1<?php
2/**
3 * HTML API: WP_HTML_Tag_Processor class
4 *
5 * Scans through an HTML document to find specific tags, then
6 * transforms those tags by adding, removing, or updating the
7 * values of the HTML attributes within that tag (opener).
8 *
9 * Does not fully parse HTML or _recurse_ into the HTML structure
10 * Instead this scans linearly through a document and only parses
11 * the HTML tag openers.
12 *
13 * ### Possible future direction for this module
14 *
15 * - Prune the whitespace when removing classes/attributes: e.g. "a b c" -> "c" not " c".
16 * This would increase the size of the changes for some operations but leave more
17 * natural-looking output HTML.
18 *
19 * @package WordPress
20 * @subpackage HTML-API
21 * @since 6.2.0
22 */
23
24/**
25 * Core class used to modify attributes in an HTML document for tags matching a query.
26 *
27 * ## Usage
28 *
29 * Use of this class requires three steps:
30 *
31 * 1. Create a new class instance with your input HTML document.
32 * 2. Find the tag(s) you are looking for.
33 * 3. Request changes to the attributes in those tag(s).
34 *
35 * Example:
36 *
37 * $tags = new WP_HTML_Tag_Processor( $html );
38 * if ( $tags->next_tag( 'option' ) ) {
39 * $tags->set_attribute( 'selected', true );
40 * }
41 *
42 * ### Finding tags
43 *
44 * The `next_tag()` function moves the internal cursor through
45 * your input HTML document until it finds a tag meeting any of
46 * the supplied restrictions in the optional query argument. If
47 * no argument is provided then it will find the next HTML tag,
48 * regardless of what kind it is.
49 *
50 * If you want to _find whatever the next tag is_:
51 *
52 * $tags->next_tag();
53 *
54 * | Goal | Query |
55 * |-----------------------------------------------------------|---------------------------------------------------------------------------------|
56 * | Find any tag. | `$tags->next_tag();` |
57 * | Find next image tag. | `$tags->next_tag( array( 'tag_name' => 'img' ) );` |
58 * | Find next image tag (without passing the array). | `$tags->next_tag( 'img' );` |
59 * | Find next tag containing the `fullwidth` CSS class. | `$tags->next_tag( array( 'class_name' => 'fullwidth' ) );` |
60 * | Find next image tag containing the `fullwidth` CSS class. | `$tags->next_tag( array( 'tag_name' => 'img', 'class_name' => 'fullwidth' ) );` |
61 *
62 * If a tag was found meeting your criteria then `next_tag()`
63 * will return `true` and you can proceed to modify it. If it
64 * returns `false`, however, it failed to find the tag and
65 * moved the cursor to the end of the file.
66 *
67 * Once the cursor reaches the end of the file the processor
68 * is done and if you want to reach an earlier tag you will
69 * need to recreate the processor and start over, as it's
70 * unable to back up or move in reverse.
71 *
72 * See the section on bookmarks for an exception to this
73 * no-backing-up rule.
74 *
75 * #### Custom queries
76 *
77 * Sometimes it's necessary to further inspect an HTML tag than
78 * the query syntax here permits. In these cases one may further
79 * inspect the search results using the read-only functions
80 * provided by the processor or external state or variables.
81 *
82 * Example:
83 *
84 * // Paint up to the first five DIV or SPAN tags marked with the "jazzy" style.
85 * $remaining_count = 5;
86 * while ( $remaining_count > 0 && $tags->next_tag() ) {
87 * if (
88 * ( 'DIV' === $tags->get_tag() || 'SPAN' === $tags->get_tag() ) &&
89 * 'jazzy' === $tags->get_attribute( 'data-style' )
90 * ) {
91 * $tags->add_class( 'theme-style-everest-jazz' );
92 * $remaining_count--;
93 * }
94 * }
95 *
96 * `get_attribute()` will return `null` if the attribute wasn't present
97 * on the tag when it was called. It may return `""` (the empty string)
98 * in cases where the attribute was present but its value was empty.
99 * For boolean attributes, those whose name is present but no value is
100 * given, it will return `true` (the only way to set `false` for an
101 * attribute is to remove it).
102 *
103 * #### When matching fails
104 *
105 * When `next_tag()` returns `false` it could mean different things:
106 *
107 * - The requested tag wasn't found in the input document.
108 * - The input document ended in the middle of an HTML syntax element.
109 *
110 * When a document ends in the middle of a syntax element it will pause
111 * the processor. This is to make it possible in the future to extend the
112 * input document and proceed - an important requirement for chunked
113 * streaming parsing of a document.
114 *
115 * Example:
116 *
117 * $processor = new WP_HTML_Tag_Processor( 'This <div is="a" partial="token' );
118 * false === $processor->next_tag();
119 *
120 * If a special element (see next section) is encountered but no closing tag
121 * is found it will count as an incomplete tag. The parser will pause as if
122 * the opening tag were incomplete.
123 *
124 * Example:
125 *
126 * $processor = new WP_HTML_Tag_Processor( '<style>// there could be more styling to come' );
127 * false === $processor->next_tag();
128 *
129 * $processor = new WP_HTML_Tag_Processor( '<style>// this is everything</style><div>' );
130 * true === $processor->next_tag( 'DIV' );
131 *
132 * #### Special self-contained elements
133 *
134 * Some HTML elements are handled in a special way; their start and end tags
135 * act like a void tag. These are special because their contents can't contain
136 * HTML markup. Everything inside these elements is handled in a special way
137 * and content that _appears_ like HTML tags inside of them isn't. There can
138 * be no nesting in these elements.
139 *
140 * In the following list, "raw text" means that all of the content in the HTML
141 * until the matching closing tag is treated verbatim without any replacements
142 * and without any parsing.
143 *
144 * - IFRAME allows no content but requires a closing tag.
145 * - NOEMBED (deprecated) content is raw text.
146 * - NOFRAMES (deprecated) content is raw text.
147 * - SCRIPT content is plaintext apart from legacy rules allowing `</script>` inside an HTML comment.
148 * - STYLE content is raw text.
149 * - TITLE content is plain text but character references are decoded.
150 * - TEXTAREA content is plain text but character references are decoded.
151 * - XMP (deprecated) content is raw text.
152 *
153 * ### Modifying HTML attributes for a found tag
154 *
155 * Once you've found the start of an opening tag you can modify
156 * any number of the attributes on that tag. You can set a new
157 * value for an attribute, remove the entire attribute, or do
158 * nothing and move on to the next opening tag.
159 *
160 * Example:
161 *
162 * if ( $tags->next_tag( array( 'class_name' => 'wp-group-block' ) ) ) {
163 * $tags->set_attribute( 'title', 'This groups the contained content.' );
164 * $tags->remove_attribute( 'data-test-id' );
165 * }
166 *
167 * If `set_attribute()` is called for an existing attribute it will
168 * overwrite the existing value. Similarly, calling `remove_attribute()`
169 * for a non-existing attribute has no effect on the document. Both
170 * of these methods are safe to call without knowing if a given attribute
171 * exists beforehand.
172 *
173 * ### Modifying CSS classes for a found tag
174 *
175 * The tag processor treats the `class` attribute as a special case.
176 * Because it's a common operation to add or remove CSS classes, this
177 * interface adds helper methods to make that easier.
178 *
179 * As with attribute values, adding or removing CSS classes is a safe
180 * operation that doesn't require checking if the attribute or class
181 * exists before making changes. If removing the only class then the
182 * entire `class` attribute will be removed.
183 *
184 * Example:
185 *
186 * // from `<span>Yippee!</span>`
187 * // to `<span class="is-active">Yippee!</span>`
188 * $tags->add_class( 'is-active' );
189 *
190 * // from `<span class="excited">Yippee!</span>`
191 * // to `<span class="excited is-active">Yippee!</span>`
192 * $tags->add_class( 'is-active' );
193 *
194 * // from `<span class="is-active heavy-accent">Yippee!</span>`
195 * // to `<span class="is-active heavy-accent">Yippee!</span>`
196 * $tags->add_class( 'is-active' );
197 *
198 * // from `<input type="text" class="is-active rugby not-disabled" length="24">`
199 * // to `<input type="text" class="is-active not-disabled" length="24">
200 * $tags->remove_class( 'rugby' );
201 *
202 * // from `<input type="text" class="rugby" length="24">`
203 * // to `<input type="text" length="24">
204 * $tags->remove_class( 'rugby' );
205 *
206 * // from `<input type="text" length="24">`
207 * // to `<input type="text" length="24">
208 * $tags->remove_class( 'rugby' );
209 *
210 * When class changes are enqueued but a direct change to `class` is made via
211 * `set_attribute` then the changes to `set_attribute` (or `remove_attribute`)
212 * will take precedence over those made through `add_class` and `remove_class`.
213 *
214 * ### Bookmarks
215 *
216 * While scanning through the input HTMl document it's possible to set
217 * a named bookmark when a particular tag is found. Later on, after
218 * continuing to scan other tags, it's possible to `seek` to one of
219 * the set bookmarks and then proceed again from that point forward.
220 *
221 * Because bookmarks create processing overhead one should avoid
222 * creating too many of them. As a rule, create only bookmarks
223 * of known string literal names; avoid creating "mark_{$index}"
224 * and so on. It's fine from a performance standpoint to create a
225 * bookmark and update it frequently, such as within a loop.
226 *
227 * $total_todos = 0;
228 * while ( $p->next_tag( array( 'tag_name' => 'UL', 'class_name' => 'todo' ) ) ) {
229 * $p->set_bookmark( 'list-start' );
230 * while ( $p->next_tag( array( 'tag_closers' => 'visit' ) ) ) {
231 * if ( 'UL' === $p->get_tag() && $p->is_tag_closer() ) {
232 * $p->set_bookmark( 'list-end' );
233 * $p->seek( 'list-start' );
234 * $p->set_attribute( 'data-contained-todos', (string) $total_todos );
235 * $total_todos = 0;
236 * $p->seek( 'list-end' );
237 * break;
238 * }
239 *
240 * if ( 'LI' === $p->get_tag() && ! $p->is_tag_closer() ) {
241 * $total_todos++;
242 * }
243 * }
244 * }
245 *
246 * ## Tokens and finer-grained processing.
247 *
248 * It's possible to scan through every lexical token in the
249 * HTML document using the `next_token()` function. This
250 * alternative form takes no argument and provides no built-in
251 * query syntax.
252 *
253 * Example:
254 *
255 * $title = '(untitled)';
256 * $text = '';
257 * while ( $processor->next_token() ) {
258 * switch ( $processor->get_token_name() ) {
259 * case '#text':
260 * $text .= $processor->get_modifiable_text();
261 * break;
262 *
263 * case 'BR':
264 * $text .= "\n";
265 * break;
266 *
267 * case 'TITLE':
268 * $title = $processor->get_modifiable_text();
269 * break;
270 * }
271 * }
272 * return trim( "# {$title}\n\n{$text}" );
273 *
274 * ### Tokens and _modifiable text_.
