| 1 |
<?php |
| 2 |
|
| 3 |
namespace Soundasleep; |
| 4 |
|
| 5 |
class Html2Text { |
| 6 |
|
| 7 |
/** @return array<string, bool | string> */ |
| 8 |
public static function defaultOptions(): array { |
| 9 |
return [ |
| 10 |
'ignore_errors' => false, |
| 11 |
'drop_links' => false, |
| 12 |
'char_set' => 'auto' |
| 13 |
]; |
| 14 |
} |
| 15 |
|
| 16 |
/** |
| 17 |
* Tries to convert the given HTML into a plain text format - best suited for |
| 18 |
* e-mail display, etc. |
| 19 |
* |
| 20 |
* <p>In particular, it tries to maintain the following features: |
| 21 |
* <ul> |
| 22 |
* <li>Links are maintained, with the 'href' copied over |
| 23 |
* <li>Information in the <head> is lost |
| 24 |
* </ul> |
| 25 |
* |
| 26 |
* @param string $html the input HTML |
| 27 |
* @param boolean|array<string, bool | string> $options if boolean, Ignore xml parsing errors, else ['ignore_errors' => false, 'drop_links' => false, 'char_set' => 'auto'] |
| 28 |
* @return string the HTML converted, as best as possible, to text |
| 29 |
* @throws Html2TextException if the HTML could not be loaded as a {@link \DOMDocument} |
| 30 |
*/ |
| 31 |
public static function convert(string $html, $options = []): string { |
| 32 |
|
| 33 |
if ($options === false || $options === true) { |
| 34 |
// Using old style (< 1.0) of passing in options |
| 35 |
$options = ['ignore_errors' => $options]; |
| 36 |
} |
| 37 |
|
| 38 |
$options = array_merge(static::defaultOptions(), $options); |
| 39 |
|
| 40 |
// check all options are valid |
| 41 |
foreach ($options as $key => $value) { |
| 42 |
if (!in_array($key, array_keys(static::defaultOptions()))) { |
| 43 |
throw new \InvalidArgumentException("Unknown html2text option '$key'. Valid options are " . implode(',', static::defaultOptions())); |
| 44 |
} |
| 45 |
} |
| 46 |
|
| 47 |
$is_office_document = self::isOfficeDocument($html); |
| 48 |
|
| 49 |
if ($is_office_document) { |
| 50 |
// remove office namespace |
| 51 |
$html = str_replace(["<o:p>", "</o:p>"], "", $html); |
| 52 |
} |
| 53 |
|
| 54 |
$html = self::fixNewlines($html); |
| 55 |
|
| 56 |
// use mb_convert_encoding for legacy versions of php |
| 57 |
if (PHP_MAJOR_VERSION * 10 + PHP_MINOR_VERSION < 81 && mb_detect_encoding($html, "UTF-8", true)) { |
| 58 |
$html = mb_convert_encoding($html, "HTML-ENTITIES", "UTF-8"); |
| 59 |
} |
| 60 |
|
| 61 |
$doc = self::getDocument($html, $options); |
| 62 |
|
| 63 |
$output = self::iterateOverNode($doc, null, false, $is_office_document, $options); |
| 64 |
|
| 65 |
// process output for whitespace/newlines |
| 66 |
$output = self::processWhitespaceNewlines($output); |
| 67 |
|
| 68 |
return $output; |
| 69 |
} |
| 70 |
|
| 71 |
/** |
| 72 |
* Unify newlines; in particular, \r\n becomes \n, and |
| 73 |
* then \r becomes \n. This means that all newlines (Unix, Windows, Mac) |
| 74 |
* all become \ns. |
| 75 |
* |
| 76 |
* @param string $text text with any number of \r, \r\n and \n combinations |
| 77 |
* @return string the fixed text |
| 78 |
*/ |
| 79 |
public static function fixNewlines(string $text): string { |
