HTML5Entities.php
1 year ago
OutputRules.php
1 year ago
README.md
1 year ago
RulesInterface.php
1 year ago
Traverser.php
1 year ago
OutputRules.php
554 lines
| 1 | <?php |
| 2 | /** |
| 3 | * @file |
| 4 | * The rules for generating output in the serializer. |
| 5 | * |
| 6 | * These output rules are likely to generate output similar to the document that |
| 7 | * was parsed. It is not intended to output exactly the document that was parsed. |
| 8 | */ |
| 9 | |
| 10 | namespace Masterminds\HTML5\Serializer; |
| 11 | |
| 12 | use Masterminds\HTML5\Elements; |
| 13 | |
| 14 | /** |
| 15 | * Generate the output html5 based on element rules. |
| 16 | */ |
| 17 | class OutputRules implements RulesInterface |
| 18 | { |
| 19 | /** |
| 20 | * Defined in http://www.w3.org/TR/html51/infrastructure.html#html-namespace-0. |
| 21 | */ |
| 22 | const NAMESPACE_HTML = 'http://www.w3.org/1999/xhtml'; |
| 23 | |
| 24 | const NAMESPACE_MATHML = 'http://www.w3.org/1998/Math/MathML'; |
| 25 | |
| 26 | const NAMESPACE_SVG = 'http://www.w3.org/2000/svg'; |
| 27 | |
| 28 | const NAMESPACE_XLINK = 'http://www.w3.org/1999/xlink'; |
| 29 | |
| 30 | const NAMESPACE_XML = 'http://www.w3.org/XML/1998/namespace'; |
| 31 | |
| 32 | const NAMESPACE_XMLNS = 'http://www.w3.org/2000/xmlns/'; |
| 33 | |
| 34 | /** |
| 35 | * Holds the HTML5 element names that causes a namespace switch. |
| 36 | * |
| 37 | * @var array |
| 38 | */ |
| 39 | protected $implicitNamespaces = array( |
| 40 | self::NAMESPACE_HTML, |
| 41 | self::NAMESPACE_SVG, |
| 42 | self::NAMESPACE_MATHML, |
| 43 | self::NAMESPACE_XML, |
| 44 | self::NAMESPACE_XMLNS, |
| 45 | ); |
| 46 | |
| 47 | const IM_IN_HTML = 1; |
| 48 | |
| 49 | const IM_IN_SVG = 2; |
| 50 | |
| 51 | const IM_IN_MATHML = 3; |
| 52 | |
| 53 | /** |
| 54 | * Used as cache to detect if is available ENT_HTML5. |
| 55 | * |
| 56 | * @var bool |
| 57 | */ |
| 58 | private $hasHTML5 = false; |
| 59 | |
| 60 | protected $traverser; |
| 61 | |
| 62 | protected $encode = false; |
| 63 | |
| 64 | protected $out; |
| 65 | |
| 66 | protected $outputMode; |
| 67 | |
| 68 | private $xpath; |
| 69 | |
| 70 | protected $nonBooleanAttributes = array( |
| 71 | /* |
| 72 | array( |
| 73 | 'nodeNamespace'=>'http://www.w3.org/1999/xhtml', |
| 74 | 'attrNamespace'=>'http://www.w3.org/1999/xhtml', |
| 75 | |
| 76 | 'nodeName'=>'img', 'nodeName'=>array('img', 'a'), |
| 77 | 'attrName'=>'alt', 'attrName'=>array('title', 'alt'), |
| 78 | ), |
| 79 | */ |
| 80 | array( |
| 81 | 'nodeNamespace' => 'http://www.w3.org/1999/xhtml', |
| 82 | 'attrName' => array('href', |
| 83 | 'hreflang', |
| 84 | 'http-equiv', |
| 85 | 'icon', |
| 86 | 'id', |
| 87 | 'keytype', |
| 88 | 'kind', |
| 89 | 'label', |
| 90 | 'lang', |
| 91 | 'language', |
| 92 | 'list', |
| 93 | 'maxlength', |
| 94 | 'media', |
| 95 | 'method', |
| 96 | 'name', |
| 97 | 'placeholder', |
| 98 | 'rel', |
| 99 | 'rows', |
| 100 | 'rowspan', |
| 101 | 'sandbox', |
| 102 | 'spellcheck', |
| 103 | 'scope', |
| 104 | 'seamless', |
| 105 | 'shape', |
| 106 | 'size', |
| 107 | 'sizes', |
| 108 | 'span', |
| 109 | 'src', |
| 110 | 'srcdoc', |
| 111 | 'srclang', |
| 112 | 'srcset', |
| 113 | 'start', |
| 114 | 'step', |
| 115 | 'style', |
| 116 | 'summary', |
| 117 | 'tabindex', |
| 118 | 'target', |
| 119 | 'title', |
| 120 | 'type', |
| 121 | 'value', |
| 122 | 'width', |
| 123 | 'border', |
| 124 | 'charset', |
| 125 | 'cite', |
| 126 | 'class', |
| 127 | 'code', |
| 128 | 'codebase', |
| 129 | 'color', |
| 130 | 'cols', |
| 131 | 'colspan', |