275 *
276 * #### Special "atomic" HTML elements.
277 *
278 * Not all HTML elements are able to contain other elements inside of them.
279 * For instance, the contents inside a TITLE element are plaintext (except
280 * that character references like & will be decoded). This means that
281 * if the string `<img>` appears inside a TITLE element, then it's not an
282 * image tag, but rather it's text describing an image tag. Likewise, the
283 * contents of a SCRIPT or STYLE element are handled entirely separately in
284 * a browser than the contents of other elements because they represent a
285 * different language than HTML.
286 *
287 * For these elements the Tag Processor treats the entire sequence as one,
288 * from the opening tag, including its contents, through its closing tag.
289 * This means that the it's not possible to match the closing tag for a
290 * SCRIPT element unless it's unexpected; the Tag Processor already matched
291 * it when it found the opening tag.
292 *
293 * The inner contents of these elements are that element's _modifiable text_.
294 *
295 * The special elements are:
296 * - `SCRIPT` whose contents are treated as raw plaintext but supports a legacy
297 * style of including JavaScript inside of HTML comments to avoid accidentally
298 * closing the SCRIPT from inside a JavaScript string. E.g. `console.log( '</script>' )`.
299 * - `TITLE` and `TEXTAREA` whose contents are treated as plaintext and then any
300 * character references are decoded. E.g. `1 < 2 < 3` becomes `1 < 2 < 3`.
301 * - `IFRAME`, `NOSCRIPT`, `NOEMBED`, `NOFRAME`, `STYLE` whose contents are treated as
302 * raw plaintext and left as-is. E.g. `1 < 2 < 3` remains `1 < 2 < 3`.
303 *
304 * #### Other tokens with modifiable text.
305 *
306 * There are also non-elements which are void/self-closing in nature and contain
307 * modifiable text that is part of that individual syntax token itself.
308 *
309 * - `#text` nodes, whose entire token _is_ the modifiable text.
310 * - HTML comments and tokens that become comments due to some syntax error. The
311 * text for these tokens is the portion of the comment inside of the syntax.
312 * E.g. for `<!-- comment -->` the text is `" comment "` (note the spaces are included).
313 * - `CDATA` sections, whose text is the content inside of the section itself. E.g. for
314 * `<![CDATA[some content]]>` the text is `"some content"` (with restrictions [1]).
315 * - "Funky comments," which are a special case of invalid closing tags whose name is
316 * invalid. The text for these nodes is the text that a browser would transform into
317 * an HTML comment when parsing. E.g. for `</%post_author>` the text is `%post_author`.
318 * - `DOCTYPE` declarations like `<DOCTYPE html>` which have no closing tag.
319 * - XML Processing instruction nodes like `<?wp __( "Like" ); ?>` (with restrictions [2]).
320 * - The empty end tag `</>` which is ignored in the browser and DOM.
321 *
322 * [1]: There are no CDATA sections in HTML. When encountering `<![CDATA[`, everything
323 * until the next `>` becomes a bogus HTML comment, meaning there can be no CDATA
324 * section in an HTML document containing `>`. The Tag Processor will first find
325 * all valid and bogus HTML comments, and then if the comment _would_ have been a
326 * CDATA section _were they to exist_, it will indicate this as the type of comment.
327 *
328 * [2]: XML allows a broader range of characters in a processing instruction's target name
329 * and disallows "xml" as a name, since it's special. The Tag Processor only recognizes
330 * target names with an ASCII-representable subset of characters. It also exhibits the
331 * same constraint as with CDATA sections, in that `>` cannot exist within the token
332 * since Processing Instructions do no exist within HTML and their syntax transforms
333 * into a bogus comment in the DOM.
334 *
335 * ## Design and limitations
336 *
337 * The Tag Processor is designed to linearly scan HTML documents and tokenize
338 * HTML tags and their attributes. It's designed to do this as efficiently as
339 * possible without compromising parsing integrity. Therefore it will be
340 * slower than some methods of modifying HTML, such as those incorporating
341 * over-simplified PCRE patterns, but will not introduce the defects and
342 * failures that those methods bring in, which lead to broken page renders
343 * and often to security vulnerabilities. On the other hand, it will be faster
344 * than full-blown HTML parsers such as DOMDocument and use considerably
345 * less memory. It requires a negligible memory overhead, enough to consider
346 * it a zero-overhead system.
347 *
348 * The performance characteristics are maintained by avoiding tree construction
349 * and semantic cleanups which are specified in HTML5. Because of this, for
350 * example, it's not possible for the Tag Processor to associate any given
351 * opening tag with its corresponding closing tag, or to return the inner markup
352 * inside an element. Systems may be built on top of the Tag Processor to do
353 * this, but the Tag Processor is and should be constrained so it can remain an
354 * efficient, low-level, and reliable HTML scanner.
355 *
356 * The Tag Processor's design incorporates a "garbage-in-garbage-out" philosophy.
357 * HTML5 specifies that certain invalid content be transformed into different forms
358 * for display, such as removing null bytes from an input document and replacing
359 * invalid characters with the Unicode replacement character `U+FFFD` (visually "οΏ½").
360 * Where errors or transformations exist within the HTML5 specification, the Tag Processor
361 * leaves those invalid inputs untouched, passing them through to the final browser
362 * to handle. While this implies that certain operations will be non-spec-compliant,
363 * such as reading the value of an attribute with invalid content, it also preserves a
364 * simplicity and efficiency for handling those error cases.
365 *
366 * Most operations within the Tag Processor are designed to minimize the difference
367 * between an input and output document for any given change. For example, the
368 * `add_class` and `remove_class` methods preserve whitespace and the class ordering
369 * within the `class` attribute; and when encountering tags with duplicated attributes,
370 * the Tag Processor will leave those invalid duplicate attributes where they are but
371 * update the proper attribute which the browser will read for parsing its value. An
372 * exception to this rule is that all attribute updates store their values as
373 * double-quoted strings, meaning that attributes on input with single-quoted or
374 * unquoted values will appear in the output with double-quotes.
375 *
376 * ### Scripting Flag
377 *
378 * The Tag Processor parses HTML with the "scripting flag" disabled. This means
379 * that it doesn't run any scripts while parsing the page. In a browser with
380 * JavaScript enabled, for example, the script can change the parse of the
381 * document as it loads. On the server, however, evaluating JavaScript is not
382 * only impractical, but also unwanted.
383 *
384 * Practically this means that the Tag Processor will descend into NOSCRIPT
385 * elements and process its child tags. Were the scripting flag enabled, such
386 * as in a typical browser, the contents of NOSCRIPT are skipped entirely.
387 *
388 * This allows the HTML API to process the content that will be presented in
389 * a browser when scripting is disabled, but it offers a different view of a
390 * page than most browser sessions will experience. E.g. the tags inside the
391 * NOSCRIPT disappear.
392 *
393 * ### Text Encoding
394 *
395 * The Tag Processor assumes that the input HTML document is encoded with a
396 * text encoding compatible with 7-bit ASCII's '<', '>', '&', ';', '/', '=',
397 * "'", '"', 'a' - 'z', 'A' - 'Z', and the whitespace characters ' ', tab,
398 * carriage-return, newline, and form-feed.
399 *
400 * In practice, this includes almost every single-byte encoding as well as
401 * UTF-8. Notably, however, it does not include UTF-16. If providing input
402 * that's incompatible, then convert the encoding beforehand.
403 *
404 * @since 6.2.0
405 * @since 6.2.1 Fix: Support for various invalid comments; attribute updates are case-insensitive.
406 * @since 6.3.2 Fix: Skip HTML-like content inside rawtext elements such as STYLE.
407 * @since 6.5.0 Pauses processor when input ends in an incomplete syntax token.
408 * Introduces "special" elements which act like void elements, e.g. TITLE, STYLE.
409 * Allows scanning through all tokens and processing modifiable text, where applicable.
410 */
411class WP_HTML_Tag_Processor {
412 /**
413 * The maximum number of bookmarks allowed to exist at
414 * any given time.
415 *
416 * @since 6.2.0
417 * @var int
418 *
419 * @see WP_HTML_Tag_Processor::set_bookmark()
420 */
421 const MAX_BOOKMARKS = 10;
422
423 /**
424 * Maximum number of times seek() can be called.
425 * Prevents accidental infinite loops.
426 *
427 * @since 6.2.0
428 * @var int
429 *
430 * @see WP_HTML_Tag_Processor::seek()
431 */
432 const MAX_SEEK_OPS = 1000;
433
434 /**
435 * The HTML document to parse.
436 *
437 * @since 6.2.0
438 * @var string
439 */
440 protected $html;
441
442 /**
443 * The last query passed to next_tag().
444 *
445 * @since 6.2.0
446 * @var array|null
447 */
448 private $last_query;
449
450 /**
451 * The tag name this processor currently scans for.
452 *
453 * @since 6.2.0
454 * @var string|null
455 */
456 private $sought_tag_name;
457
458 /**
459 * The CSS class name this processor currently scans for.
460 *
461 * @since 6.2.0
462 * @var string|null
463 */
464 private $sought_class_name;
465
466 /**
467 * The match offset this processor currently scans for.
468 *
469 * @since 6.2.0
470 * @var int|null
471 */
472 private $sought_match_offset;
473
474 /**
475 * Whether to visit tag closers, e.g. </div>, when walking an input document.
476 *
477 * @since 6.2.0
478 * @var bool
479 */
480 private $stop_on_tag_closers;
481
482 /**
483 * Specifies mode of operation of the parser at any given time.
484 *
485 * | State | Meaning |
486 * | ----------------|----------------------------------------------------------------------|
487 * | *Ready* | The parser is ready to run. |
488 * | *Complete* | There is nothing left to parse. |
489 * | *Incomplete* | The HTML ended in the middle of a token; nothing more can be parsed. |
490 * | *Matched tag* | Found an HTML tag; it's possible to modify its attributes. |
491 * | *Text node* | Found a #text node; this is plaintext and modifiable. |
492 * | *CDATA node* | Found a CDATA section; this is modifiable. |
493 * | *Comment* | Found a comment or bogus comment; this is modifiable. |
494 * | *Presumptuous* | Found an empty tag closer: `</>`. |
495 * | *Funky comment* | Found a tag closer with an invalid tag name; this is modifiable. |
496 *
497 * @since 6.5.0
498 *
499 * @see WP_HTML_Tag_Processor::STATE_READY
500 * @see WP_HTML_Tag_Processor::STATE_COMPLETE
501 * @see WP_HTML_Tag_Processor::STATE_INCOMPLETE_INPUT
502 * @see WP_HTML_Tag_Processor::STATE_MATCHED_TAG
503 * @see WP_HTML_Tag_Processor::STATE_TEXT_NODE
504 * @see WP_HTML_Tag_Processor::STATE_CDATA_NODE
505 * @see WP_HTML_Tag_Processor::STATE_COMMENT
506 * @see WP_HTML_Tag_Processor::STATE_DOCTYPE
507 * @see WP_HTML_Tag_Processor::STATE_PRESUMPTUOUS_TAG
508 * @see WP_HTML_Tag_Processor::STATE_FUNKY_COMMENT
509 *
510 * @var string
511 */
512 protected $parser_state = self::STATE_READY;
513
514 /**
515 * Indicates if the document is in quirks mode or no-quirks mode.