| 80 |
// replace \r\n to \n |
| 81 |
$text = str_replace("\r\n", "\n", $text); |
| 82 |
// remove \rs |
| 83 |
$text = str_replace("\r", "\n", $text); |
| 84 |
|
| 85 |
return $text; |
| 86 |
} |
| 87 |
|
| 88 |
/** @return array<string> */ |
| 89 |
public static function nbspCodes(): array { |
| 90 |
return [ |
| 91 |
"\xc2\xa0", |
| 92 |
"\u00a0", |
| 93 |
]; |
| 94 |
} |
| 95 |
|
| 96 |
/** @return array<string> */ |
| 97 |
public static function zwnjCodes(): array { |
| 98 |
return [ |
| 99 |
"\xe2\x80\x8c", |
| 100 |
"\u200c", |
| 101 |
]; |
| 102 |
} |
| 103 |
|
| 104 |
/** |
| 105 |
* Remove leading or trailing spaces and excess empty lines from provided multiline text |
| 106 |
* |
| 107 |
* @param string $text multiline text any number of leading or trailing spaces or excess lines |
| 108 |
* @return string the fixed text |
| 109 |
*/ |
| 110 |
public static function processWhitespaceNewlines(string $text): string { |
| 111 |
|
| 112 |
// remove excess spaces around tabs |
| 113 |
$text = preg_replace("/ *\t */im", "\t", $text); |
| 114 |
|
| 115 |
// remove leading whitespace |
| 116 |
$text = ltrim($text); |
| 117 |
|
| 118 |
// remove leading spaces on each line |
| 119 |
$text = preg_replace("/\n[ \t]*/im", "\n", $text); |
| 120 |
|
| 121 |
// convert non-breaking spaces to regular spaces to prevent output issues, |
| 122 |
// do it here so they do NOT get removed with other leading spaces, as they |
| 123 |
// are sometimes used for indentation |
| 124 |
$text = self::renderText($text); |
| 125 |
|
| 126 |
// remove trailing whitespace |
| 127 |
$text = rtrim($text); |
| 128 |
|
| 129 |
// remove trailing spaces on each line |
| 130 |
$text = preg_replace("/[ \t]*\n/im", "\n", $text); |
| 131 |
|
| 132 |
// unarmor pre blocks |
| 133 |
$text = self::fixNewLines($text); |
| 134 |
|
| 135 |
// remove unnecessary empty lines |
| 136 |
$text = preg_replace("/\n\n\n*/im", "\n\n", $text); |
| 137 |
|
| 138 |
return $text; |
| 139 |
} |
| 140 |
|
| 141 |
/** |
| 142 |
* Can we guess that this HTML is generated by Microsoft Office? |
| 143 |
*/ |
| 144 |
public static function isOfficeDocument(string $html): bool { |
| 145 |
return strpos($html, "urn:schemas-microsoft-com:office") !== false; |
| 146 |
} |
| 147 |
|
| 148 |
public static function isWhitespace(string $text): bool { |
| 149 |
return strlen(trim(self::renderText($text), "\n\r\t ")) === 0; |
| 150 |
} |
| 151 |
|
| 152 |
/** |
| 153 |
* Parse HTML into a DOMDocument |
| 154 |
* |
| 155 |
* @param string $html the input HTML |
| 156 |
* @param array<string, bool | string> $options |
| 157 |
* @return \DOMDocument the parsed document tree |
| 158 |
*/ |
| 159 |
private static function getDocument(string $html, array $options): \DOMDocument { |
| 160 |
|
| 161 |
$doc = new \DOMDocument(); |
| 162 |
|
| 163 |
$html = trim($html); |
| 164 |
|
| 165 |
if (!$html) { |
| 166 |
// DOMDocument doesn't support empty value and throws an error |
| 167 |
// Return empty document instead |
| 168 |
return $doc; |
| 169 |
} |