| 132 | 'content', |
| 133 | 'coords', |
| 134 | 'data', |
| 135 | 'datetime', |
| 136 | 'default', |
| 137 | 'dir', |
| 138 | 'dirname', |
| 139 | 'enctype', |
| 140 | 'for', |
| 141 | 'form', |
| 142 | 'formaction', |
| 143 | 'headers', |
| 144 | 'height', |
| 145 | 'accept', |
| 146 | 'accept-charset', |
| 147 | 'accesskey', |
| 148 | 'action', |
| 149 | 'align', |
| 150 | 'alt', |
| 151 | 'bgcolor', |
| 152 | ), |
| 153 | ), |
| 154 | array( |
| 155 | 'nodeNamespace' => 'http://www.w3.org/1999/xhtml', |
| 156 | 'xpath' => 'starts-with(local-name(), \'data-\')', |
| 157 | ), |
| 158 | ); |
| 159 | |
| 160 | const DOCTYPE = '<!DOCTYPE html>'; |
| 161 | |
| 162 | public function __construct($output, $options = array()) |
| 163 | { |
| 164 | if (isset($options['encode_entities'])) { |
| 165 | $this->encode = $options['encode_entities']; |
| 166 | } |
| 167 | |
| 168 | $this->outputMode = static::IM_IN_HTML; |
| 169 | $this->out = $output; |
| 170 | $this->hasHTML5 = defined('ENT_HTML5'); |
| 171 | } |
| 172 | |
| 173 | public function addRule(array $rule) |
| 174 | { |
| 175 | $this->nonBooleanAttributes[] = $rule; |
| 176 | } |
| 177 | |
| 178 | public function setTraverser(Traverser $traverser) |
| 179 | { |
| 180 | $this->traverser = $traverser; |
| 181 | |
| 182 | return $this; |
| 183 | } |
| 184 | |
| 185 | public function unsetTraverser() |
| 186 | { |
| 187 | $this->traverser = null; |
| 188 | |
| 189 | return $this; |
| 190 | } |
| 191 | |
| 192 | public function document($dom) |
| 193 | { |
| 194 | $this->doctype(); |
| 195 | if ($dom->documentElement) { |
| 196 | foreach ($dom->childNodes as $node) { |
| 197 | $this->traverser->node($node); |
| 198 | } |
| 199 | $this->nl(); |
| 200 | } |
| 201 | } |
| 202 | |
| 203 | protected function doctype() |
| 204 | { |
| 205 | $this->wr(static::DOCTYPE); |
| 206 | $this->nl(); |
| 207 | } |
| 208 | |
| 209 | public function element($ele) |
| 210 | { |
| 211 | $name = $ele->tagName; |
| 212 | |
| 213 | // Per spec: |
| 214 | // If the element has a declared namespace in the HTML, MathML or |
| 215 | // SVG namespaces, we use the lname instead of the tagName. |
| 216 | if ($this->traverser->isLocalElement($ele)) { |
| 217 | $name = $ele->localName; |
| 218 | } |
| 219 | |
| 220 | // If we are in SVG or MathML there is special handling. |
| 221 | // Using if/elseif instead of switch because it's faster in PHP. |
| 222 | if ('svg' == $name) { |
| 223 | $this->outputMode = static::IM_IN_SVG; |
| 224 | $name = Elements::normalizeSvgElement($name); |
| 225 | } elseif ('math' == $name) { |
| 226 | $this->outputMode = static::IM_IN_MATHML; |
| 227 | } |
| 228 | |
| 229 | $this->openTag($ele); |
| 230 | if (Elements::isA($name, Elements::TEXT_RAW)) { |
| 231 | foreach ($ele->childNodes as $child) { |
| 232 | if ($child instanceof \DOMCharacterData) { |
| 233 | $this->wr($child->data); |
| 234 | } elseif ($child instanceof \DOMElement) { |
| 235 | $this->element($child); |
| 236 | } |
| 237 | } |
| 238 | } else { |
| 239 | // Handle children. |
| 240 | if ($ele->hasChildNodes()) { |
| 241 | $this->traverser->children($ele->childNodes); |
| 242 | } |
| 243 | |
| 244 | // Close out the SVG or MathML special handling. |
| 245 | if ('svg' == $name || 'math' == $name) { |
| 246 | $this->outputMode = static::IM_IN_HTML; |
| 247 | } |
| 248 | } |
| 249 | |
| 250 | // If not unary, add a closing tag. |
| 251 | if (!Elements::isA($name, Elements::VOID_TAG)) { |
| 252 | $this->closeTag($ele); |
| 253 | } |
| 254 | } |
| 255 | |
| 256 | /** |