516 *
517 * Impact on HTML parsing:
518 *
519 * - In `NO_QUIRKS_MODE` (also known as "standard mode"):
520 * - CSS class and ID selectors match byte-for-byte (case-sensitively).
521 * - A TABLE start tag `<table>` implicitly closes any open `P` element.
522 *
523 * - In `QUIRKS_MODE`:
524 * - CSS class and ID selectors match match in an ASCII case-insensitive manner.
525 * - A TABLE start tag `<table>` opens a `TABLE` element as a child of a `P`
526 * element if one is open.
527 *
528 * Quirks and no-quirks mode are thus mostly about styling, but have an impact when
529 * tables are found inside paragraph elements.
530 *
531 * @see self::QUIRKS_MODE
532 * @see self::NO_QUIRKS_MODE
533 *
534 * @since 6.7.0
535 *
536 * @var string
537 */
538 protected $compat_mode = self::NO_QUIRKS_MODE;
539
540 /**
541 * Indicates whether the parser is inside foreign content,
542 * e.g. inside an SVG or MathML element.
543 *
544 * One of 'html', 'svg', or 'math'.
545 *
546 * Several parsing rules change based on whether the parser
547 * is inside foreign content, including whether CDATA sections
548 * are allowed and whether a self-closing flag indicates that
549 * an element has no content.
550 *
551 * @since 6.7.0
552 *
553 * @var string
554 */
555 private $parsing_namespace = 'html';
556
557 /**
558 * What kind of syntax token became an HTML comment.
559 *
560 * Since there are many ways in which HTML syntax can create an HTML comment,
561 * this indicates which of those caused it. This allows the Tag Processor to
562 * represent more from the original input document than would appear in the DOM.
563 *
564 * @since 6.5.0
565 *
566 * @var string|null
567 */
568 protected $comment_type = null;
569
570 /**
571 * What kind of text the matched text node represents, if it was subdivided.
572 *
573 * @see self::TEXT_IS_NULL_SEQUENCE
574 * @see self::TEXT_IS_WHITESPACE
575 * @see self::TEXT_IS_GENERIC
576 * @see self::subdivide_text_appropriately
577 *
578 * @since 6.7.0
579 *
580 * @var string
581 */
582 protected $text_node_classification = self::TEXT_IS_GENERIC;
583
584 /**
585 * How many bytes from the original HTML document have been read and parsed.
586 *
587 * This value points to the latest byte offset in the input document which
588 * has been already parsed. It is the internal cursor for the Tag Processor
589 * and updates while scanning through the HTML tokens.
590 *
591 * @since 6.2.0
592 * @var int
593 */
594 private $bytes_already_parsed = 0;
595
596 /**
597 * Byte offset in input document where current token starts.
598 *
599 * Example:
600 *
601 * <div id="test">...
602 * 01234
603 * - token starts at 0
604 *
605 * @since 6.5.0
606 *
607 * @var int|null
608 */
609 private $token_starts_at;
610
611 /**
612 * Byte length of current token.
613 *
614 * Example:
615 *
616 * <div id="test">...
617 * 012345678901234
618 * - token length is 14 - 0 = 14
619 *
620 * a <!-- comment --> is a token.
621 * 0123456789 123456789 123456789
622 * - token length is 17 - 2 = 15
623 *
624 * @since 6.5.0
625 *
626 * @var int|null
627 */
628 private $token_length;
629
630 /**
631 * Byte offset in input document where current tag name starts.
632 *
633 * Example:
634 *
635 * <div id="test">...
636 * 01234
637 * - tag name starts at 1
638 *
639 * @since 6.2.0
640 *
641 * @var int|null
642 */
643 private $tag_name_starts_at;
644
645 /**
646 * Byte length of current tag name.
647 *
648 * Example:
649 *
650 * <div id="test">...
651 * 01234
652 * --- tag name length is 3
653 *
654 * @since 6.2.0
655 *
656 * @var int|null
657 */
658 private $tag_name_length;
659
660 /**
661 * Byte offset into input document where current modifiable text starts.
662 *
663 * @since 6.5.0
664 *
665 * @var int
666 */
667 private $text_starts_at;
668
669 /**
670 * Byte length of modifiable text.
671 *
672 * @since 6.5.0
673 *
674 * @var int
675 */
676 private $text_length;
677
678 /**
679 * Whether the current tag is an opening tag, e.g. <div>, or a closing tag, e.g. </div>.
680 *
681 * @var bool
682 */
683 private $is_closing_tag;
684
685 /**
686 * Lazily-built index of attributes found within an HTML tag, keyed by the attribute name.
687 *
688 * Example:
689 *
690 * // Supposing the parser is working through this content
691 * // and stops after recognizing the `id` attribute.
692 * // <div id="test-4" class=outline title="data:text/plain;base64=asdk3nk1j3fo8">
693 * // ^ parsing will continue from this point.
694 * $this->attributes = array(
695 * 'id' => new WP_HTML_Attribute_Token( 'id', 9, 6, 5, 11, false )
696 * );
697 *
698 * // When picking up parsing again, or when asking to find the
699 * // `class` attribute we will continue and add to this array.
700 * $this->attributes = array(
701 * 'id' => new WP_HTML_Attribute_Token( 'id', 9, 6, 5, 11, false ),
702 * 'class' => new WP_HTML_Attribute_Token( 'class', 23, 7, 17, 13, false )
703 * );
704 *
705 * // Note that only the `class` attribute value is stored in the index.
706 * // That's because it is the only value used by this class at the moment.
707 *
708 * @since 6.2.0
709 * @var WP_HTML_Attribute_Token[]
710 */
711 private $attributes = array();
712
713 /**
714 * Tracks spans of duplicate attributes on a given tag, used for removing
715 * all copies of an attribute when calling `remove_attribute()`.
716 *
717 * @since 6.3.2
718 *
719 * @var (WP_HTML_Span[])[]|null
720 */
721 private $duplicate_attributes = null;
722
723 /**
724 * Which class names to add or remove from a tag.
725 *
726 * These are tracked separately from attribute updates because they are
727 * semantically distinct, whereas this interface exists for the common
728 * case of adding and removing class names while other attributes are
729 * generally modified as with DOM `setAttribute` calls.
730 *
731 * When modifying an HTML document these will eventually be collapsed
732 * into a single `set_attribute( 'class', $changes )` call.
733 *
734 * Example:
735 *
736 * // Add the `wp-block-group` class, remove the `wp-group` class.
737 * $classname_updates = array(
738 * // Indexed by a comparable class name.
739 * 'wp-block-group' => WP_HTML_Tag_Processor::ADD_CLASS,
740 * 'wp-group' => WP_HTML_Tag_Processor::REMOVE_CLASS
741 * );
742 *
743 * @since 6.2.0
744 * @var bool[]
745 */
746 private $classname_updates = array();
747
748 /**
749 * Tracks a semantic location in the original HTML which
750 * shifts with updates as they are applied to the document.
751 *
752 * @since 6.2.0
753 * @var WP_HTML_Span[]
754 */
755 protected $bookmarks = array();
756
757 const ADD_CLASS = true;
758 const REMOVE_CLASS = false;
759 const SKIP_CLASS = null;
760
761 /**
762 * Lexical replacements to apply to input HTML document.
763 *
764 * "Lexical" in this class refers to the part of this class which
765 * operates on pure text _as text_ and not as HTML. There's a line
766 * between the public interface, with HTML-semantic methods like
767 * `set_attribute` and `add_class`, and an internal state that tracks
768 * text offsets in the input document.
769 *
770 * When higher-level HTML methods are called, those have to transform their
771 * operations (such as setting an attribute's value) into text diffing
772 * operations (such as replacing the sub-string from indices A to B with
773 * some given new string). These text-diffing operations are the lexical
774 * updates.
775 *
776 * As new higher-level methods are added they need to collapse their
777 * operations into these lower-level lexical updates since that's the
778 * Tag Processor's internal language of change. Any code which creates
779 * these lexical updates must ensure that they do not cross HTML syntax
780 * boundaries, however, so these should never be exposed outside of this
781 * class or any classes which intentionally expand its functionality.
782 *
783 * These are enqueued while editing the document instead of being immediately
784 * applied to avoid processing overhead, string allocations, and string
785 * copies when applying many updates to a single document.
786 *
787 * Example:
788 *
789 * // Replace an attribute stored with a new value, indices
790 * // sourced from the lazily-parsed HTML recognizer.
791 * $start = $attributes['src']->start;
792 * $length = $attributes['src']->length;
793 * $modifications[] = new WP_HTML_Text_Replacement( $start, $length, $new_value );
794 *
795 * // Correspondingly, something like this will appear in this array.
796 * $lexical_updates = array(
797 * WP_HTML_Text_Replacement( 14, 28, 'https://my-site.my-domain/wp-content/uploads/2014/08/kittens.jpg' )
798 * );
799 *
800 * @since 6.2.0
801 * @var WP_HTML_Text_Replacement[]
802 */
803 protected $lexical_updates = array();
804
805 /**
806 * Tracks and limits `seek()` calls to prevent accidental infinite loops.
807 *
808 * @since 6.2.0
809 * @var int
810 *
811 * @see WP_HTML_Tag_Processor::seek()
812 */
813 protected $seek_count = 0;
814
815 /**
816 * Whether the parser should skip over an immediately-following linefeed
817 * character, as is the case with LISTING, PRE, and TEXTAREA.
818 *
819 * > If the next token is a U+000A LINE FEED (LF) character token, then
820 * > ignore that token and move on to the next one. (Newlines at the start
821 * > of [these] elements are ignored as an authoring convenience.)
822 *
823 * @since 6.7.0
824 *
825 * @var int|null
826 */
827 private $skip_newline_at = null;
828
829 /**
830 * Constructor.
831 *
832 * @since 6.2.0
833 *
834 * @param string $html HTML to process.
835 */
836 public function __construct( $html ) {
837 if ( ! is_string( $html ) ) {
838 _doing_it_wrong(
839 __METHOD__,
840 __( 'The HTML parameter must be a string.' ),
841 '6.9.0'
842 );
843 $html = '';
844 }
845 $this->html = $html;
846 }
847
848 /**
849 * Switches parsing mode into a new namespace, such as when
850 * encountering an SVG tag and entering foreign content.
851 *
852 * @since 6.7.0
853 *
854 * @param string $new_namespace One of 'html', 'svg', or 'math' indicating into what
855 * namespace the next tokens will be processed.