| 170 |
|
| 171 |
if ($html[0] !== '<') { |
| 172 |
// If HTML does not begin with a tag, we put a body tag around it. |
| 173 |
// If we do not do this, PHP will insert a paragraph tag around |
| 174 |
// the first block of text for some reason which can mess up |
| 175 |
// the newlines. See pre.html test for an example. |
| 176 |
$html = '<body>' . $html . '</body>'; |
| 177 |
} |
| 178 |
|
| 179 |
$header = ''; |
| 180 |
// use char sets for modern versions of php |
| 181 |
if (PHP_MAJOR_VERSION * 10 + PHP_MINOR_VERSION >= 81) { |
| 182 |
// use specified char_set, or auto detect if not set |
| 183 |
$char_set = ! empty($options['char_set']) ? $options['char_set'] : 'auto'; |
| 184 |
if ('auto' === $char_set) { |
| 185 |
$char_set = mb_detect_encoding($html); |
| 186 |
} else if (strpos($char_set, ',')) { |
| 187 |
mb_detect_order($char_set); |
| 188 |
$char_set = mb_detect_encoding($html); |
| 189 |
} |
| 190 |
// turn off error detection for Windows-1252 legacy html |
| 191 |
if (strpos($char_set, '1252')) { |
| 192 |
$options['ignore_errors'] = true; |
| 193 |
} |
| 194 |
$header = '<?xml version="1.0" encoding="' . $char_set . '">'; |
| 195 |
} |
| 196 |
|
| 197 |
if (! empty($options['ignore_errors'])) { |
| 198 |
$doc->strictErrorChecking = false; |
| 199 |
$doc->recover = true; |
| 200 |
$doc->xmlStandalone = true; |
| 201 |
$old_internal_errors = libxml_use_internal_errors(true); |
| 202 |
$load_result = $doc->loadHTML($header . $html, LIBXML_NOWARNING | LIBXML_NOERROR | LIBXML_NONET | LIBXML_PARSEHUGE); |
| 203 |
libxml_use_internal_errors($old_internal_errors); |
| 204 |
} |
| 205 |
else { |
| 206 |
$load_result = $doc->loadHTML($header . $html); |
| 207 |
} |
| 208 |
|
| 209 |
if (!$load_result) { |
| 210 |
throw new Html2TextException("Could not load HTML - badly formed?", $html); |
| 211 |
} |
| 212 |
|
| 213 |
return $doc; |
| 214 |
} |
| 215 |
|
| 216 |
/** |
| 217 |
* Replace any special characters with simple text versions, to prevent output issues: |
| 218 |
* - Convert non-breaking spaces to regular spaces; and |
| 219 |
* - Convert zero-width non-joiners to '' (nothing). |
| 220 |
* |
| 221 |
* This is to match our goal of rendering documents as they would be rendered |
| 222 |
* by a browser. |
| 223 |
*/ |
| 224 |
private static function renderText(string $text): string { |
| 225 |
$text = str_replace(self::nbspCodes(), " ", $text); |
| 226 |
$text = str_replace(self::zwnjCodes(), "", $text); |
| 227 |
return $text; |
| 228 |
} |
| 229 |
|
| 230 |
private static function nextChildName(?\DOMNode $node): ?string { |
| 231 |
// get the next child |
| 232 |
$nextNode = $node->nextSibling; |
| 233 |
while ($nextNode != null) { |
| 234 |
if ($nextNode instanceof \DOMText) { |
| 235 |
if (!self::isWhitespace($nextNode->wholeText)) { |
| 236 |
break; |
| 237 |
} |
| 238 |
} |
| 239 |
|
| 240 |
if ($nextNode instanceof \DOMElement) { |
| 241 |
break; |
| 242 |
} |
| 243 |
|
| 244 |
$nextNode = $nextNode->nextSibling; |
| 245 |
} |
| 246 |
|
| 247 |
$nextName = null; |
| 248 |