| 257 | * Write a text node. |
| 258 | * |
| 259 | * @param \DOMText $ele The text node to write. |
| 260 | */ |
| 261 | public function text($ele) |
| 262 | { |
| 263 | if (isset($ele->parentNode) && isset($ele->parentNode->tagName) && Elements::isA($ele->parentNode->localName, Elements::TEXT_RAW)) { |
| 264 | $this->wr($ele->data); |
| 265 | |
| 266 | return; |
| 267 | } |
| 268 | |
| 269 | // FIXME: This probably needs some flags set. |
| 270 | $this->wr($this->enc($ele->data)); |
| 271 | } |
| 272 | |
| 273 | public function cdata($ele) |
| 274 | { |
| 275 | // This encodes CDATA. |
| 276 | $this->wr($ele->ownerDocument->saveXML($ele)); |
| 277 | } |
| 278 | |
| 279 | public function comment($ele) |
| 280 | { |
| 281 | // These produce identical output. |
| 282 | // $this->wr('<!--')->wr($ele->data)->wr('-->'); |
| 283 | $this->wr($ele->ownerDocument->saveXML($ele)); |
| 284 | } |
| 285 | |
| 286 | public function processorInstruction($ele) |
| 287 | { |
| 288 | $this->wr('<?') |
| 289 | ->wr($ele->target) |
| 290 | ->wr(' ') |
| 291 | ->wr($ele->data) |
| 292 | ->wr('?>'); |
| 293 | } |
| 294 | |
| 295 | /** |
| 296 | * Write the namespace attributes. |
| 297 | * |
| 298 | * @param \DOMNode $ele The element being written. |
| 299 | */ |
| 300 | protected function namespaceAttrs($ele) |
| 301 | { |
| 302 | if (!$this->xpath || $this->xpath->document !== $ele->ownerDocument) { |
| 303 | $this->xpath = new \DOMXPath($ele->ownerDocument); |
| 304 | } |
| 305 | |
| 306 | foreach ($this->xpath->query('namespace::*[not(.=../../namespace::*)]', $ele) as $nsNode) { |
| 307 | if (!in_array($nsNode->nodeValue, $this->implicitNamespaces)) { |
| 308 | $this->wr(' ')->wr($nsNode->nodeName)->wr('="')->wr($nsNode->nodeValue)->wr('"'); |
| 309 | } |
| 310 | } |
| 311 | } |
| 312 | |
| 313 | /** |
| 314 | * Write the opening tag. |
| 315 | * |
| 316 | * Tags for HTML, MathML, and SVG are in the local name. Otherwise, use the |
| 317 | * qualified name (8.3). |
| 318 | * |
| 319 | * @param \DOMNode $ele The element being written. |
| 320 | */ |
| 321 | protected function openTag($ele) |
| 322 | { |
| 323 | $this->wr('<')->wr($this->traverser->isLocalElement($ele) ? $ele->localName : $ele->tagName); |
| 324 | |
| 325 | $this->attrs($ele); |
| 326 | $this->namespaceAttrs($ele); |
| 327 | |
| 328 | if ($this->outputMode == static::IM_IN_HTML) { |
| 329 | $this->wr('>'); |
| 330 | } // If we are not in html mode we are in SVG, MathML, or XML embedded content. |
| 331 | else { |
| 332 | if ($ele->hasChildNodes()) { |
| 333 | $this->wr('>'); |
| 334 | } // If there are no children this is self closing. |
| 335 | else { |
| 336 | $this->wr(' />'); |
| 337 | } |
| 338 | } |
| 339 | } |
| 340 | |
| 341 | protected function attrs($ele) |
| 342 | { |
| 343 | // FIXME: Needs support for xml, xmlns, xlink, and namespaced elements. |
| 344 | if (!$ele->hasAttributes()) { |
| 345 | return $this; |
| 346 | } |
| 347 | |
| 348 | // TODO: Currently, this always writes name="value", and does not do |
| 349 | // value-less attributes. |
| 350 | $map = $ele->attributes; |
| 351 | $len = $map->length; |
| 352 | for ($i = 0; $i < $len; ++$i) { |
| 353 | $node = $map->item($i); |
| 354 | $val = $this->enc($node->value, true); |
| 355 | |
| 356 | // XXX: The spec says that we need to ensure that anything in |
| 357 | // the XML, XMLNS, or XLink NS's should use the canonical |
| 358 | // prefix. It seems that DOM does this for us already, but there |
| 359 | // may be exceptions. |
| 360 | $name = $node->nodeName; |
| 361 | |