856 * @return bool Whether the namespace was valid and changed.
857 */
858 public function change_parsing_namespace( string $new_namespace ): bool {
859 if ( ! in_array( $new_namespace, array( 'html', 'math', 'svg' ), true ) ) {
860 return false;
861 }
862
863 $this->parsing_namespace = $new_namespace;
864 return true;
865 }
866
867 /**
868 * Finds the next tag matching the $query.
869 *
870 * @since 6.2.0
871 * @since 6.5.0 No longer processes incomplete tokens at end of document; pauses the processor at start of token.
872 *
873 * @param array|string|null $query {
874 * Optional. Which tag name to find, having which class, etc. Default is to find any tag.
875 *
876 * @type string|null $tag_name Which tag to find, or `null` for "any tag."
877 * @type int|null $match_offset Find the Nth tag matching all search criteria.
878 * 1 for "first" tag, 3 for "third," etc.
879 * Defaults to first tag.
880 * @type string|null $class_name Tag must contain this whole class name to match.
881 * @type string|null $tag_closers "visit" or "skip": whether to stop on tag closers, e.g. </div>.
882 * }
883 * @return bool Whether a tag was matched.
884 */
885 public function next_tag( $query = null ): bool {
886 $this->parse_query( $query );
887 $already_found = 0;
888
889 do {
890 if ( false === $this->next_token() ) {
891 return false;
892 }
893
894 if ( self::STATE_MATCHED_TAG !== $this->parser_state ) {
895 continue;
896 }
897
898 if ( $this->matches() ) {
899 ++$already_found;
900 }
901 } while ( $already_found < $this->sought_match_offset );
902
903 return true;
904 }
905
906 /**
907 * Finds the next token in the HTML document.
908 *
909 * An HTML document can be viewed as a stream of tokens,
910 * where tokens are things like HTML tags, HTML comments,
911 * text nodes, etc. This method finds the next token in
912 * the HTML document and returns whether it found one.
913 *
914 * If it starts parsing a token and reaches the end of the
915 * document then it will seek to the start of the last
916 * token and pause, returning `false` to indicate that it
917 * failed to find a complete token.
918 *
919 * Possible token types, based on the HTML specification:
920 *
921 * - an HTML tag, whether opening, closing, or void.
922 * - a text node - the plaintext inside tags.
923 * - an HTML comment.
924 * - a DOCTYPE declaration.
925 * - a processing instruction, e.g. `<?xml version="1.0" ?>`.
926 *
927 * The Tag Processor currently only supports the tag token.
928 *
929 * @since 6.5.0
930 * @since 6.7.0 Recognizes CDATA sections within foreign content.
931 *
932 * @return bool Whether a token was parsed.
933 */
934 public function next_token(): bool {
935 return $this->base_class_next_token();
936 }
937
938 /**
939 * Internal method which finds the next token in the HTML document.
940 *
941 * This method is a protected internal function which implements the logic for
942 * finding the next token in a document. It exists so that the parser can update
943 * its state without affecting the location of the cursor in the document and
944 * without triggering subclass methods for things like `next_token()`, e.g. when
945 * applying patches before searching for the next token.
946 *
947 * @since 6.5.0
948 *
949 * @access private
950 *
951 * @return bool Whether a token was parsed.
952 */
953 private function base_class_next_token(): bool {
954 $was_at = $this->bytes_already_parsed;
955 $this->after_tag();
956
957 // Don't proceed if there's nothing more to scan.
958 if (
959 self::STATE_COMPLETE === $this->parser_state ||
960 self::STATE_INCOMPLETE_INPUT === $this->parser_state
961 ) {
962 return false;
963 }
964
965 /*
966 * The next step in the parsing loop determines the parsing state;
967 * clear it so that state doesn't linger from the previous step.
968 */
969 $this->parser_state = self::STATE_READY;
970
971 if ( $this->bytes_already_parsed >= strlen( $this->html ) ) {
972 $this->parser_state = self::STATE_COMPLETE;
973 return false;
974 }
975
976 // Find the next tag if it exists.
977 if ( false === $this->parse_next_tag() ) {
978 if ( self::STATE_INCOMPLETE_INPUT === $this->parser_state ) {
979 $this->bytes_already_parsed = $was_at;
980 }
981
982 return false;
983 }
984
985 /*
986 * For legacy reasons the rest of this function handles tags and their
987 * attributes. If the processor has reached the end of the document
988 * or if it matched any other token then it should return here to avoid
989 * attempting to process tag-specific syntax.
990 */
991 if (
992 self::STATE_INCOMPLETE_INPUT !== $this->parser_state &&
993 self::STATE_COMPLETE !== $this->parser_state &&
994 self::STATE_MATCHED_TAG !== $this->parser_state
995 ) {
996 return true;
997 }
998
999 // Parse all of its attributes.
1000 while ( $this->parse_next_attribute() ) {
1001 continue;
1002 }
1003
1004 // Ensure that the tag closes before the end of the document.
1005 if (
1006 self::STATE_INCOMPLETE_INPUT === $this->parser_state ||
1007 $this->bytes_already_parsed >= strlen( $this->html )
1008 ) {
1009 // Does this appropriately clear state (parsed attributes)?
1010 $this->parser_state = self::STATE_INCOMPLETE_INPUT;
1011 $this->bytes_already_parsed = $was_at;
1012
1013 return false;
1014 }
1015
1016 $tag_ends_at = strpos( $this->html, '>', $this->bytes_already_parsed );
1017 if ( false === $tag_ends_at ) {
1018 $this->parser_state = self::STATE_INCOMPLETE_INPUT;
1019 $this->bytes_already_parsed = $was_at;
1020
1021 return false;
1022 }
1023 $this->parser_state = self::STATE_MATCHED_TAG;
1024 $this->bytes_already_parsed = $tag_ends_at + 1;
1025 $this->token_length = $this->bytes_already_parsed - $this->token_starts_at;
1026
1027 /*
1028 * Certain tags require additional processing. The first-letter pre-check
1029 * avoids unnecessary string allocation when comparing the tag names.
1030 *
1031 * - IFRAME
1032 * - LISTING (deprecated)
1033 * - NOEMBED (deprecated)
1034 * - NOFRAMES (deprecated)
1035 * - PRE
1036 * - SCRIPT
1037 * - STYLE
1038 * - TEXTAREA
1039 * - TITLE
1040 * - XMP (deprecated)
1041 */
1042 if (
1043 $this->is_closing_tag ||
1044 'html' !== $this->parsing_namespace ||
1045 1 !== strspn( $this->html, 'iIlLnNpPsStTxX', $this->tag_name_starts_at, 1 )
1046 ) {
1047 return true;
1048 }
1049
1050 $tag_name = $this->get_tag();
1051
1052 /*
1053 * For LISTING, PRE, and TEXTAREA, the first linefeed of an immediately-following
1054 * text node is ignored as an authoring convenience.
1055 *
1056 * @see static::skip_newline_at
1057 */
1058 if ( 'LISTING' === $tag_name || 'PRE' === $tag_name ) {
1059 $this->skip_newline_at = $this->bytes_already_parsed;
1060 return true;
1061 }
1062
1063 /*
1064 * There are certain elements whose children are not DATA but are instead
1065 * RCDATA or RAWTEXT. These cannot contain other elements, and the contents
1066 * are parsed as plaintext, with character references decoded in RCDATA but
1067 * not in RAWTEXT.
1068 *
1069 * These elements are described here as "self-contained" or special atomic
1070 * elements whose end tag is consumed with the opening tag, and they will
1071 * contain modifiable text inside of them.
1072 *
1073 * Preserve the opening tag pointers, as these will be overwritten
1074 * when finding the closing tag. They will be reset after finding
1075 * the closing to tag to point to the opening of the special atomic
1076 * tag sequence.
1077 */
1078 $tag_name_starts_at = $this->tag_name_starts_at;
1079 $tag_name_length = $this->tag_name_length;
1080 $tag_ends_at = $this->token_starts_at + $this->token_length;
1081 $attributes = $this->attributes;
1082 $duplicate_attributes = $this->duplicate_attributes;
1083
1084 // Find the closing tag if necessary.
1085 switch ( $tag_name ) {
1086 case 'SCRIPT':
1087 $found_closer = $this->skip_script_data();
1088 break;
1089
1090 case 'TEXTAREA':
1091 case 'TITLE':
1092 $found_closer = $this->skip_rcdata( $tag_name );
1093 break;
1094
1095 /*
1096 * In the browser this list would include the NOSCRIPT element,
1097 * but the Tag Processor is an environment with the scripting
1098 * flag disabled, meaning that it needs to descend into the
1099 * NOSCRIPT element to be able to properly process what will be
1100 * sent to a browser.
1101 *
1102 * Note that this rule makes HTML5 syntax incompatible with XML,
1103 * because the parsing of this token depends on client application.
1104 * The NOSCRIPT element cannot be represented in the XHTML syntax.
1105 */
1106 case 'IFRAME':
1107 case 'NOEMBED':
1108 case 'NOFRAMES':
1109 case 'STYLE':
1110 case 'XMP':
1111 $found_closer = $this->skip_rawtext( $tag_name );
1112 break;
1113
1114 // No other tags should be treated in their entirety here.
1115 default:
1116 return true;
1117 }
1118
1119 if ( ! $found_closer ) {
1120 $this->parser_state = self::STATE_INCOMPLETE_INPUT;
1121 $this->bytes_already_parsed = $was_at;
1122 return false;
1123 }
1124
1125 /*
1126 * The values here look like they reference the opening tag but they reference
1127 * the closing tag instead. This is why the opening tag values were stored
1128 * above in a variable. It reads confusingly here, but that's because the
1129 * functions that skip the contents have moved all the internal cursors past
1130 * the inner content of the tag.
1131 */
1132 $this->token_starts_at = $was_at;
1133 $this->token_length = $this->bytes_already_parsed - $this->token_starts_at;
1134 $this->text_starts_at = $tag_ends_at;
1135 $this->text_length = $this->tag_name_starts_at - $this->text_starts_at;
1136 $this->tag_name_starts_at = $tag_name_starts_at;
1137 $this->tag_name_length = $tag_name_length;
1138 $this->attributes = $attributes;
1139 $this->duplicate_attributes = $duplicate_attributes;
1140
1141 return true;
1142 }
1143
1144 /**
1145 * Whether the processor paused because the input HTML document ended
1146 * in the middle of a syntax element, such as in the middle of a tag.
1147 *
1148 * Example:
1149 *
1150 * $processor = new WP_HTML_Tag_Processor( '<input type="text" value="Th' );
1151 * false === $processor->get_next_tag();
1152 * true === $processor->paused_at_incomplete_token();
1153 *
1154 * @since 6.5.0
1155 *
1156 * @return bool Whether the parse paused at the start of an incomplete token.
