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

/** Content adapters. Never reserialize arbitrary HTML or edit shared/dynamic blocks. */
class Linkbraid_Content {
    public static function editor($id) {
        if (get_post_meta($id, '_elementor_edit_mode', true) === 'builder' || get_post_meta($id, '_elementor_data', true)) return 'elementor';
        if (get_post_meta($id, '_fl_builder_data', true)) return 'beaver';
        $post = get_post($id);
        if ($post && get_post_meta($id, '_et_pb_use_builder', true) === 'on' && strpos($post->post_content, '[et_pb_') !== false) return 'divi';
        if ($post && strpos($post->post_content, '[vc_') !== false) return 'wpbakery';
        return $post && has_blocks($post->post_content) ? 'gutenberg' : 'classic';
    }

    public static function snapshot($id) {
        $post = get_post($id);
        return ['content' => $post ? $post->post_content : '', 'elementor' => get_post_meta($id, '_elementor_data', true)];
    }

    public static function hash($id) { return hash('sha256', wp_json_encode(self::snapshot($id))); }

    public static function fragments($id) {
        $post = get_post($id);
        if (!$post) return [];
        $editor = self::editor($id);
        $out = [];
        if ($editor === 'elementor') {
            $data = get_post_meta($id, '_elementor_data', true);
            $data = is_string($data) ? json_decode($data, true) : $data;
            if (is_array($data)) self::elementor_fragments($data, [], $out);
        } elseif ($editor === 'gutenberg') {
            self::block_fragments(parse_blocks($post->post_content), [], $out);
        } elseif ($editor === 'divi') {
            self::divi_fragments($post->post_content, $out);
        } elseif ($editor === 'classic') {
            $out[] = ['location' => __( 'Post content', 'linkbraid' ), 'path' => [], 'html' => $post->post_content];
        }
        // Unknown builder sources are intentionally read-only until a dedicated adapter exists.
        return $out;
    }

    /**
     * Divi stores the builder as shortcodes in post_content. The text module
     * ([et_pb_text]) has a known HTML contract; its inner content is addressed
     * byte-precisely (offset + length), so the write is a plain substr_replace
     * and every other byte of the layout stays untouched. Modules that contain
     * nested builder shortcodes stay read-only.
     */
    private static function divi_fragments($content, &$out) {
        if (!function_exists('get_shortcode_regex')) return;
        if (!preg_match_all('/' . get_shortcode_regex(['et_pb_text']) . '/s', $content, $matches, PREG_OFFSET_CAPTURE)) return;
        foreach ($matches[5] as $i => $inner) {
            if (($matches[1][$i][0] ?? '') === '[') continue; // escaped [[et_pb_text]]
            if (!is_array($inner) || $inner[1] < 0) continue; // self-closing module
            $html = $inner[0];
            if (trim($html) === '' || strpos($html, '[et_pb_') !== false) continue;
            /* translators: %d: dynamic value inserted at runtime. */
            $out[] = ['path' => ['divi', $inner[1], strlen($html)], 'location' => sprintf( __( 'Divi · text module %d', 'linkbraid' ), $i + 1 ), 'html' => $html];
        }
    }

    private static function block_fragments($blocks, $parent, &$out) {
        foreach ($blocks as $i => $block) {
            $path = array_merge($parent, [$i]);
            if (!empty($block['attrs']['metadata']['bindings'])) continue;
            if (in_array($block['blockName'], ['core/paragraph', 'core/list-item', 'core/list'], true) && empty($block['innerBlocks'])) {
                /* translators: %1$s, %2$s: dynamic values in display order. */
                $out[] = ['path' => $path, 'location' => sprintf( __( 'Block %1$s · %2$s', 'linkbraid' ), implode('.', $path), $block['blockName'] ), 'html' => $block['innerHTML']];
            }
            if (in_array($block['blockName'], ['core/group', 'core/columns', 'core/column', 'core/list'], true)) {
                self::block_fragments($block['innerBlocks'], array_merge($path, ['innerBlocks']), $out);
            }
        }
    }

    private static function elementor_fragments($elements, $parent, &$out) {
        foreach ($elements as $i => $el) {
            $path = array_merge($parent, [$i]);
            // Rich-text widgets have a known HTML contract. Titles, HTML/code widgets and menus do not.
            if (($el['widgetType'] ?? '') === 'text-editor' && isset($el['settings']['editor']) && is_string($el['settings']['editor'])) {
                /* translators: %s: dynamic value inserted at runtime. */
                $out[] = ['path' => array_merge($path, ['settings', 'editor']), 'location' => sprintf( __( 'Elementor · %s', 'linkbraid' ), $el['id'] ?? implode('.', $path) ), 'html' => $el['settings']['editor']];
            }
            if (!empty($el['elements']) && is_array($el['elements'])) self::elementor_fragments($el['elements'], array_merge($path, ['elements']), $out);
        }
    }

