| 1 |
<?php |
| 2 |
// If this file is called directly, abort. |
| 3 |
if (!defined('ABSPATH')) { |
| 4 |
exit; |
| 5 |
} |
| 6 |
|
| 7 |
/** |
| 8 |
* HTML to Markdown converter utility. |
| 9 |
* |
| 10 |
* Reusable utility class for converting WordPress/HTML content into |
| 11 |
* clean markdown. Used by the LLMs.txt generator and the |
| 12 |
* wordpress_get_post_markdown MCP tool. |
| 13 |
* |
| 14 |
* @package Metasync |
| 15 |
* @subpackage Metasync/includes |
| 16 |
*/ |
| 17 |
class Metasync_Html_To_Markdown |
| 18 |
{ |
| 19 |
/** |
| 20 |
* Convert an HTML string to markdown. |
| 21 |
* |
| 22 |
* @param string $html Raw HTML. |
| 23 |
* @return string Markdown output. |
| 24 |
*/ |
| 25 |
public static function convert($html) |
| 26 |
{ |
| 27 |
if (!is_string($html) || $html === '') { |
| 28 |
return ''; |
| 29 |
} |
| 30 |
|
| 31 |
// Normalise whitespace and strip dangerous / noisy elements first. |
| 32 |
$html = preg_replace('/<!--(.|\s)*?-->/', '', $html); |
| 33 |
$html = preg_replace('#<script\b[^>]*>(.*?)</script>#is', '', $html); |
| 34 |
$html = preg_replace('#<style\b[^>]*>(.*?)</style>#is', '', $html); |
| 35 |
$html = preg_replace('#<noscript\b[^>]*>(.*?)</noscript>#is', '', $html); |
| 36 |
|
| 37 |
// Convert YouTube/Vimeo iframes to markdown links (must run before generic iframe strip). |
| 38 |
$html = preg_replace_callback( |
| 39 |
'#<iframe\b[^>]*src=["\']([^"\']+)["\'][^>]*(?:title=["\']([^"\']*)["\'])?[^>]*>.*?</iframe>#is', |
| 40 |
function ($m) { |
| 41 |
$src = $m[1]; |
| 42 |
$title = isset($m[2]) && $m[2] !== '' ? $m[2] : 'Video'; |
| 43 |
if (preg_match('#(youtube\.com|youtu\.be|vimeo\.com)#i', $src)) { |
| 44 |
return "\n[" . $title . '](' . $src . ")\n"; |
| 45 |
} |
| 46 |
return ''; |
| 47 |
}, |
| 48 |
$html |
| 49 |
); |
| 50 |
|
| 51 |
// Convert <br> to newline and <hr> to ---. |
| 52 |
$html = preg_replace('#<br\s*/?>#i', "\n", $html); |
| 53 |
$html = preg_replace('#<hr\s*/?>#i', "\n\n---\n\n", $html); |
| 54 |
|
| 55 |
// Headings h1-h6. |
| 56 |
for ($i = 1; $i <= 6; $i++) { |
| 57 |
$hashes = str_repeat('#', $i); |
| 58 |
$html = preg_replace_callback( |
| 59 |
'#<h' . $i . '\b[^>]*>(.*?)</h' . $i . '>#is', |
| 60 |
function ($m) use ($hashes) { |
| 61 |
$text = self::strip_tags_preserve(trim($m[1])); |
| 62 |
return "\n\n" . $hashes . ' ' . $text . "\n\n"; |
| 63 |
}, |
| 64 |
$html |
| 65 |
); |
| 66 |
} |
| 67 |
|
| 68 |
// Bold. |
| 69 |
$html = preg_replace_callback( |
| 70 |
'#<(?:strong|b)\b[^>]*>(.*?)</(?:strong|b)>#is', |
| 71 |
function ($m) { return '**' . trim(self::strip_tags_preserve($m[1])) . '**'; }, |
| 72 |
$html |
| 73 |
); |
| 74 |
|
| 75 |
// Italic. |
| 76 |
$html = preg_replace_callback( |
| 77 |
'#<(?:em|i)\b[^>]*>(.*?)</(?:em|i)>#is', |
| 78 |
function ($m) { return '*' . trim(self::strip_tags_preserve($m[1])) . '*'; }, |
| 79 |
$html |
| 80 |
); |
| 81 |
|
| 82 |
// Images -> . |
| 83 |
$html = preg_replace_callback( |
| 84 |
'#<img\b([^>]*)/?>#i', |
| 85 |
function ($m) { |
| 86 |
$attrs = $m[1]; |
| 87 |
$alt = ''; |
| 88 |
$src = ''; |
| 89 |
if (preg_match('#\balt=["\']([^"\']*)["\']#i', $attrs, $a)) { |
| 90 |
$alt = $a[1]; |
| 91 |
} |