1157 */
1158 public function paused_at_incomplete_token(): bool {
1159 return self::STATE_INCOMPLETE_INPUT === $this->parser_state;
1160 }
1161
1162 /**
1163 * Generator for a foreach loop to step through each class name for the matched tag.
1164 *
1165 * This generator function is designed to be used inside a "foreach" loop.
1166 *
1167 * Example:
1168 *
1169 * $p = new WP_HTML_Tag_Processor( "<div class='free <egg<\tlang-en'>" );
1170 * $p->next_tag();
1171 * foreach ( $p->class_list() as $class_name ) {
1172 * echo "{$class_name} ";
1173 * }
1174 * // Outputs: "free <egg> lang-en "
1175 *
1176 * @since 6.4.0
1177 */
1178 public function class_list() {
1179 if ( self::STATE_MATCHED_TAG !== $this->parser_state ) {
1180 return;
1181 }
1182
1183 /** @var string $class contains the string value of the class attribute, with character references decoded. */
1184 $class = $this->get_attribute( 'class' );
1185
1186 if ( ! is_string( $class ) ) {
1187 return;
1188 }
1189
1190 $seen = array();
1191
1192 $is_quirks = self::QUIRKS_MODE === $this->compat_mode;
1193
1194 $at = 0;
1195 while ( $at < strlen( $class ) ) {
1196 // Skip past any initial boundary characters.
1197 $at += strspn( $class, " \t\f\r\n", $at );
1198 if ( $at >= strlen( $class ) ) {
1199 return;
1200 }
1201
1202 // Find the byte length until the next boundary.
1203 $length = strcspn( $class, " \t\f\r\n", $at );
1204 if ( 0 === $length ) {
1205 return;
1206 }
1207
1208 $name = str_replace( "\x00", "\u{FFFD}", substr( $class, $at, $length ) );
1209 if ( $is_quirks ) {
1210 $name = strtolower( $name );
1211 }
1212 $at += $length;
1213
1214 /*
1215 * It's expected that the number of class names for a given tag is relatively small.
1216 * Given this, it is probably faster overall to scan an array for a value rather
1217 * than to use the class name as a key and check if it's a key of $seen.
1218 */
1219 if ( in_array( $name, $seen, true ) ) {
1220 continue;
1221 }
1222
1223 $seen[] = $name;
1224 yield $name;
1225 }
1226 }
1227
1228
1229 /**
1230 * Returns if a matched tag contains the given ASCII case-insensitive class name.
1231 *
1232 * @since 6.4.0
1233 *
1234 * @param string $wanted_class Look for this CSS class name, ASCII case-insensitive.
1235 * @return bool|null Whether the matched tag contains the given class name, or null if not matched.
1236 */
1237 public function has_class( $wanted_class ): ?bool {
1238 if ( self::STATE_MATCHED_TAG !== $this->parser_state ) {
1239 return null;
1240 }
1241
1242 $case_insensitive = self::QUIRKS_MODE === $this->compat_mode;
1243
1244 $wanted_length = strlen( $wanted_class );
1245 foreach ( $this->class_list() as $class_name ) {
1246 if (
1247 strlen( $class_name ) === $wanted_length &&
1248 0 === substr_compare( $class_name, $wanted_class, 0, strlen( $wanted_class ), $case_insensitive )
1249 ) {
1250 return true;
1251 }
1252 }
1253
1254 return false;
1255 }
1256
1257
1258 /**
1259 * Sets a bookmark in the HTML document.
1260 *
1261 * Bookmarks represent specific places or tokens in the HTML
1262 * document, such as a tag opener or closer. When applying
1263 * edits to a document, such as setting an attribute, the
1264 * text offsets of that token may shift; the bookmark is
1265 * kept updated with those shifts and remains stable unless
1266 * the entire span of text in which the token sits is removed.
1267 *
1268 * Release bookmarks when they are no longer needed.
1269 *
1270 * Example:
1271 *
1272 * <main><h2>Surprising fact you may not know!</h2></main>
1273 * ^ ^
1274 * \-|-- this `H2` opener bookmark tracks the token
1275 *
1276 * <main class="clickbait"><h2>Surprising fact you may noβ¦
1277 * ^ ^
1278 * \-|-- it shifts with edits
1279 *
1280 * Bookmarks provide the ability to seek to a previously-scanned
1281 * place in the HTML document. This avoids the need to re-scan
1282 * the entire document.
1283 *
1284 * Example:
1285 *
1286 * <ul><li>One</li><li>Two</li><li>Three</li></ul>
1287 * ^^^^
1288 * want to note this last item
1289 *
1290 * $p = new WP_HTML_Tag_Processor( $html );
1291 * $in_list = false;
1292 * while ( $p->next_tag( array( 'tag_closers' => $in_list ? 'visit' : 'skip' ) ) ) {
1293 * if ( 'UL' === $p->get_tag() ) {
1294 * if ( $p->is_tag_closer() ) {
1295 * $in_list = false;
1296 * $p->set_bookmark( 'resume' );
1297 * if ( $p->seek( 'last-li' ) ) {
1298 * $p->add_class( 'last-li' );
1299 * }
1300 * $p->seek( 'resume' );
1301 * $p->release_bookmark( 'last-li' );
1302 * $p->release_bookmark( 'resume' );
1303 * } else {
1304 * $in_list = true;
1305 * }
1306 * }
1307 *
1308 * if ( 'LI' === $p->get_tag() ) {
1309 * $p->set_bookmark( 'last-li' );
1310 * }
1311 * }
1312 *
1313 * Bookmarks intentionally hide the internal string offsets
1314 * to which they refer. They are maintained internally as
1315 * updates are applied to the HTML document and therefore
1316 * retain their "position" - the location to which they
1317 * originally pointed. The inability to use bookmarks with
1318 * functions like `substr` is therefore intentional to guard
1319 * against accidentally breaking the HTML.
1320 *
1321 * Because bookmarks allocate memory and require processing
1322 * for every applied update, they are limited and require
1323 * a name. They should not be created with programmatically-made
1324 * names, such as "li_{$index}" with some loop. As a general
1325 * rule they should only be created with string-literal names
1326 * like "start-of-section" or "last-paragraph".
1327 *
1328 * Bookmarks are a powerful tool to enable complicated behavior.
1329 * Consider double-checking that you need this tool if you are
1330 * reaching for it, as inappropriate use could lead to broken
1331 * HTML structure or unwanted processing overhead.
1332 *
1333 * @since 6.2.0
1334 *
1335 * @param string $name Identifies this particular bookmark.
1336 * @return bool Whether the bookmark was successfully created.
1337 */
1338 public function set_bookmark( $name ): bool {
1339 // It only makes sense to set a bookmark if the parser has paused on a concrete token.
1340 if (
1341 self::STATE_COMPLETE === $this->parser_state ||
1342 self::STATE_INCOMPLETE_INPUT === $this->parser_state
1343 ) {
1344 return false;
1345 }
1346
1347 if ( ! array_key_exists( $name, $this->bookmarks ) && count( $this->bookmarks ) >= static::MAX_BOOKMARKS ) {
1348 _doing_it_wrong(
1349 __METHOD__,
1350 __( 'Too many bookmarks: cannot create any more.' ),
1351 '6.2.0'
1352 );
1353 return false;
1354 }
1355
1356 $this->bookmarks[ $name ] = new WP_HTML_Span( $this->token_starts_at, $this->token_length );
1357
1358 return true;
1359 }
1360
1361
1362 /**
1363 * Removes a bookmark that is no longer needed.
1364 *
1365 * Releasing a bookmark frees up the small
1366 * performance overhead it requires.
1367 *
1368 * @param string $name Name of the bookmark to remove.
1369 * @return bool Whether the bookmark already existed before removal.
1370 */
1371 public function release_bookmark( $name ): bool {
1372 if ( ! array_key_exists( $name, $this->bookmarks ) ) {
1373 return false;
1374 }
1375
1376 unset( $this->bookmarks[ $name ] );
1377
1378 return true;
1379 }
1380
1381 /**
1382 * Skips contents of generic rawtext elements.
1383 *
1384 * @since 6.3.2
1385 *
1386 * @see https://html.spec.whatwg.org/#generic-raw-text-element-parsing-algorithm
1387 *
1388 * @param string $tag_name The uppercase tag name which will close the RAWTEXT region.
1389 * @return bool Whether an end to the RAWTEXT region was found before the end of the document.
1390 */
1391 private function skip_rawtext( string $tag_name ): bool {
1392 /*
1393 * These two functions distinguish themselves on whether character references are
1394 * decoded, and since functionality to read the inner markup isn't supported, it's
1395 * not necessary to implement these two functions separately.
1396 */
1397 return $this->skip_rcdata( $tag_name );
1398 }
1399
1400 /**
1401 * Skips contents of RCDATA elements, namely title and textarea tags.
1402 *
1403 * @since 6.2.0
1404 *
1405 * @see https://html.spec.whatwg.org/multipage/parsing.html#rcdata-state
1406 *
1407 * @param string $tag_name The uppercase tag name which will close the RCDATA region.
1408 * @return bool Whether an end to the RCDATA region was found before the end of the document.
1409 */
1410 private function skip_rcdata( string $tag_name ): bool {
1411 $html = $this->html;
1412 $doc_length = strlen( $html );
1413 $tag_length = strlen( $tag_name );
1414
1415 $at = $this->bytes_already_parsed;
1416
1417 while ( false !== $at && $at < $doc_length ) {
1418 $at = strpos( $this->html, '</', $at );
1419 $this->tag_name_starts_at = $at;
1420
1421 // Fail if there is no possible tag closer.
1422 if ( false === $at || ( $at + $tag_length ) >= $doc_length ) {
1423 return false;
1424 }
1425
1426 $at += 2;
1427
1428 /*
1429 * Find a case-insensitive match to the tag name.
1430 *
1431 * Because tag names are limited to US-ASCII there is no
1432 * need to perform any kind of Unicode normalization when
1433 * comparing; any character which could be impacted by such
1434 * normalization could not be part of a tag name.
1435 */
1436 for ( $i = 0; $i < $tag_length; $i++ ) {
1437 $tag_char = $tag_name[ $i ];
1438 $html_char = $html[ $at + $i ];
1439
1440 if ( $html_char !== $tag_char && strtoupper( $html_char ) !== $tag_char ) {
1441 $at += $i;
1442 continue 2;
1443 }
1444 }
1445
1446 $at += $tag_length;
1447 $this->bytes_already_parsed = $at;
1448
1449 if ( $at >= strlen( $html ) ) {
1450 return false;
1451 }
1452
1453 /*
1454 * Ensure that the tag name terminates to avoid matching on
1455 * substrings of a longer tag name. For example, the sequence
1456 * "</textarearug" should not match for "</textarea" even
1457 * though "textarea" is found within the text.