    /** Tokens preserve exact bytes, including attributes, JSON comments and entities. */
    public static function tokens($html, $include_links = false) {
        preg_match_all('~<!--[\s\S]*?(?:-->|$)|<(script|style|textarea|pre|code|template|noscript)\b[^>]*>[\s\S]*?(?:</\1\s*>|$)|<(?:(?:"[^"]*"|\x27[^\x27]*\x27)|[^\x27">])*>|[^<]+|<~i', $html, $matches, PREG_OFFSET_CAPTURE);
        $out = []; $stack = [];
        foreach ($matches[0] as $match) {
            list($raw, $offset) = $match;
            if ($raw[0] === '<') {
                if (preg_match('~^</([a-z0-9]+)\s*>$~i', $raw, $m)) {
                    $tag = strtolower($m[1]);
                    for ($i = count($stack) - 1; $i >= 0; $i--) if ($stack[$i]['tag'] === $tag) { $stack = array_slice($stack, 0, $i); break; }
                } elseif (preg_match('~^<([a-z0-9]+)\b~i', $raw, $m) && !preg_match('~^<(script|style|textarea|pre|code|template|noscript)\b~i', $raw)) {
                    $tag = strtolower($m[1]);
                    $blocked = ($tag === 'a' && !$include_links) || in_array($tag, ['h1','h2','h3','h4','h5','h6','nav','header','footer','button','select','svg','math'], true)
                        || preg_match('~\s(?:hidden|inert)(?:\s|=|>)|aria-hidden\s*=\s*["\x27]true|display\s*:\s*none|visibility\s*:\s*hidden~i', $raw);
                    if (!in_array($tag, ['br','hr','img','input','meta','link','source','wbr','area','base','embed','param','track','col'], true) && substr($raw, -2) !== '/>') $stack[] = ['tag' => $tag, 'blocked' => $blocked];
                }
                continue;
            }
            $blocked = false;
            foreach ($stack as $entry) if ($entry['blocked']) $blocked = true;
            if (!$blocked && trim($raw) !== '') {
                // Shortcode attributes are code, even though they are not enclosed in HTML tags.
                $parts = preg_split('/(\[(?:\/?)[a-zA-Z_][^\]]*\])/', $raw, -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_OFFSET_CAPTURE);
                foreach ($parts as $part) {
                    if (trim($part[0]) === '' || preg_match('/^\[\/?[a-zA-Z_]/', $part[0])) continue;
                    $out[] = ['raw' => $part[0], 'offset' => $offset + $part[1], 'text' => html_entity_decode($part[0], ENT_QUOTES | ENT_HTML5, 'UTF-8')];
                }
            }
        }
        return $out;
    }

    public static function plain($html, $include_links = false) {
        // A newly inserted anchor must not add whitespace before punctuation in
        // the context used to locate the remaining suggestions from this analysis.
        $text = ''; $previous = null;
        foreach (self::tokens($html, $include_links) as $token) {
            $separator = ' ';
            if ($include_links && $previous !== null) {
                $end = $previous['offset'] + strlen($previous['raw']);
                $gap = substr($html, $end, $token['offset'] - $end);
                $between = preg_replace('~</?a\b(?:(?:"[^"]*"|\x27[^\x27]*\x27)|[^\x27">])*>~i', '', $gap);
                if ($between !== $gap && trim($between) === '') $separator = $between;
            }
            $text .= ($previous === null ? '' : $separator) . $token['text'];
            $previous = $token;
        }
        return trim(preg_replace('/[\s\x{00a0}]+/u', ' ', $text));
    }

    public static function context_matches($html, $context) {
        if (mb_strpos(self::plain($html, true), $context) !== false) return true;
        // Keep proposals generated before the context whitespace correction usable.
        $legacy = trim(preg_replace('/[\s\x{00a0}]+/u', ' ', implode(' ', array_column(self::tokens($html, true), 'text'))));
        return mb_strpos($legacy, $context) !== false;
    }