if (($nextNode instanceof \DOMElement || $nextNode instanceof \DOMText) && $nextNode != null) { |
| 249 |
$nextName = strtolower($nextNode->nodeName); |
| 250 |
} |
| 251 |
|
| 252 |
return $nextName; |
| 253 |
} |
| 254 |
|
| 255 |
/** @param array<string, bool | string> $options */ |
| 256 |
private static function iterateOverNode(\DOMNode $node, ?string $prevName, bool $in_pre, bool $is_office_document, array $options): string { |
| 257 |
if ($node instanceof \DOMText) { |
| 258 |
// Replace whitespace characters with a space (equivilant to \s) |
| 259 |
if ($in_pre) { |
| 260 |
$text = "\n" . trim(self::renderText($node->wholeText), "\n\r\t ") . "\n"; |
| 261 |
|
| 262 |
// Remove trailing whitespace only |
| 263 |
$text = preg_replace("/[ \t]*\n/im", "\n", $text); |
| 264 |
|
| 265 |
// armor newlines with \r. |
| 266 |
return str_replace("\n", "\r", $text); |
| 267 |
|
| 268 |
} |
| 269 |
$text = self::renderText($node->wholeText); |
| 270 |
$text = preg_replace("/[\\t\\n\\f\\r ]+/im", " ", $text); |
| 271 |
|
| 272 |
if (!self::isWhitespace($text) && ($prevName == 'p' || $prevName == 'div')) { |
| 273 |
return "\n" . $text; |
| 274 |
} |
| 275 |
return $text; |
| 276 |
} |
| 277 |
|
| 278 |
if ($node instanceof \DOMDocumentType || $node instanceof \DOMProcessingInstruction) { |
| 279 |
// ignore |
| 280 |
return ""; |
| 281 |
} |
| 282 |
|
| 283 |
$name = strtolower($node->nodeName); |
| 284 |
$nextName = self::nextChildName($node); |
| 285 |
|
| 286 |
// start whitespace |
| 287 |
switch ($name) { |
| 288 |
case "hr": |
| 289 |
$prefix = ''; |
| 290 |
if ($prevName != null) { |
| 291 |
$prefix = "\n"; |
| 292 |
} |
| 293 |
return $prefix . "---------------------------------------------------------------\n"; |
| 294 |
|
| 295 |
case "style": |
| 296 |
case "head": |
| 297 |
case "title": |
| 298 |
case "meta": |
| 299 |
case "script": |
| 300 |
// ignore these tags |
| 301 |
return ""; |
| 302 |
|
| 303 |
case "h1": |
| 304 |
case "h2": |
| 305 |
case "h3": |
| 306 |
case "h4": |
| 307 |
case "h5": |
| 308 |
case "h6": |
| 309 |
case "ol": |
| 310 |
case "ul": |
| 311 |
case "pre": |
| 312 |
// add two newlines |
| 313 |
$output = "\n\n"; |
| 314 |
break; |
| 315 |
|
| 316 |
case "td": |
| 317 |
case "th": |
| 318 |
// add tab char to separate table fields |
| 319 |
$output = "\t"; |
| 320 |
break; |
| 321 |
|
| 322 |
case "p": |
| 323 |
// Microsoft exchange emails often include HTML which, when passed through |
| 324 |
// html2text, results in lots of double line returns everywhere. |
| 325 |
// |
| 326 |
// To fix this, for any p element with a className of `MsoNormal` (the standard |
| 327 |
// classname in any Microsoft export or outlook for a paragraph that behaves |
| 328 |
// like a line return) we skip the first line returns and set the name to br. |
| 329 |
// @phpstan-ignore-next-line |
| 330 |
if ($is_office_document && $node->getAttribute('class') == 'MsoNormal') { |
| 331 |
$output = ""; |
| 332 |
$name = 'br'; |
| 333 |
break; |
| 334 |
} |
| 335 |
|
| 336 |
// add two lines |
| 337 |
$output = "\n\n"; |