| 92 |
if (preg_match('#\bsrc=["\']([^"\']+)["\']#i', $attrs, $s)) { |
| 93 |
$src = $s[1]; |
| 94 |
} |
| 95 |
if ($src === '') { |
| 96 |
return ''; |
| 97 |
} |
| 98 |
return ''; |
| 99 |
}, |
| 100 |
$html |
| 101 |
); |
| 102 |
|
| 103 |
// Links -> [text](url). |
| 104 |
$html = preg_replace_callback( |
| 105 |
'#<a\b[^>]*href=["\']([^"\']+)["\'][^>]*>(.*?)</a>#is', |
| 106 |
function ($m) { |
| 107 |
$href = $m[1]; |
| 108 |
$text = trim(self::strip_tags_preserve($m[2])); |
| 109 |
if ($text === '') { |
| 110 |
$text = $href; |
| 111 |
} |
| 112 |
return '[' . $text . '](' . $href . ')'; |
| 113 |
}, |
| 114 |
$html |
| 115 |
); |
| 116 |
|
| 117 |
// Fenced code blocks. |
| 118 |
$html = preg_replace_callback( |
| 119 |
'#<pre\b[^>]*>\s*<code\b[^>]*>(.*?)</code>\s*</pre>#is', |
| 120 |
function ($m) { |
| 121 |
$code = html_entity_decode($m[1], ENT_QUOTES | ENT_HTML5, 'UTF-8'); |
| 122 |
return "\n\n```\n" . trim($code) . "\n```\n\n"; |
| 123 |
}, |
| 124 |
$html |
| 125 |
); |
| 126 |
$html = preg_replace_callback( |
| 127 |
'#<pre\b[^>]*>(.*?)</pre>#is', |
| 128 |
function ($m) { |
| 129 |
$code = html_entity_decode(strip_tags($m[1]), ENT_QUOTES | ENT_HTML5, 'UTF-8'); |
| 130 |
return "\n\n```\n" . trim($code) . "\n```\n\n"; |
| 131 |
}, |
| 132 |
$html |
| 133 |
); |
| 134 |
|
| 135 |
// Inline code. |
| 136 |
$html = preg_replace_callback( |
| 137 |
'#<code\b[^>]*>(.*?)</code>#is', |
| 138 |
function ($m) { |
| 139 |
$code = html_entity_decode(strip_tags($m[1]), ENT_QUOTES | ENT_HTML5, 'UTF-8'); |
| 140 |
return '`' . $code . '`'; |
| 141 |
}, |
| 142 |
$html |
| 143 |
); |
| 144 |
|
| 145 |
// Blockquotes. |
| 146 |
$html = preg_replace_callback( |
| 147 |
'#<blockquote\b[^>]*>(.*?)</blockquote>#is', |
| 148 |
function ($m) { |
| 149 |
$inner = trim(self::strip_tags_preserve($m[1])); |
| 150 |
$lines = preg_split('/\r?\n/', $inner); |
| 151 |
$lines = array_map(function ($l) { return '> ' . ltrim($l); }, $lines); |
| 152 |
return "\n\n" . implode("\n", $lines) . "\n\n"; |
| 153 |
}, |
| 154 |
$html |
| 155 |
); |
| 156 |
|
| 157 |
// Tables. |
| 158 |
$html = preg_replace_callback( |
| 159 |
'#<table\b[^>]*>(.*?)</table>#is', |
| 160 |
[self::class, 'convert_table'], |
| 161 |
$html |
| 162 |
); |
| 163 |
|
| 164 |
// Lists (nested supported via recursion). |
| 165 |
$html = self::convert_lists($html); |
| 166 |
|
| 167 |
// Paragraphs. |
| 168 |
$html = preg_replace_callback( |
| 169 |
'#<p\b[^>]*>(.*?)</p>#is', |
| 170 |
function ($m) { |
| 171 |
$text = trim(self::strip_tags_preserve($m[1])); |
| 172 |
if ($text === '') { |
| 173 |
return ''; |
| 174 |
} |
| 175 |
return "\n\n" . $text . "\n\n"; |
| 176 |
}, |
| 177 |
$html |
| 178 |
); |
| 179 |
|
| 180 |
// Strip any remaining tags but keep the inner text. |
| 181 |
$html = self::strip_tags_preserve($html); |
| 182 |
|
| 183 |
// Decode entities after stripping tags. |
| 184 |
$html = html_entity_decode($html, ENT_QUOTES | ENT_HTML5, 'UTF-8'); |
| 185 |
|
| 186 |
// Collapse excessive blank lines. |
| 187 |
$html = preg_replace("/\n{3,}/", "\n\n", $html); |
| 188 |
|
| 189 |
return trim($html); |
| 190 |
} |
| 191 |
|
| 192 |
/** |
| 193 |
* Convert a WordPress post to markdown. |