1458 */
1459 $c = $html[ $at ];
1460 if ( ' ' !== $c && "\t" !== $c && "\r" !== $c && "\n" !== $c && '/' !== $c && '>' !== $c ) {
1461 continue;
1462 }
1463
1464 while ( $this->parse_next_attribute() ) {
1465 continue;
1466 }
1467
1468 $at = $this->bytes_already_parsed;
1469 if ( $at >= strlen( $this->html ) ) {
1470 return false;
1471 }
1472
1473 if ( '>' === $html[ $at ] ) {
1474 $this->bytes_already_parsed = $at + 1;
1475 return true;
1476 }
1477
1478 if ( $at + 1 >= strlen( $this->html ) ) {
1479 return false;
1480 }
1481
1482 if ( '/' === $html[ $at ] && '>' === $html[ $at + 1 ] ) {
1483 $this->bytes_already_parsed = $at + 2;
1484 return true;
1485 }
1486 }
1487
1488 return false;
1489 }
1490
1491 /**
1492 * Skips contents of script tags.
1493 *
1494 * @since 6.2.0
1495 *
1496 * @return bool Whether the script tag was closed before the end of the document.
1497 */
1498 private function skip_script_data(): bool {
1499 $state = 'unescaped';
1500 $html = $this->html;
1501 $doc_length = strlen( $html );
1502 $at = $this->bytes_already_parsed;
1503
1504 while ( false !== $at && $at < $doc_length ) {
1505 $at += strcspn( $html, '-<', $at );
1506
1507 /*
1508 * Optimization: Terminating a complete script element requires at least eight
1509 * additional bytes in the document. Some checks below may cause local escaped
1510 * state transitions when processing shorter strings, but those transitions are
1511 * irrelevant if the script tag is incomplete and the function must return false.
1512 *
1513 * This may need updating if those transitions become significant or exported from
1514 * this function in some way, such as when building safe methods to embed JavaScript
1515 * or data inside a SCRIPT element.
1516 *
1517 * $at may be here.
1518 * β
1519 * ...</script>
1520 * β°βββ¬ββββ―
1521 * $at + 8 additional bytes are required for a non-false return value.
1522 *
1523 * This single check eliminates the need to check lengths for the shorter spans:
1524 *
1525 * $at may be here.
1526 * β
1527 * <script><!-- --></script>
1528 * ββ―
1529 * $at + 2 additional characters does not require a length check.
1530 *
1531 * The transition from "escaped" to "unescaped" is not relevant if the document ends:
1532 *
1533 * $at may be here.
1534 * β
1535 * <script><!-- -->[[END-OF-DOCUMENT]]
1536 * β°βββ¬ββββ―
1537 * $at + 8 additional bytes is not satisfied, return false.
1538 */
1539 if ( $at + 8 >= $doc_length ) {
1540 return false;
1541 }
1542
1543 /*
1544 * For all script states a "-->" transitions
1545 * back into the normal unescaped script mode,
1546 * even if that's the current state.
1547 */
1548 if (
1549 '-' === $html[ $at ] &&
1550 '-' === $html[ $at + 1 ] &&
1551 '>' === $html[ $at + 2 ]
1552 ) {
1553 $at += 3;
1554 $state = 'unescaped';
1555 continue;
1556 }
1557
1558 /*
1559 * Everything of interest past here starts with "<".
1560 * Check this character and advance position regardless.
1561 */
1562 if ( '<' !== $html[ $at++ ] ) {
1563 continue;
1564 }
1565
1566 /*
1567 * "<!--" only transitions from _unescaped_ to _escaped_. This byte sequence is only
1568 * significant in the _unescaped_ state and is ignored in any other state.
1569 */
1570 if (
1571 'unescaped' === $state &&
1572 '!' === $html[ $at ] &&
1573 '-' === $html[ $at + 1 ] &&
1574 '-' === $html[ $at + 2 ]
1575 ) {
1576 $at += 3;
1577
1578 /*
1579 * The parser is ready to enter the _escaped_ state, but may remain in the
1580 * _unescaped_ state. This occurs when "<!--" is immediately followed by a
1581 * sequence of 0 or more "-" followed by ">". This is similar to abruptly closed
1582 * HTML comments like "<!-->" or "<!--->".
1583 *
1584 * Note that this check may advance the position significantly and requires a
1585 * length check to prevent bad offsets on inputs like `<script><!---------`.
1586 */
1587 $at += strspn( $html, '-', $at );
1588 if ( $at < $doc_length && '>' === $html[ $at ] ) {
1589 ++$at;
1590 continue;
1591 }
1592
1593 $state = 'escaped';
1594 continue;
1595 }
1596
1597 if ( '/' === $html[ $at ] ) {
1598 $closer_potentially_starts_at = $at - 1;
1599 $is_closing = true;
1600 ++$at;
1601 } else {
1602 $is_closing = false;
1603 }
1604
1605 /*
1606 * At this point the only remaining state-changes occur with the
1607 * <script> and </script> tags; unless one of these appears next,
1608 * proceed scanning to the next potential token in the text.
1609 */
1610 if ( ! (
1611 ( 's' === $html[ $at ] || 'S' === $html[ $at ] ) &&
1612 ( 'c' === $html[ $at + 1 ] || 'C' === $html[ $at + 1 ] ) &&
1613 ( 'r' === $html[ $at + 2 ] || 'R' === $html[ $at + 2 ] ) &&
1614 ( 'i' === $html[ $at + 3 ] || 'I' === $html[ $at + 3 ] ) &&
1615 ( 'p' === $html[ $at + 4 ] || 'P' === $html[ $at + 4 ] ) &&
1616 ( 't' === $html[ $at + 5 ] || 'T' === $html[ $at + 5 ] )
1617 ) ) {
1618 ++$at;
1619 continue;
1620 }
1621
1622 /*
1623 * Ensure that the script tag terminates to avoid matching on
1624 * substrings of a non-match. For example, the sequence
1625 * "<script123" should not end a script region even though
1626 * "<script" is found within the text.
1627 */
1628 $at += 6;
1629 $c = $html[ $at ];
1630 if (
1631 /**
1632 * These characters trigger state transitions of interest:
1633 *
1634 * - @see {https://html.spec.whatwg.org/multipage/parsing.html#script-data-end-tag-name-state}
1635 * - @see {https://html.spec.whatwg.org/multipage/parsing.html#script-data-escaped-end-tag-name-state}
1636 * - @see {https://html.spec.whatwg.org/multipage/parsing.html#script-data-double-escape-start-state}
1637 * - @see {https://html.spec.whatwg.org/multipage/parsing.html#script-data-double-escape-end-state}
1638 *
1639 * The "\r" character is not present in the above references. However, "\r" must be
1640 * treated the same as "\n". This is because the HTML Standard requires newline
1641 * normalization during preprocessing which applies this replacement.
1642 *
1643 * - @see https://html.spec.whatwg.org/multipage/parsing.html#preprocessing-the-input-stream
1644 * - @see https://infra.spec.whatwg.org/#normalize-newlines
1645 */
1646 '>' !== $c &&
1647 ' ' !== $c &&
1648 "\n" !== $c &&
1649 '/' !== $c &&
1650 "\t" !== $c &&
1651 "\f" !== $c &&
1652 "\r" !== $c
1653 ) {
1654 continue;
1655 }
1656
1657 if ( 'escaped' === $state && ! $is_closing ) {
1658 $state = 'double-escaped';
1659 continue;
1660 }
1661
1662 if ( 'double-escaped' === $state && $is_closing ) {
1663 $state = 'escaped';
1664 continue;
1665 }
1666
1667 if ( $is_closing ) {
1668 $this->bytes_already_parsed = $closer_potentially_starts_at;
1669 $this->tag_name_starts_at = $closer_potentially_starts_at;
1670 if ( $this->bytes_already_parsed >= $doc_length ) {
1671 return false;
1672 }
1673
1674 while ( $this->parse_next_attribute() ) {
1675 continue;
1676 }
1677
1678 if ( $this->bytes_already_parsed >= $doc_length ) {
1679 return false;
1680 }
1681
1682 if ( '>' === $html[ $this->bytes_already_parsed ] ) {
1683 ++$this->bytes_already_parsed;
1684 return true;
1685 }
1686 }
1687
1688 ++$at;
1689 }
1690
1691 return false;
1692 }
1693
1694 /**
1695 * Parses the next tag.
1696 *
1697 * This will find and start parsing the next tag, including
1698 * the opening `<`, the potential closer `/`, and the tag
1699 * name. It does not parse the attributes or scan to the
1700 * closing `>`; these are left for other methods.
1701 *
1702 * @since 6.2.0
1703 * @since 6.2.1 Support abruptly-closed comments, invalid-tag-closer-comments, and empty elements.
1704 *
1705 * @return bool Whether a tag was found before the end of the document.
1706 */
1707 private function parse_next_tag(): bool {
1708 $this->after_tag();
1709
1710 $html = $this->html;
1711 $doc_length = strlen( $html );
1712 $was_at = $this->bytes_already_parsed;
1713 $at = $was_at;
1714
1715 while ( $at < $doc_length ) {
1716 $at = strpos( $html, '<', $at );
1717 if ( false === $at ) {
1718 break;
1719 }
1720
1721 if ( $at > $was_at ) {
1722 /*
1723 * A "<" normally starts a new HTML tag or syntax token, but in cases where the
1724 * following character can't produce a valid token, the "<" is instead treated
1725 * as plaintext and the parser should skip over it. This avoids a problem when
1726 * following earlier practices of typing emoji with text, e.g. "<3". This
1727 * should be a heart, not a tag. It's supposed to be rendered, not hidden.
1728 *
1729 * At this point the parser checks if this is one of those cases and if it is
1730 * will continue searching for the next "<" in search of a token boundary.
1731 *
1732 * @see https://html.spec.whatwg.org/#tag-open-state
1733 */
1734 if ( 1 !== strspn( $html, '!/?abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ', $at + 1, 1 ) ) {
1735 ++$at;
1736 continue;
1737 }
1738
1739 $this->parser_state = self::STATE_TEXT_NODE;
1740 $this->token_starts_at = $was_at;
1741 $this->token_length = $at - $was_at;
1742 $this->text_starts_at = $was_at;
1743 $this->text_length = $this->token_length;
1744 $this->bytes_already_parsed = $at;
1745 return true;
1746 }
1747
1748 $this->token_starts_at = $at;
1749
1750 if ( $at + 1 < $doc_length && '/' === $this->html[ $at + 1 ] ) {
1751 $this->is_closing_tag = true;
1752 ++$at;
1753 } else {
1754 $this->is_closing_tag = false;
1755 }
1756
1757 /*
1758 * HTML tag names must start with [a-zA-Z] otherwise they are not tags.
1759 * For example, "<3" is rendered as text, not a tag opener. If at least
1760 * one letter follows the "<" then _it is_ a tag, but if the following
1761 * character is anything else it _is not a tag_.
1762 *
1763 * It's not uncommon to find non-tags starting with `<` in an HTML
1764 * document, so it's good for performance to make this pre-check before
1765 * continuing to attempt to parse a tag name.