    public static function append_sentence($html, $sentence) {
        $tokens = self::tokens($html);
        if (!$tokens) return null;
        $last = end($tokens); $end = $last['offset'] + strlen($last['raw']);
        // Find the containing paragraph/list-item boundary, skipping comments, raw code and quoted attributes.
        preg_match_all('~<!--[\s\S]*?(?:-->|$)|<(script|style|textarea|pre|code|template|noscript)\b[^>]*>[\s\S]*?(?:</\1\s*>|$)|<(?:(?:"[^"]*"|\x27[^\x27]*\x27)|[^\x27">])*>~i', substr($html, $end), $matches, PREG_OFFSET_CAPTURE);
        foreach ($matches[0] as $m) {
            if (preg_match('~^</(?:p|li)\s*>$~i', $m[0])) {
                $offset = $end + $m[1];
                return substr($html, 0, $offset) . ' ' . $sentence . substr($html, $offset);
            }
            if (preg_match('~^<(?:p|li|nav|header|footer)\b~i', $m[0])) return null;
        }
        return $end === strlen($html) ? $html . ' ' . $sentence : null;
    }

    public static function text($id) {
        return implode("\n\n", array_map(function($f) { return self::plain($f['html']); }, self::fragments($id)));
    }

    public static function insert($html, $anchor, $url) {
        if (trim($anchor) === '') return null;
        $pattern = '~(?<![\p{L}\p{N}_])' . implode('[\s\x{00a0}]+', array_map(function($s) { return preg_quote($s, '~'); }, preg_split('/\s+/u', trim($anchor)))) . '(?![\p{L}\p{N}_])~iu';
        $match = self::insert_match($html, $pattern, $url);
        return $match ? $match['html'] : null;
    }

    /**
     * Regex for a phrase that also matches natural inflected forms: words of
     * 5-6 letters vary in the last letter, 7+ letters in the last two (plus up
     * to a few extra trailing letters), short words stay literal. Conservative
     * on purpose — precision beats recall for automatic rules.
     */
    public static function flexible_pattern($phrase) {
        $words = preg_split('/\s+/u', trim((string) $phrase), -1, PREG_SPLIT_NO_EMPTY);
        if (!$words) return null;
        $parts = [];
        foreach ($words as $word) {
            $length = mb_strlen($word);
            if ($length <= 4) { $parts[] = preg_quote($word, '~'); continue; }
            if ($length <= 6) { $parts[] = preg_quote(mb_substr($word, 0, $length - 1), '~') . '\p{L}{0,3}'; continue; }
            $parts[] = preg_quote(mb_substr($word, 0, $length - 2), '~') . '\p{L}{0,4}';
        }
        return '~(?<![\p{L}\p{N}_])' . implode('[\s\x{00a0}]+', $parts) . '(?![\p{L}\p{N}_])~iu';
    }

    /**
     * Like insert(), but the phrase may match inflected forms. Returns the new
     * HTML together with the SURFACE form found in the text — that form becomes
     * the anchor, so everything downstream stays a literal match.
     */
    public static function insert_flexible($html, $phrase, $url) {
        $pattern = self::flexible_pattern($phrase);
        return $pattern ? self::insert_match($html, $pattern, $url) : null;
    }

    private static function insert_match($html, $pattern, $url) {
        foreach (self::tokens($html) as $token) {
            if (!preg_match($pattern, $token['text'], $m, PREG_OFFSET_CAPTURE)) continue;
            // Map decoded byte boundaries to original HTML, preserving every original entity.
            $raw = $token['raw']; $map = [0 => 0]; $decoded = 0;
            preg_match_all('/&(?:#[0-9]+|#x[0-9a-f]+|[a-z][a-z0-9]+);|./isu', $raw, $units, PREG_OFFSET_CAPTURE);
            foreach ($units[0] as $unit) { $decoded += strlen(html_entity_decode($unit[0], ENT_QUOTES | ENT_HTML5, 'UTF-8')); $map[$decoded] = $unit[1] + strlen($unit[0]); }
            $start = $m[0][1]; $end = $start + strlen($m[0][0]);
            if (!isset($map[$start], $map[$end])) continue;
            $a = $token['offset'] + $map[$start]; $b = $token['offset'] + $map[$end];
            return [
                'html'   => substr($html, 0, $a) . '<a href="' . esc_url($url) . '">' . substr($html, $a, $b - $a) . '</a>' . substr($html, $b),
                'anchor' => trim(preg_replace('/[\s\x{00a0}]+/u', ' ', $m[0][0])),
            ];
        }
        return null;
    }