| 194 |
* |
| 195 |
* Runs `the_content` filter and do_shortcode() first so page-builder |
| 196 |
* content (Elementor, Divi, Beaver Builder, etc.) is rendered before |
| 197 |
* conversion. |
| 198 |
* |
| 199 |
* @param int $post_id Post ID. |
| 200 |
* @param array $options { |
| 201 |
* Optional. { |
| 202 |
* @type bool $include_frontmatter Prepend YAML frontmatter. Default true. |
| 203 |
* @type bool $include_featured_image Prepend featured image as markdown. Default true. |
| 204 |
* } |
| 205 |
* } |
| 206 |
* @return string Markdown output (empty string if post not found). |
| 207 |
*/ |
| 208 |
public static function convert_post($post_id, $options = []) |
| 209 |
{ |
| 210 |
$post = get_post(absint($post_id)); |
| 211 |
if (!$post) { |
| 212 |
return ''; |
| 213 |
} |
| 214 |
|
| 215 |
$options = wp_parse_args($options, [ |
| 216 |
'include_frontmatter' => true, |
| 217 |
'include_featured_image' => true, |
| 218 |
]); |
| 219 |
|
| 220 |
$raw_content = $post->post_content; |
| 221 |
$filtered = apply_filters('the_content', $raw_content); |
| 222 |
$filtered = do_shortcode($filtered); |
| 223 |
|
| 224 |
$markdown = self::convert($filtered); |
| 225 |
|
| 226 |
$prefix = ''; |
| 227 |
|
| 228 |
if (!empty($options['include_frontmatter'])) { |
| 229 |
$author_name = ''; |
| 230 |
if (!empty($post->post_author)) { |
| 231 |
$author = get_userdata($post->post_author); |
| 232 |
if ($author && !empty($author->display_name)) { |
| 233 |
$author_name = $author->display_name; |
| 234 |
} |
| 235 |
} |
| 236 |
|
| 237 |
$frontmatter = "---\n"; |
| 238 |
$frontmatter .= 'title: ' . self::yaml_escape($post->post_title) . "\n"; |
| 239 |
$frontmatter .= 'slug: ' . self::yaml_escape($post->post_name) . "\n"; |
| 240 |
$frontmatter .= 'date: ' . self::yaml_escape($post->post_date) . "\n"; |
| 241 |
if ($author_name !== '') { |
| 242 |
$frontmatter .= 'author: ' . self::yaml_escape($author_name) . "\n"; |
| 243 |
} |
| 244 |
$frontmatter .= "---\n\n"; |
| 245 |
$prefix .= $frontmatter; |
| 246 |
} |
| 247 |
|
| 248 |
if (!empty($options['include_featured_image'])) { |
| 249 |
$thumb_id = get_post_thumbnail_id($post_id); |
| 250 |
if ($thumb_id) { |
| 251 |
$src = wp_get_attachment_url($thumb_id); |
| 252 |
if ($src) { |
| 253 |
$alt = get_post_meta($thumb_id, '_wp_attachment_image_alt', true); |
| 254 |
if (!is_string($alt)) { |
| 255 |
$alt = ''; |
| 256 |
} |
| 257 |
if ($alt === '') { |
| 258 |
$alt = $post->post_title; |
| 259 |
} |
| 260 |
$prefix .= '![' . $alt . '](' . $src . ")\n\n"; |
| 261 |
} |
| 262 |
} |
| 263 |
} |
| 264 |
|
| 265 |
return $prefix . $markdown; |
| 266 |
} |
| 267 |
|
| 268 |
/** |
| 269 |
* Escape a string for safe use as a YAML scalar value. |
| 270 |
* |
| 271 |
* @param string $value |
| 272 |
* @return string |
| 273 |
*/ |
| 274 |
private static function yaml_escape($value) |
| 275 |
{ |
| 276 |
$value = (string) $value; |
| 277 |
// Wrap in double quotes and escape backslashes and quotes. |
| 278 |
$escaped = str_replace(['\\', '"'], ['\\\\', '\\"'], $value); |
| 279 |
return '"' . $escaped . '"'; |
| 280 |
} |
| 281 |
|
| 282 |
/** |
| 283 |