1766 *
1767 * Reference:
1768 * * https://html.spec.whatwg.org/multipage/parsing.html#data-state
1769 * * https://html.spec.whatwg.org/multipage/parsing.html#tag-open-state
1770 */
1771 $tag_name_prefix_length = strspn( $html, 'abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ', $at + 1 );
1772 if ( $tag_name_prefix_length > 0 ) {
1773 ++$at;
1774 $this->parser_state = self::STATE_MATCHED_TAG;
1775 $this->tag_name_starts_at = $at;
1776 $this->tag_name_length = $tag_name_prefix_length + strcspn( $html, " \t\f\r\n/>", $at + $tag_name_prefix_length );
1777 $this->bytes_already_parsed = $at + $this->tag_name_length;
1778 return true;
1779 }
1780
1781 /*
1782 * Abort if no tag is found before the end of
1783 * the document. There is nothing left to parse.
1784 */
1785 if ( $at + 1 >= $doc_length ) {
1786 $this->parser_state = self::STATE_INCOMPLETE_INPUT;
1787
1788 return false;
1789 }
1790
1791 /*
1792 * `<!` transitions to markup declaration open state
1793 * https://html.spec.whatwg.org/multipage/parsing.html#markup-declaration-open-state
1794 */
1795 if ( ! $this->is_closing_tag && '!' === $html[ $at + 1 ] ) {
1796 /*
1797 * `<!--` transitions to a comment state β apply further comment rules.
1798 * https://html.spec.whatwg.org/multipage/parsing.html#tag-open-state
1799 */
1800 if ( 0 === substr_compare( $html, '--', $at + 2, 2 ) ) {
1801 $closer_at = $at + 4;
1802 // If it's not possible to close the comment then there is nothing more to scan.
1803 if ( $doc_length <= $closer_at ) {
1804 $this->parser_state = self::STATE_INCOMPLETE_INPUT;
1805
1806 return false;
1807 }
1808
1809 // Abruptly-closed empty comments are a sequence of dashes followed by `>`.
1810 $span_of_dashes = strspn( $html, '-', $closer_at );
1811 if ( '>' === $html[ $closer_at + $span_of_dashes ] ) {
1812 /*
1813 * @todo When implementing `set_modifiable_text()` ensure that updates to this token
1814 * don't break the syntax for short comments, e.g. `<!--->`. Unlike other comment
1815 * and bogus comment syntax, these leave no clear insertion point for text and
1816 * they need to be modified specially in order to contain text. E.g. to store
1817 * `?` as the modifiable text, the `<!--->` needs to become `<!--?-->`, which
1818 * involves inserting an additional `-` into the token after the modifiable text.
1819 */
1820 $this->parser_state = self::STATE_COMMENT;
1821 $this->comment_type = self::COMMENT_AS_ABRUPTLY_CLOSED_COMMENT;
1822 $this->token_length = $closer_at + $span_of_dashes + 1 - $this->token_starts_at;
1823
1824 // Only provide modifiable text if the token is long enough to contain it.
1825 if ( $span_of_dashes >= 2 ) {
1826 $this->comment_type = self::COMMENT_AS_HTML_COMMENT;
1827 $this->text_starts_at = $this->token_starts_at + 4;
1828 $this->text_length = $span_of_dashes - 2;
1829 }
1830
1831 $this->bytes_already_parsed = $closer_at + $span_of_dashes + 1;
1832 return true;
1833 }
1834
1835 /*
1836 * Comments may be closed by either a --> or an invalid --!>.
1837 * The first occurrence closes the comment.
1838 *
1839 * See https://html.spec.whatwg.org/#parse-error-incorrectly-closed-comment
1840 */
1841 --$closer_at; // Pre-increment inside condition below reduces risk of accidental infinite looping.
1842 while ( ++$closer_at < $doc_length ) {
1843 $closer_at = strpos( $html, '--', $closer_at );
1844 if ( false === $closer_at ) {
1845 $this->parser_state = self::STATE_INCOMPLETE_INPUT;
1846
1847 return false;
1848 }
1849
1850 if ( $closer_at + 2 < $doc_length && '>' === $html[ $closer_at + 2 ] ) {
1851 $this->parser_state = self::STATE_COMMENT;
1852 $this->comment_type = self::COMMENT_AS_HTML_COMMENT;
1853 $this->token_length = $closer_at + 3 - $this->token_starts_at;
1854 $this->text_starts_at = $this->token_starts_at + 4;
1855 $this->text_length = $closer_at - $this->text_starts_at;
1856 $this->bytes_already_parsed = $closer_at + 3;
1857 return true;
1858 }
1859
1860 if (
1861 $closer_at + 3 < $doc_length &&
1862 '!' === $html[ $closer_at + 2 ] &&
1863 '>' === $html[ $closer_at + 3 ]
1864 ) {
1865 $this->parser_state = self::STATE_COMMENT;
1866 $this->comment_type = self::COMMENT_AS_HTML_COMMENT;
1867 $this->token_length = $closer_at + 4 - $this->token_starts_at;
1868 $this->text_starts_at = $this->token_starts_at + 4;
1869 $this->text_length = $closer_at - $this->text_starts_at;
1870 $this->bytes_already_parsed = $closer_at + 4;
1871 return true;
1872 }
1873 }
1874 }
1875
1876 /*
1877 * `<!DOCTYPE` transitions to DOCTYPE state β skip to the nearest >
1878 * These are ASCII-case-insensitive.
1879 * https://html.spec.whatwg.org/multipage/parsing.html#tag-open-state
1880 */
1881 if (
1882 $doc_length > $at + 8 &&
1883 ( 'D' === $html[ $at + 2 ] || 'd' === $html[ $at + 2 ] ) &&
1884 ( 'O' === $html[ $at + 3 ] || 'o' === $html[ $at + 3 ] ) &&
1885 ( 'C' === $html[ $at + 4 ] || 'c' === $html[ $at + 4 ] ) &&
1886 ( 'T' === $html[ $at + 5 ] || 't' === $html[ $at + 5 ] ) &&
1887 ( 'Y' === $html[ $at + 6 ] || 'y' === $html[ $at + 6 ] ) &&
1888 ( 'P' === $html[ $at + 7 ] || 'p' === $html[ $at + 7 ] ) &&
1889 ( 'E' === $html[ $at + 8 ] || 'e' === $html[ $at + 8 ] )
1890 ) {
1891 $closer_at = strpos( $html, '>', $at + 9 );
1892 if ( false === $closer_at ) {
1893 $this->parser_state = self::STATE_INCOMPLETE_INPUT;
1894
1895 return false;
1896 }
1897
1898 $this->parser_state = self::STATE_DOCTYPE;
1899 $this->token_length = $closer_at + 1 - $this->token_starts_at;
1900 $this->text_starts_at = $this->token_starts_at + 9;
1901 $this->text_length = $closer_at - $this->text_starts_at;
1902 $this->bytes_already_parsed = $closer_at + 1;
1903 return true;
1904 }
1905
1906 if (
1907 'html' !== $this->parsing_namespace &&
1908 strlen( $html ) > $at + 8 &&
1909 '[' === $html[ $at + 2 ] &&
1910 'C' === $html[ $at + 3 ] &&
1911 'D' === $html[ $at + 4 ] &&
1912 'A' === $html[ $at + 5 ] &&
1913 'T' === $html[ $at + 6 ] &&
1914 'A' === $html[ $at + 7 ] &&
1915 '[' === $html[ $at + 8 ]
1916 ) {
1917 $closer_at = strpos( $html, ']]>', $at + 9 );
1918 if ( false === $closer_at ) {
1919 $this->parser_state = self::STATE_INCOMPLETE_INPUT;
1920
1921 return false;
1922 }
1923
1924 $this->parser_state = self::STATE_CDATA_NODE;
1925 $this->text_starts_at = $at + 9;
1926 $this->text_length = $closer_at - $this->text_starts_at;
1927 $this->token_length = $closer_at + 3 - $this->token_starts_at;
1928 $this->bytes_already_parsed = $closer_at + 3;
1929 return true;
1930 }
1931
1932 /*
1933 * Anything else here is an incorrectly-opened comment and transitions
1934 * to the bogus comment state - skip to the nearest >. If no closer is
1935 * found then the HTML was truncated inside the markup declaration.
1936 */
1937 $closer_at = strpos( $html, '>', $at + 1 );
1938 if ( false === $closer_at ) {
1939 $this->parser_state = self::STATE_INCOMPLETE_INPUT;
1940
1941 return false;
1942 }
1943
1944 $this->parser_state = self::STATE_COMMENT;
1945 $this->comment_type = self::COMMENT_AS_INVALID_HTML;
1946 $this->token_length = $closer_at + 1 - $this->token_starts_at;
1947 $this->text_starts_at = $this->token_starts_at + 2;
1948 $this->text_length = $closer_at - $this->text_starts_at;
1949 $this->bytes_already_parsed = $closer_at + 1;
1950
1951 /*
1952 * Identify nodes that would be CDATA if HTML had CDATA sections.
1953 *
1954 * This section must occur after identifying the bogus comment end
1955 * because in an HTML parser it will span to the nearest `>`, even
1956 * if there's no `]]>` as would be required in an XML document. It
1957 * is therefore not possible to parse a CDATA section containing
1958 * a `>` in the HTML syntax.
1959 *
1960 * Inside foreign elements there is a discrepancy between browsers
1961 * and the specification on this.
1962 *
1963 * @todo Track whether the Tag Processor is inside a foreign element
1964 * and require the proper closing `]]>` in those cases.
1965 */
1966 if (
1967 $this->token_length >= 10 &&
1968 '[' === $html[ $this->token_starts_at + 2 ] &&
1969 'C' === $html[ $this->token_starts_at + 3 ] &&
1970 'D' === $html[ $this->token_starts_at + 4 ] &&
1971 'A' === $html[ $this->token_starts_at + 5 ] &&
1972 'T' === $html[ $this->token_starts_at + 6 ] &&
1973 'A' === $html[ $this->token_starts_at + 7 ] &&
1974 '[' === $html[ $this->token_starts_at + 8 ] &&
1975 ']' === $html[ $closer_at - 1 ] &&
1976 ']' === $html[ $closer_at - 2 ]
1977 ) {
1978 $this->parser_state = self::STATE_COMMENT;
1979 $this->comment_type = self::COMMENT_AS_CDATA_LOOKALIKE;
1980 $this->text_starts_at += 7;
1981 $this->text_length -= 9;
1982 }
1983
1984 return true;
1985 }
1986
1987 /*
1988 * </> is a missing end tag name, which is ignored.
1989 *
1990 * This was also known as the "presumptuous empty tag"
1991 * in early discussions as it was proposed to close
1992 * the nearest previous opening tag.
1993 *
1994 * See https://html.spec.whatwg.org/#parse-error-missing-end-tag-name
1995 */
1996 if ( '>' === $html[ $at + 1 ] ) {
1997 // `<>` is interpreted as plaintext.