    /**
     * Removes the first <a> pointing at $url, keeping its inner content.
     * String surgery on the original HTML (like insert()) — the DOM is never
     * serialized back into a post. Returns the modified HTML or null.
     */
    public static function unlink($html, $url) {
        $needle = untrailingslashit(strtolower(trim(html_entity_decode((string) $url, ENT_QUOTES | ENT_HTML5, 'UTF-8'))));
        if ($needle === '') return null;
        if (!preg_match_all('~<a\b((?:(?:"[^"]*"|\x27[^\x27]*\x27)|[^\x27">])*)>([\s\S]*?)</a\s*>~i', $html, $matches, PREG_OFFSET_CAPTURE)) return null;
        foreach ($matches[0] as $i => $full) {
            if (!preg_match('~\bhref\s*=\s*(?:"([^"]*)"|\x27([^\x27]*)\x27|([^\s">]+))~i', $matches[1][$i][0], $m)) continue;
            $href = $m[1] !== '' ? $m[1] : (isset($m[2]) && $m[2] !== '' ? $m[2] : ($m[3] ?? ''));
            $href = untrailingslashit(strtolower(trim(html_entity_decode($href, ENT_QUOTES | ENT_HTML5, 'UTF-8'))));
            if ($href !== $needle) continue;
            $offset = $full[1];
            return substr($html, 0, $offset) . $matches[2][$i][0] . substr($html, $offset + strlen($full[0]));
        }
        return null;
    }

    /**
     * Replaces the FIRST href equal to $old with $new (case- and trailing-slash-
     * insensitive comparison, entity-decoded). Only the attribute value changes;
     * anchor text, other attributes and markup stay byte-identical. Returns the
     * modified HTML or null when no link matches.
     */
    public static function replace_url($html, $old, $new) {
        $needle = untrailingslashit(strtolower(trim(html_entity_decode((string) $old, ENT_QUOTES | ENT_HTML5, 'UTF-8'))));
        $new = trim((string) $new);
        if ($needle === '' || $new === '') return null;
        if (!preg_match_all('~<a\b((?:(?:"[^"]*"|\x27[^\x27]*\x27)|[^\x27">])*)>~i', $html, $matches, PREG_OFFSET_CAPTURE)) return null;
        foreach ($matches[1] as $attrs) {
            if (!preg_match('~\bhref\s*=\s*(?:"([^"]*)"|\x27([^\x27]*)\x27|([^\s">]+))~i', $attrs[0], $m, PREG_OFFSET_CAPTURE)) continue;
            $value = null;
            foreach ([1, 2, 3] as $group) {
                if (isset($m[$group]) && $m[$group][1] !== -1) { $value = $m[$group]; break; }
            }
            if ($value === null) continue;
            $href = untrailingslashit(strtolower(trim(html_entity_decode($value[0], ENT_QUOTES | ENT_HTML5, 'UTF-8'))));
            if ($href !== $needle) continue;
            $start = $attrs[1] + $value[1];
            return substr($html, 0, $start) . esc_url($new) . substr($html, $start + strlen($value[0]));
        }
        return null;
    }

    public static function resolve_url($href) {
        $href = trim(html_entity_decode($href, ENT_QUOTES | ENT_HTML5, 'UTF-8'));
        if ($href === '' || $href[0] === '#') return false;
        $home = wp_parse_url(home_url('/'));
        if (strpos($href, '//') === 0) $href = $home['scheme'] . ':' . $href;
        elseif ($href[0] === '/') $href = $home['scheme'] . '://' . $home['host'] . (isset($home['port']) ? ':' . $home['port'] : '') . $href;
        $parts = wp_parse_url($href);
        if (!$parts || !in_array(strtolower($parts['scheme'] ?? ''), ['http','https'], true) || strtolower($parts['host'] ?? '') !== strtolower($home['host']) || isset($parts['user']) || isset($parts['pass']) || ($parts['port'] ?? null) !== ($home['port'] ?? null)) return false;
        return esc_url_raw($href);
    }

    public static function links($html) {
        $out = [];
        foreach (self::links_all($html) as $link) if ($link['internal']) $out[] = $link;
        return $out;
    }