| 362 | // Special handling for attributes in SVG and MathML. |
| 363 | // Using if/elseif instead of switch because it's faster in PHP. |
| 364 | if ($this->outputMode == static::IM_IN_SVG) { |
| 365 | $name = Elements::normalizeSvgAttribute($name); |
| 366 | } elseif ($this->outputMode == static::IM_IN_MATHML) { |
| 367 | $name = Elements::normalizeMathMlAttribute($name); |
| 368 | } |
| 369 | |
| 370 | $this->wr(' ')->wr($name); |
| 371 | |
| 372 | if ((isset($val) && '' !== $val) || $this->nonBooleanAttribute($node)) { |
| 373 | $this->wr('="')->wr($val)->wr('"'); |
| 374 | } |
| 375 | } |
| 376 | } |
| 377 | |
| 378 | protected function nonBooleanAttribute(\DOMAttr $attr) |
| 379 | { |
| 380 | $ele = $attr->ownerElement; |
| 381 | foreach ($this->nonBooleanAttributes as $rule) { |
| 382 | if (isset($rule['nodeNamespace']) && $rule['nodeNamespace'] !== $ele->namespaceURI) { |
| 383 | continue; |
| 384 | } |
| 385 | if (isset($rule['attNamespace']) && $rule['attNamespace'] !== $attr->namespaceURI) { |
| 386 | continue; |
| 387 | } |
| 388 | if (isset($rule['nodeName']) && !is_array($rule['nodeName']) && $rule['nodeName'] !== $ele->localName) { |
| 389 | continue; |
| 390 | } |
| 391 | if (isset($rule['nodeName']) && is_array($rule['nodeName']) && !in_array($ele->localName, $rule['nodeName'], true)) { |
| 392 | continue; |
| 393 | } |
| 394 | if (isset($rule['attrName']) && !is_array($rule['attrName']) && $rule['attrName'] !== $attr->localName) { |
| 395 | continue; |
| 396 | } |
| 397 | if (isset($rule['attrName']) && is_array($rule['attrName']) && !in_array($attr->localName, $rule['attrName'], true)) { |
| 398 | continue; |
| 399 | } |
| 400 | if (isset($rule['xpath'])) { |
| 401 | $xp = $this->getXPath($attr); |
| 402 | if (isset($rule['prefixes'])) { |
| 403 | foreach ($rule['prefixes'] as $nsPrefix => $ns) { |
| 404 | $xp->registerNamespace($nsPrefix, $ns); |
| 405 | } |
| 406 | } |
| 407 | if (!$xp->evaluate($rule['xpath'], $attr)) { |
| 408 | continue; |
| 409 | } |
| 410 | } |
| 411 | |
| 412 | return true; |
| 413 | } |
| 414 | |
| 415 | return false; |
| 416 | } |
| 417 | |
| 418 | private function getXPath(\DOMNode $node) |
| 419 | { |
| 420 | if (!$this->xpath) { |
| 421 | $this->xpath = new \DOMXPath($node->ownerDocument); |
| 422 | } |
| 423 | |
| 424 | return $this->xpath; |
| 425 | } |
| 426 | |
| 427 | /** |
| 428 | * Write the closing tag. |
| 429 | * |
| 430 | * Tags for HTML, MathML, and SVG are in the local name. Otherwise, use the |
| 431 | * qualified name (8.3). |
| 432 | * |
| 433 | * @param \DOMNode $ele The element being written. |
| 434 | */ |
| 435 | protected function closeTag($ele) |
| 436 | { |
| 437 | if ($this->outputMode == static::IM_IN_HTML || $ele->hasChildNodes()) { |
| 438 | $this->wr('</')->wr($this->traverser->isLocalElement($ele) ? $ele->localName : $ele->tagName)->wr('>'); |
| 439 | } |
| 440 | } |
| 441 | |
| 442 | /** |
| 443 | * Write to the output. |
| 444 | * |
| 445 | * @param string $text The string to put into the output |
| 446 | * |
| 447 | * @return $this |
| 448 | */ |
| 449 | protected function wr($text) |
| 450 | { |
| 451 | fwrite($this->out, $text); |
| 452 | |
| 453 | return $this; |
| 454 | } |
| 455 | |
| 456 | /** |
| 457 | * Write a new line character. |
| 458 | * |
| 459 | * @return $this |
| 460 | */ |
| 461 | protected function nl() |
| 462 | { |
| 463 | fwrite($this->out, PHP_EOL); |
| 464 | |
| 465 | return $this; |
| 466 | } |
| 467 | |
| 468 | /** |
| 469 | * Encode text. |
| 470 | * |