* Strip HTML tags but leave a usable inline form (decodes entities later). |
| 284 |
* |
| 285 |
* @param string $html |
| 286 |
* @return string |
| 287 |
*/ |
| 288 |
private static function strip_tags_preserve($html) |
| 289 |
{ |
| 290 |
// Replace block-ish containers with newlines so stripping doesn't mash content. |
| 291 |
$html = preg_replace('#</(?:div|section|article|aside|header|footer|main|nav|figure|figcaption)>#i', "\n", $html); |
| 292 |
$html = preg_replace('#<(?:div|section|article|aside|header|footer|main|nav|figure|figcaption)\b[^>]*>#i', '', $html); |
| 293 |
// Strip remaining tags. |
| 294 |
return strip_tags($html); |
| 295 |
} |
| 296 |
|
| 297 |
/** |
| 298 |
* Convert a <table> block into a pipe-style markdown table. |
| 299 |
* |
| 300 |
* @param array $matches Regex matches from convert(); index 1 is inner HTML. |
| 301 |
* @return string |
| 302 |
*/ |
| 303 |
private static function convert_table($matches) |
| 304 |
{ |
| 305 |
$inner = $matches[1]; |
| 306 |
|
| 307 |
// Flatten any thead/tbody wrappers for easier row extraction. |
| 308 |
$inner = preg_replace('#</?(?:thead|tbody|tfoot)\b[^>]*>#i', '', $inner); |
| 309 |
|
| 310 |
if (!preg_match_all('#<tr\b[^>]*>(.*?)</tr>#is', $inner, $rows_match)) { |
| 311 |
return ''; |
| 312 |
} |
| 313 |
|
| 314 |
$rows = []; |
| 315 |
$header_used = false; |
| 316 |
foreach ($rows_match[1] as $row_index => $row_html) { |
| 317 |
$cells = []; |
| 318 |
$is_header_row = false; |
| 319 |
if (preg_match_all('#<(th|td)\b[^>]*>(.*?)</\1>#is', $row_html, $cell_match, PREG_SET_ORDER)) { |
| 320 |
foreach ($cell_match as $cm) { |
| 321 |
$cell_text = trim(self::strip_tags_preserve($cm[2])); |
| 322 |
$cell_text = html_entity_decode($cell_text, ENT_QUOTES | ENT_HTML5, 'UTF-8'); |
| 323 |
$cell_text = str_replace(['|', "\n", "\r"], ['\\|', ' ', ' '], $cell_text); |
| 324 |
$cells[] = $cell_text; |
| 325 |
if (strtolower($cm[1]) === 'th') { |
| 326 |
$is_header_row = true; |
| 327 |
} |
| 328 |
} |
| 329 |
} |
| 330 |
if (empty($cells)) { |
| 331 |
continue; |
| 332 |
} |
| 333 |
$rows[] = ['cells' => $cells, 'is_header' => $is_header_row]; |
| 334 |
} |
| 335 |
|
| 336 |
if (empty($rows)) { |
| 337 |
return ''; |
| 338 |
} |
| 339 |
|
| 340 |
$output = "\n\n"; |
| 341 |
$col_count = 0; |
| 342 |
foreach ($rows as $row) { |
| 343 |
if (count($row['cells']) > $col_count) { |
| 344 |
$col_count = count($row['cells']); |
| 345 |
} |
| 346 |
} |
| 347 |
|
| 348 |
// Header row: use first header row, else synthesise from first row. |
| 349 |
$header_row = null; |
| 350 |
foreach ($rows as $idx => $row) { |
| 351 |
if ($row['is_header']) { |
| 352 |
$header_row = $row; |
| 353 |
unset($rows[$idx]); |
| 354 |
break; |
| 355 |
} |
| 356 |
} |
| 357 |
if ($header_row === null) { |
| 358 |
$header_row = array_shift($rows); |
| 359 |
} |
| 360 |
|
| 361 |
$header_cells = array_pad($header_row['cells'], $col_count, ''); |
| 362 |
$output .= '| ' . implode(' | ', $header_cells) . " |\n"; |
| 363 |
$output .= '|' . str_repeat(' --- |', $col_count) . "\n"; |
| 364 |
|
| 365 |
foreach ($rows as $row) { |
| 366 |
$cells = array_pad($row['cells'], $col_count, ''); |
| 367 |
$output .= '| ' . implode(' | ', $cells) . " |\n"; |