    /** All visible links of a fragment: internal (normalized URL) and external http(s). */
    public static function links_all($html) {
        // DOM is used only for reading; never serialize a DOM tree back into a post.
        if (!class_exists('DOMDocument')) return [];
        $old = libxml_use_internal_errors(true);
        $dom = new DOMDocument();
        $dom->loadHTML('<?xml encoding="UTF-8"><div>' . $html . '</div>', LIBXML_NONET | LIBXML_NOERROR | LIBXML_NOWARNING);
        libxml_clear_errors(); libxml_use_internal_errors($old);
        $home_host = strtolower(wp_parse_url(home_url('/'))['host'] ?? '');
        $out = [];
        foreach ($dom->getElementsByTagName('a') as $a) {
            $skip = false;
            for ($p = $a; $p instanceof DOMElement; $p = $p->parentNode) {
                if (in_array(strtolower($p->tagName), ['script','style','template','noscript','code','pre','textarea','nav','header','footer','svg'], true) || $p->hasAttribute('hidden') || $p->hasAttribute('inert') || $p->getAttribute('aria-hidden') === 'true' || preg_match('/display\s*:\s*none|visibility\s*:\s*hidden/i', $p->getAttribute('style'))) $skip = true;
            }
            if ($skip || trim($a->textContent) === '') continue;
            $anchor = trim($a->textContent);
            $internal = self::resolve_url($a->getAttribute('href'));
            if ($internal) { $out[] = ['url' => $internal, 'anchor' => $anchor, 'internal' => true, 'host' => $home_host]; continue; }
            $href = trim(html_entity_decode($a->getAttribute('href'), ENT_QUOTES | ENT_HTML5, 'UTF-8'));
            $parts = $href === '' ? false : wp_parse_url($href);
            if (!$parts || !in_array(strtolower($parts['scheme'] ?? ''), ['http', 'https'], true) || empty($parts['host']) || isset($parts['user']) || isset($parts['pass'])) continue;
            if (strtolower($parts['host']) === $home_host) continue; // internal, but rejected by resolve_url (a different port, for example)
            $out[] = ['url' => esc_url_raw($href), 'anchor' => $anchor, 'internal' => false, 'host' => strtolower($parts['host'])];
        }
        return $out;
    }

    public static function &at(&$data, $path) {
        $ref =& $data;
        foreach ($path as $key) $ref =& $ref[$key];
        return $ref;
    }

    /**
     * ACF text carriers of a post (text, textarea, wysiwyg — also inside
     * repeaters, groups and flexible content), read-only. They feed the link
     * scan and the semantic index; suggestions are still written only into the
     * main editor content.
     */
    public static function acf_fragments($post_id) {
        if (!function_exists('get_field_objects') || !get_option('lbraid_scan_acf', 1)) return [];
        $objects = get_field_objects($post_id);
        if (!is_array($objects) || !$objects) return [];
        $out = [];
        foreach ($objects as $field) self::acf_collect((array) $field, $field['value'] ?? null, $out);
        return apply_filters('lbraid_acf_fragments', $out, $post_id);
    }

    /** Plain text of the ACF carriers, capped — for the embedding context. */
    public static function acf_text($post_id, $max = 4000) {
        $parts = [];
        foreach (self::acf_fragments($post_id) as $fragment) {
            $parts[] = trim(wp_strip_all_tags((string) $fragment['html']));
        }
        return mb_substr(trim(preg_replace('/\s+/u', ' ', implode(' ', array_filter($parts)))), 0, $max);
    }

    public static function acf_collect(array $field, $value, array &$out) {
        $type = (string) ($field['type'] ?? '');
        if (in_array($type, ['text', 'textarea', 'wysiwyg'], true)) {
            if (is_string($value) && trim($value) !== '') {
                $out[] = ['html' => $value, 'location' => 'ACF: ' . (string) ($field['label'] ?? $field['name'] ?? 'field')];
            }
            return;
        }
        if (!in_array($type, ['repeater', 'group', 'flexible_content'], true) || !is_array($value)) return;
        $subs = [];
        foreach ((array) ($field['sub_fields'] ?? []) as $sub) $subs[$sub['name']] = $sub;
        if ($type === 'flexible_content') {
            foreach ((array) ($field['layouts'] ?? []) as $layout) {
                foreach ((array) ($layout['sub_fields'] ?? []) as $sub) $subs[$sub['name']] = $sub;
            }
        }
        $rows = $type === 'group' ? [$value] : $value;
        foreach ($rows as $row) {
            if (!is_array($row)) continue;
            foreach ($row as $name => $sub_value) {
                if (isset($subs[$name])) self::acf_collect((array) $subs[$name], $sub_value, $out);
            }
        }
    }
}