| 471 | * When encode is set to false, the default value, the text passed in is |
| 472 | * escaped per section 8.3 of the html5 spec. For details on how text is |
| 473 | * escaped see the escape() method. |
| 474 | * |
| 475 | * When encoding is set to true the text is converted to named character |
| 476 | * references where appropriate. Section 8.1.4 Character references of the |
| 477 | * html5 spec refers to using named character references. This is useful for |
| 478 | * characters that can't otherwise legally be used in the text. |
| 479 | * |
| 480 | * The named character references are listed in section 8.5. |
| 481 | * |
| 482 | * @see http://www.w3.org/TR/2013/CR-html5-20130806/syntax.html#named-character-references True encoding will turn all named character references into their entities. |
| 483 | * This includes such characters as +.# and many other common ones. By default |
| 484 | * encoding here will just escape &'<>". |
| 485 | * |
| 486 | * Note, PHP 5.4+ has better html5 encoding. |
| 487 | * |
| 488 | * @todo Use the Entities class in php 5.3 to have html5 entities. |
| 489 | * |
| 490 | * @param string $text Text to encode. |
| 491 | * @param bool $attribute True if we are encoding an attrubute, false otherwise. |
| 492 | * |
| 493 | * @return string The encoded text. |
| 494 | */ |
| 495 | protected function enc($text, $attribute = false) |
| 496 | { |
| 497 | // Escape the text rather than convert to named character references. |
| 498 | if (!$this->encode) { |
| 499 | return $this->escape($text, $attribute); |
| 500 | } |
| 501 | |
| 502 | // If we are in PHP 5.4+ we can use the native html5 entity functionality to |
| 503 | // convert the named character references. |
| 504 | |
| 505 | if ($this->hasHTML5) { |
| 506 | return htmlentities($text, ENT_HTML5 | ENT_SUBSTITUTE | ENT_QUOTES, 'UTF-8', false); |
| 507 | } // If a version earlier than 5.4 html5 entities are not entirely handled. |
| 508 | // This manually handles them. |
| 509 | else { |
| 510 | return strtr($text, HTML5Entities::$map); |
| 511 | } |
| 512 | } |
| 513 | |
| 514 | /** |
| 515 | * Escape test. |
| 516 | * |
| 517 | * According to the html5 spec section 8.3 Serializing HTML fragments, text |
| 518 | * within tags that are not style, script, xmp, iframe, noembed, and noframes |
| 519 | * need to be properly escaped. |
| 520 | * |
| 521 | * The & should be converted to &, no breaking space unicode characters |
| 522 | * converted to , when in attribute mode the " should be converted to |
| 523 | * ", and when not in attribute mode the < and > should be converted to |
| 524 | * < and >. |
| 525 | * |
| 526 | * @see http://www.w3.org/TR/2013/CR-html5-20130806/syntax.html#escapingString |
| 527 | * |
| 528 | * @param string $text Text to escape. |
| 529 | * @param bool $attribute True if we are escaping an attrubute, false otherwise. |
| 530 | */ |
| 531 | protected function escape($text, $attribute = false) |
| 532 | { |
| 533 | // Not using htmlspecialchars because, while it does escaping, it doesn't |
| 534 | // match the requirements of section 8.5. For example, it doesn't handle |
| 535 | // non-breaking spaces. |
| 536 | if ($attribute) { |
| 537 | $replace = array( |
| 538 | '"' => '"', |
| 539 | '&' => '&', |
| 540 | "\xc2\xa0" => ' ', |
| 541 | ); |
| 542 | } else { |
| 543 | $replace = array( |
| 544 | '<' => '<', |
| 545 | '>' => '>', |
| 546 | '&' => '&', |
| 547 | "\xc2\xa0" => ' ', |
| 548 | ); |
| 549 | } |
| 550 | |
| 551 | return strtr($text, $replace); |
| 552 | } |
| 553 | } |
| 554 |