| 338 |
break; |
| 339 |
|
| 340 |
case "tr": |
| 341 |
// add one line |
| 342 |
$output = "\n"; |
| 343 |
break; |
| 344 |
|
| 345 |
case "div": |
| 346 |
$output = ""; |
| 347 |
if ($prevName !== null) { |
| 348 |
// add one line |
| 349 |
$output .= "\n"; |
| 350 |
} |
| 351 |
break; |
| 352 |
|
| 353 |
case "li": |
| 354 |
$output = "- "; |
| 355 |
break; |
| 356 |
|
| 357 |
default: |
| 358 |
// print out contents of unknown tags |
| 359 |
$output = ""; |
| 360 |
break; |
| 361 |
} |
| 362 |
|
| 363 |
// debug |
| 364 |
//$output .= "[$name,$nextName]"; |
| 365 |
|
| 366 |
if (isset($node->childNodes)) { |
| 367 |
|
| 368 |
$n = $node->childNodes->item(0); |
| 369 |
$previousSiblingNames = []; |
| 370 |
$previousSiblingName = null; |
| 371 |
|
| 372 |
$parts = []; |
| 373 |
$trailing_whitespace = 0; |
| 374 |
|
| 375 |
while ($n != null) { |
| 376 |
|
| 377 |
$text = self::iterateOverNode($n, $previousSiblingName, $in_pre || $name == 'pre', $is_office_document, $options); |
| 378 |
|
| 379 |
// Pass current node name to next child, as previousSibling does not appear to get populated |
| 380 |
if ($n instanceof \DOMDocumentType |
| 381 |
|| $n instanceof \DOMProcessingInstruction |
| 382 |
|| ($n instanceof \DOMText && self::isWhitespace($text))) { |
| 383 |
// Keep current previousSiblingName, these are invisible |
| 384 |
$trailing_whitespace++; |
| 385 |
} |
| 386 |
else { |
| 387 |
$previousSiblingName = strtolower($n->nodeName); |
| 388 |
$previousSiblingNames[] = $previousSiblingName; |
| 389 |
$trailing_whitespace = 0; |
| 390 |
} |
| 391 |
|
| 392 |
$node->removeChild($n); |
| 393 |
$n = $node->childNodes->item(0); |
| 394 |
|
| 395 |
$parts[] = $text; |
| 396 |
} |
| 397 |
|
| 398 |
// Remove trailing whitespace, important for the br check below |
| 399 |
while ($trailing_whitespace-- > 0) { |
| 400 |
array_pop($parts); |
| 401 |
} |
| 402 |
|
| 403 |
// suppress last br tag inside a node list if follows text |
| 404 |
$last_name = array_pop($previousSiblingNames); |
| 405 |
if ($last_name === 'br') { |
| 406 |
$last_name = array_pop($previousSiblingNames); |
| 407 |
if ($last_name === '#text') { |
| 408 |
array_pop($parts); |
| 409 |
} |
| 410 |
} |
| 411 |
|
| 412 |
$output .= implode('', $parts); |
| 413 |
} |
| 414 |
|
| 415 |
// end whitespace |
| 416 |
switch ($name) { |
| 417 |
case "h1": |
| 418 |
case "h2": |
| 419 |
case "h3": |
| 420 |
case "h4": |
| 421 |
case "h5": |
| 422 |
case "h6": |
| 423 |
case "pre": |
| 424 |
case "p": |
| 425 |
// add two lines |
| 426 |
$output .= "\n\n"; |
| 427 |
break; |
| 428 |
|
| 429 |
case "br": |
| 430 |
// add one line |
| 431 |
$output .= "\n"; |
| 432 |
break; |
| 433 |
|
| 434 |
case "div": |
| 435 |
break; |
| 436 |
|
| 437 |
case "a": |
| 438 |
// links are returned in [text](link) format |
| 439 |
// @phpstan-ignore-next-line |
| 440 |
$href = $node->getAttribute("href"); |
| 441 |
|
| 442 |
$output = trim($output); |
| 443 |
|
| 444 |
// remove double [[ ]] s from linking images |
| 445 |
if (substr($output, 0, 1) == "[" && substr($output, -1) == "]") { |