| 368 |
} |
| 369 |
|
| 370 |
return $output . "\n"; |
| 371 |
} |
| 372 |
|
| 373 |
/** |
| 374 |
* Convert <ul>/<ol> lists to markdown, preserving nesting. |
| 375 |
* |
| 376 |
* Iteratively replaces the inner-most list until no lists remain. |
| 377 |
* |
| 378 |
* @param string $html |
| 379 |
* @return string |
| 380 |
*/ |
| 381 |
private static function convert_lists($html) |
| 382 |
{ |
| 383 |
// Loop until there are no lists left (handles nesting). |
| 384 |
$guard = 0; |
| 385 |
while ($guard < 50 && preg_match('#<(ul|ol)\b[^>]*>((?:(?!<ul|<ol).)*?)</\1>#is', $html, $m, PREG_OFFSET_CAPTURE)) { |
| 386 |
$tag = strtolower($m[1][0]); |
| 387 |
$inner = $m[2][0]; |
| 388 |
$full = $m[0][0]; |
| 389 |
$offset = $m[0][1]; |
| 390 |
|
| 391 |
$converted = self::render_list_items($inner, $tag, 0); |
| 392 |
$html = substr($html, 0, $offset) . "\n\n" . $converted . "\n\n" . substr($html, $offset + strlen($full)); |
| 393 |
$guard++; |
| 394 |
} |
| 395 |
|
| 396 |
// Clean up anything left unconverted (defensive). |
| 397 |
$html = preg_replace('#</?(?:ul|ol)\b[^>]*>#i', '', $html); |
| 398 |
$html = preg_replace('#<li\b[^>]*>#i', "- ", $html); |
| 399 |
$html = preg_replace('#</li>#i', "\n", $html); |
| 400 |
|
| 401 |
return $html; |
| 402 |
} |
| 403 |
|
| 404 |
/** |
| 405 |
* Render <li> items for a given list tag. |
| 406 |
* |
| 407 |
* @param string $inner Inner HTML of the list. |
| 408 |
* @param string $tag 'ul' or 'ol'. |
| 409 |
* @param int $depth Current nesting depth. |
| 410 |
* @return string |
| 411 |
*/ |
| 412 |
private static function render_list_items($inner, $tag, $depth) |
| 413 |
{ |
| 414 |
if (!preg_match_all('#<li\b[^>]*>(.*?)</li>#is', $inner, $items)) { |
| 415 |
return ''; |
| 416 |
} |
| 417 |
|
| 418 |
$out = ''; |
| 419 |
$indent = str_repeat(' ', max(0, $depth)); |
| 420 |
$counter = 1; |
| 421 |
|
| 422 |
foreach ($items[1] as $item_html) { |
| 423 |
// Recursively convert nested lists inside the <li>. |
| 424 |
$nested_converted = $item_html; |
| 425 |
while (preg_match('#<(ul|ol)\b[^>]*>((?:(?!<ul|<ol).)*?)</\1>#is', $nested_converted, $m, PREG_OFFSET_CAPTURE)) { |
| 426 |
$inner_nested = $m[2][0]; |
| 427 |
$inner_tag = strtolower($m[1][0]); |
| 428 |
$sub = self::render_list_items($inner_nested, $inner_tag, $depth + 1); |
| 429 |
$nested_converted = substr($nested_converted, 0, $m[0][1]) |
| 430 |
. "\n" . $sub |
| 431 |
. substr($nested_converted, $m[0][1] + strlen($m[0][0])); |
| 432 |
} |
| 433 |
|
| 434 |
$text = trim(self::strip_tags_preserve($nested_converted)); |
| 435 |
if ($text === '') { |
| 436 |
continue; |
| 437 |
} |
| 438 |
|
| 439 |
$marker = ($tag === 'ol') ? ($counter . '.') : '-'; |
| 440 |
// The line itself. |
| 441 |
$lines = preg_split('/\r?\n/', $text); |
| 442 |
$first = array_shift($lines); |
| 443 |
$out .= $indent . $marker . ' ' . ltrim($first) . "\n"; |
| 444 |
foreach ($lines as $line) { |
| 445 |
if (trim($line) === '') { |
| 446 |
continue; |
| 447 |
} |
| 448 |
$out .= $indent . ' ' . ltrim($line) . "\n"; |
| 449 |
} |
| 450 |
$counter++; |
| 451 |
} |
| 452 |
|
| 453 |
return rtrim($out, "\n"); |
| 454 |
} |
| 455 |
} |
| 456 |
|
