preg_replace('/<(?:\/?(?:address|article|aside|blockquote|div|details|figcaption|figure|footer|form|h[1-6]|header|hr|li|main|nav|ol|p|pre|section|summary|table|tbody|td|tfoot|th|thead|tr|ul)|br)\b[^>]*>/i', ' $0 ', $html);
        $text = html_entity_decode(wp_strip_all_tags($html), ENT_QUOTES | ENT_HTML5, 'UTF-8');
        $text = preg_replace('/\s+/u', ' ', trim($text));
        return self::lower($text);
    }

    protected static function word_count($html) {
        $text = self::visible_text($html);
        if ($text === '') return 0;
        preg_match_all('/[\p{L}\p{N}][\p{L}\p{N}\p{M}\x{2019}\'\-]*/u', $text, $matches);
        return count($matches[0]);
    }

    protected static function headings($html) {
        preg_match_all('/<h([1-3])\b[^>]*>(.*?)<\/h\1>/is', (string)$html, $matches, PREG_SET_ORDER);
        $headings = [];
        foreach ($matches as $match) {
            $headings[] = 'h' . $match[1] . ':' . self::visible_text($match[2]);
        }
        return $headings;
    }

    protected static function links($html) {
        preg_match_all('/<a\b[^>]*\bhref\s*=\s*(["\'])(.*?)\1/is', (string)$html, $matches);
        return array_values(array_unique(array_map('html_entity_decode', $matches[2] ?? [])));
    }

    protected static function section_count($html) {
        preg_match_all('/<(?:section|article|h[1-3])\b/i', (string)$html, $matches);
        return count($matches[0]);
    }

    protected static function critical_phrases($html) {
        $text = self::visible_text($html);
        preg_match_all('/[^.!?\n]{35,}[.!?]/u', $text, $matches);
        $phrases = array_slice($matches[0] ?? [], 0, 24);
        $protected_terms = [
            'ai webmaster is a free wordpress plugin',
            'aiwos cloud is a separate future platform',
            'aiwos cloud is not a paid unlock of ai webmaster',
            'administrator-supervised workflows',
            'api keys', 'external services', 'security & privacy',
            'approval center', 'rollback', 'monitoring', 'scheduled reports',
            'limitation of liability', 'user responsibility',
        ];
        foreach ($protected_terms as $term) {
            if (strpos($text, $term) !== false) $phrases[] = $term;
        }
        if (!$phrases && $text !== '') $phrases[] = self::substr($text, 0, 180);
        return array_values(array_unique(array_filter(array_map('trim', $phrases))));
    }

    protected static function missing_values($candidate, $required) {
        $counts = array_count_values($candidate);
        $missing = [];
        foreach ($required as $item) {
            if (empty($counts[$item])) {
                $missing[] = $item;
            } else {
                $counts[$item]--;
            }
        }
        return $missing;
    }

    protected static function phrases_present($candidate_text, $phrases) {
        foreach ($phrases as $phrase) {
            if ($phrase !== '' && strpos($candidate_text, self::lower($phrase)) === false) return false;
        }
        return true;
    }

    protected static function valid_html_structure($html) {
        $stack = [];
        $void = ['area','base','br','col','embed','hr','img','input','link','meta','param','source','track','wbr'];
        preg_match_all('/<\/?([a-z][a-z0-9:-]*)\b[^>]*>/i', (string)$html, $matches, PREG_SET_ORDER);
        foreach ($matches as $match) {
            $tag = strtolower($match[1]);
            $token = $match[0];
            if (in_array($tag, $void, true) || preg_match('/\/\s*>$/', $token)) continue;
            if (strpos($token, '</') === 0) {
                if (!$stack || array_pop($stack) !== $tag) return false;
            } else {
                if ($tag === 'a' && in_array('a', $stack, true)) return false;
                if ($tag === 'p' && in_array('p', $stack, true)) return false;
                if (in_array($tag, ['div','section','article','h1','h2','h3','ul','ol','table'], true) && end($stack) === 'p') return false;
                $stack[] = $tag;
            }
        }
        return empty($stack);
    }

    protected static function valid_block_comments($html) {
        preg_match_all('/<!--\s*(\/?)wp:([a-z0-9_-]+(?:\/[a-z0-9_-]+)?)(?:\s+[^>]*?)?\s*(\/?)-->/i', (string)$html, $matches, PREG_SET_ORDER);
        $stack = [];
        foreach ($matches as $match) {
            if ($match[1] === '/') {
                if (!$stack || array_pop($stack) !== strtolower($match[2])) return false;
            } elseif ($match[3] !== '/') {
                $stack[] = strtolower($match[2]);
            }
        }
        return empty($stack);
    }

    protected static function is_strategic($context) {
        $haystack = self::lower(($context['post_title'] ?? '') . ' ' . ($context['post_name'] ?? ''));
        $terms = [
            'home','homepage','start','o nas','about','kontakt','contact','polityka','policy','privacy','regulamin','terms','security','bezpieczeń',
            'aiwos cloud','ai webmaster','api keys','openai api','external services','changelog','roadmap','faq','support','documentation','dokumentacja','operator','approval center','ai webmaster vs aiwos',
        ];
        foreach ($terms as $term) {
            if (strpos($haystack, $term) !== false) return true;
        }
        if (!empty($context['post_id']) && function_exists('get_option') && intval(get_option('wp_page_for_privacy_policy', 0)) === intval($context['post_id'])) return true;
        return false;
    }

    protected static function block_structure($html) {
        preg_match_all('/<!--\s*\/?wp:[a-z0-9_-]+(?:\/[a-z0-9_-]+)?[^>]*-->|<\/?(?:section|article|div|p|h[1-6]|ul|ol|li|table|thead|tbody|tr|th|td|a)\b[^>]*>/i', (string)$html, $matches);
        return implode('|', array_map(function($token) {
            return preg_replace('/\s+/u', ' ', trim(self::lower($token)));
        }, $matches[0] ?? []));
    }

    protected static function lower($text) {
        return function_exists('mb_strtolower') ? mb_strtolower((string)$text, 'UTF-8') : strtolower((string)$text);
    }

    protected static function substr($text, $start, $length) {
        return function_exists('mb_substr') ? mb_substr($text, $start, $length) : substr($text, $start, $length);
    }
}