| 446 |
$output = substr($output, 1, strlen($output) - 2); |
| 447 |
|
| 448 |
// for linking images, the title of the <a> overrides the title of the <img> |
| 449 |
// @phpstan-ignore-next-line |
| 450 |
if ($node->getAttribute("title")) { |
| 451 |
// @phpstan-ignore-next-line |
| 452 |
$output = $node->getAttribute("title"); |
| 453 |
} |
| 454 |
} |
| 455 |
|
| 456 |
// if there is no link text, but a title attr |
| 457 |
// @phpstan-ignore-next-line |
| 458 |
if (!$output && $node->getAttribute("title")) { |
| 459 |
// @phpstan-ignore-next-line |
| 460 |
$output = $node->getAttribute("title"); |
| 461 |
} |
| 462 |
|
| 463 |
if ($href == null) { |
| 464 |
// it doesn't link anywhere |
| 465 |
// @phpstan-ignore-next-line |
| 466 |
if ($node->getAttribute("name") != null) { |
| 467 |
if ($options['drop_links']) { |
| 468 |
$output = "$output"; |
| 469 |
} else { |
| 470 |
$output = "[$output]"; |
| 471 |
} |
| 472 |
} |
| 473 |
} else { |
| 474 |
if ($href == $output || $href == "mailto:$output" || $href == "http://$output" || $href == "https://$output") { |
| 475 |
// link to the same address: just use link |
| 476 |
$output = "$output"; |
| 477 |
} else { |
| 478 |
// replace it |
| 479 |
if ($output) { |
| 480 |
if ($options['drop_links']) { |
| 481 |
$output = "$output"; |
| 482 |
} else { |
| 483 |
$output = "[$output]($href)"; |
| 484 |
} |
| 485 |
} else { |
| 486 |
// empty string |
| 487 |
$output = "$href"; |
| 488 |
} |
| 489 |
} |
| 490 |
} |
| 491 |
|
| 492 |
// does the next node require additional whitespace? |
| 493 |
switch ($nextName) { |
| 494 |
case "h1": case "h2": case "h3": case "h4": case "h5": case "h6": |
| 495 |
$output .= "\n"; |
| 496 |
break; |
| 497 |
} |
| 498 |
break; |
| 499 |
|
| 500 |
case "img": |
| 501 |
// @phpstan-ignore-next-line |
| 502 |
if ($node->getAttribute("title")) { |
| 503 |
// @phpstan-ignore-next-line |
| 504 |
$output = "[" . $node->getAttribute("title") . "]"; |
| 505 |
// @phpstan-ignore-next-line |
| 506 |
} elseif ($node->getAttribute("alt")) { |
| 507 |
// @phpstan-ignore-next-line |
| 508 |
$output = "[" . $node->getAttribute("alt") . "]"; |
| 509 |
} else { |
| 510 |
$output = ""; |
| 511 |
} |
| 512 |
break; |
| 513 |
|
| 514 |
case "li": |
| 515 |
$output .= "\n"; |
| 516 |
break; |
| 517 |
|
| 518 |
case "blockquote": |
| 519 |
// process quoted text for whitespace/newlines |
| 520 |
$output = self::processWhitespaceNewlines($output); |
| 521 |
|
| 522 |
// add leading newline |
| 523 |
$output = "\n" . $output; |
| 524 |
|
| 525 |
// prepend '> ' at the beginning of all lines |
| 526 |
$output = preg_replace("/\n/im", "\n> ", $output); |
| 527 |
|
| 528 |
// replace leading '> >' with '>>' |
| 529 |
$output = preg_replace("/\n> >/im", "\n>>", $output); |
| 530 |
|
| 531 |
// add another leading newline and trailing newlines |
| 532 |
$output = "\n" . $output . "\n\n"; |
| 533 |
break; |
| 534 |
default: |
| 535 |
// do nothing |
| 536 |
} |
| 537 |
|
| 538 |
return $output; |
| 539 |
} |
| 540 |
} |
| 541 |
|