1998 if ( ! $this->is_closing_tag ) {
1999 ++$at;
2000 continue;
2001 }
2002
2003 $this->parser_state = self::STATE_PRESUMPTUOUS_TAG;
2004 $this->token_length = $at + 2 - $this->token_starts_at;
2005 $this->bytes_already_parsed = $at + 2;
2006 return true;
2007 }
2008
2009 /*
2010 * `<?` transitions to a bogus comment state β skip to the nearest >
2011 * See https://html.spec.whatwg.org/multipage/parsing.html#tag-open-state
2012 */
2013 if ( ! $this->is_closing_tag && '?' === $html[ $at + 1 ] ) {
2014 $closer_at = strpos( $html, '>', $at + 2 );
2015 if ( false === $closer_at ) {
2016 $this->parser_state = self::STATE_INCOMPLETE_INPUT;
2017
2018 return false;
2019 }
2020
2021 $this->parser_state = self::STATE_COMMENT;
2022 $this->comment_type = self::COMMENT_AS_INVALID_HTML;
2023 $this->token_length = $closer_at + 1 - $this->token_starts_at;
2024 $this->text_starts_at = $this->token_starts_at + 2;
2025 $this->text_length = $closer_at - $this->text_starts_at;
2026 $this->bytes_already_parsed = $closer_at + 1;
2027
2028 /*
2029 * Identify a Processing Instruction node were HTML to have them.
2030 *
2031 * This section must occur after identifying the bogus comment end
2032 * because in an HTML parser it will span to the nearest `>`, even
2033 * if there's no `?>` as would be required in an XML document. It
2034 * is therefore not possible to parse a Processing Instruction node
2035 * containing a `>` in the HTML syntax.
2036 *
2037 * XML allows for more target names, but this code only identifies
2038 * those with ASCII-representable target names. This means that it
2039 * may identify some Processing Instruction nodes as bogus comments,
2040 * but it will not misinterpret the HTML structure. By limiting the
2041 * identification to these target names the Tag Processor can avoid
2042 * the need to start parsing UTF-8 sequences.
2043 *
2044 * > NameStartChar ::= ":" | [A-Z] | "_" | [a-z] | [#xC0-#xD6] | [#xD8-#xF6] | [#xF8-#x2FF] |
2045 * [#x370-#x37D] | [#x37F-#x1FFF] | [#x200C-#x200D] | [#x2070-#x218F] |
2046 * [#x2C00-#x2FEF] | [#x3001-#xD7FF] | [#xF900-#xFDCF] | [#xFDF0-#xFFFD] |
2047 * [#x10000-#xEFFFF]
2048 * > NameChar ::= NameStartChar | "-" | "." | [0-9] | #xB7 | [#x0300-#x036F] | [#x203F-#x2040]
2049 *
2050 * @todo Processing instruction nodes in SGML may contain any kind of markup. XML defines a
2051 * special case with `<?xml ... ?>` syntax, but the `?` is part of the bogus comment.
2052 *
2053 * @see https://www.w3.org/TR/2006/REC-xml11-20060816/#NT-PITarget
2054 */
2055 if ( $this->token_length >= 5 && '?' === $html[ $closer_at - 1 ] ) {
2056 $comment_text = substr( $html, $this->token_starts_at + 2, $this->token_length - 4 );
2057 $pi_target_length = strspn( $comment_text, 'abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ:_' );
2058
2059 if ( 0 < $pi_target_length ) {
2060 $pi_target_length += strspn( $comment_text, 'abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789:_-.', $pi_target_length );
2061
2062 $this->comment_type = self::COMMENT_AS_PI_NODE_LOOKALIKE;
2063 $this->tag_name_starts_at = $this->token_starts_at + 2;
2064 $this->tag_name_length = $pi_target_length;
2065 $this->text_starts_at += $pi_target_length;
2066 $this->text_length -= $pi_target_length + 1;
2067 }
2068 }
2069
2070 return true;
2071 }
2072
2073 /*
2074 * If a non-alpha starts the tag name in a tag closer it's a comment.
2075 * Find the first `>`, which closes the comment.
2076 *
2077 * This parser classifies these particular comments as special "funky comments"
2078 * which are made available for further processing.
2079 *
2080 * See https://html.spec.whatwg.org/#parse-error-invalid-first-character-of-tag-name
2081 */
2082 if ( $this->is_closing_tag ) {
2083 // No chance of finding a closer.
2084 if ( $at + 3 > $doc_length ) {
2085 $this->parser_state = self::STATE_INCOMPLETE_INPUT;
2086
2087 return false;
2088 }
2089
2090 $closer_at = strpos( $html, '>', $at + 2 );
2091 if ( false === $closer_at ) {
2092 $this->parser_state = self::STATE_INCOMPLETE_INPUT;
2093
2094 return false;
2095 }
2096
2097 $this->parser_state = self::STATE_FUNKY_COMMENT;
2098 $this->token_length = $closer_at + 1 - $this->token_starts_at;
2099 $this->text_starts_at = $this->token_starts_at + 2;
2100 $this->text_length = $closer_at - $this->text_starts_at;
2101 $this->bytes_already_parsed = $closer_at + 1;
2102 return true;
2103 }
2104
2105 ++$at;
2106 }
2107
2108 /*
2109 * This does not imply an incomplete parse; it indicates that there
2110 * can be nothing left in the document other than a #text node.
2111 */
2112 $this->parser_state = self::STATE_TEXT_NODE;
2113 $this->token_starts_at = $was_at;
2114 $this->token_length = $doc_length - $was_at;
2115 $this->text_starts_at = $was_at;
2116 $this->text_length = $this->token_length;
2117 $this->bytes_already_parsed = $doc_length;
2118 return true;
2119 }
2120
2121 /**
2122 * Parses the next attribute.
2123 *
2124 * @since 6.2.0
2125 *
2126 * @return bool Whether an attribute was found before the end of the document.
2127 */
2128 private function parse_next_attribute(): bool {
2129 $doc_length = strlen( $this->html );
2130
2131 // Skip whitespace and slashes.
2132 $this->bytes_already_parsed += strspn( $this->html, " \t\f\r\n/", $this->bytes_already_parsed );
2133 if ( $this->bytes_already_parsed >= $doc_length ) {
2134 $this->parser_state = self::STATE_INCOMPLETE_INPUT;
2135
2136 return false;
2137 }
2138
2139 /*
2140 * Treat the equal sign as a part of the attribute
2141 * name if it is the first encountered byte.
2142 *
2143 * @see https://html.spec.whatwg.org/multipage/parsing.html#before-attribute-name-state
2144 */
2145 $name_length = '=' === $this->html[ $this->bytes_already_parsed ]
2146 ? 1 + strcspn( $this->html, "=/> \t\f\r\n", $this->bytes_already_parsed + 1 )
2147 : strcspn( $this->html, "=/> \t\f\r\n", $this->bytes_already_parsed );
2148
2149 // No attribute, just tag closer.
2150 if ( 0 === $name_length || $this->bytes_already_parsed + $name_length >= $doc_length ) {
2151 return false;
2152 }
2153
2154 $attribute_start = $this->bytes_already_parsed;
2155 $attribute_name = substr( $this->html, $attribute_start, $name_length );
2156 $this->bytes_already_parsed += $name_length;
2157 if ( $this->bytes_already_parsed >= $doc_length ) {
2158 $this->parser_state = self::STATE_INCOMPLETE_INPUT;
2159
2160 return false;
2161 }
2162
2163 $this->skip_whitespace();
2164 if ( $this->bytes_already_parsed >= $doc_length ) {
2165 $this->parser_state = self::STATE_INCOMPLETE_INPUT;
2166
2167 return false;
2168 }
2169
2170 $has_value = '=' === $this->html[ $this->bytes_already_parsed ];
2171 if ( $has_value ) {
2172 ++$this->bytes_already_parsed;
2173 $this->skip_whitespace();
2174 if ( $this->bytes_already_parsed >= $doc_length ) {
2175 $this->parser_state = self::STATE_INCOMPLETE_INPUT;
2176
2177 return false;
2178 }
2179
2180 switch ( $this->html[ $this->bytes_already_parsed ] ) {
2181 case "'":
2182 case '"':
2183 $quote = $this->html[ $this->bytes_already_parsed ];
2184 $value_start = $this->bytes_already_parsed + 1;
2185 $end_quote_at = strpos( $this->html, $quote, $value_start );
2186 $end_quote_at = false === $end_quote_at ? $doc_length : $end_quote_at;
2187 $value_length = $end_quote_at - $value_start;
2188 $attribute_end = $end_quote_at + 1;
2189 $this->bytes_already_parsed = $attribute_end;
2190 break;
2191
2192 default:
2193 $value_start = $this->bytes_already_parsed;
2194 $value_length = strcspn( $this->html, "> \t\f\r\n", $value_start );
2195 $attribute_end = $value_start + $value_length;
2196 $this->bytes_already_parsed = $attribute_end;
2197 }
2198 } else {
2199 $value_start = $this->bytes_already_parsed;
2200 $value_length = 0;
2201 $attribute_end = $attribute_start + $name_length;
2202 }
2203
2204 if ( $attribute_end >= $doc_length ) {
2205 $this->parser_state = self::STATE_INCOMPLETE_INPUT;
2206
2207 return false;
2208 }
2209
2210 if ( $this->is_closing_tag ) {
2211 return true;
2212 }
2213
2214 /*
2215 * > There must never be two or more attributes on
2216 * > the same start tag whose names are an ASCII
2217 * > case-insensitive match for each other.
2218 * - HTML 5 spec
2219 *
2220 * @see https://html.spec.whatwg.org/multipage/syntax.html#attributes-2:ascii-case-insensitive
2221 */
2222 $comparable_name = strtolower( $attribute_name );
2223
2224 // If an attribute is listed many times, only use the first declaration and ignore the rest.
2225 if ( ! isset( $this->attributes[ $comparable_name ] ) ) {
2226 $this->attributes[ $comparable_name ] = new WP_HTML_Attribute_Token(
2227 $attribute_name,
2228 $value_start,
2229 $value_length,
2230 $attribute_start,
2231 $attribute_end - $attribute_start,
2232 ! $has_value
2233 );
2234
2235 return true;
2236 }
2237
2238 /*
2239 * Track the duplicate attributes so if we remove it, all disappear together.
2240 *
2241 * While `$this->duplicated_attributes` could always be stored as an `array()`,
2242 * which would simplify the logic here, storing a `null` and only allocating
2243 * an array when encountering duplicates avoids needless allocations in the
2244 * normative case of parsing tags with no duplicate attributes.
2245 */
2246 $duplicate_span = new WP_HTML_Span( $attribute_start, $attribute_end - $attribute_start );
2247 if ( null === $this->duplicate_attributes ) {
2248 $this->duplicate_attributes = array( $comparable_name => array( $duplicate_span ) );
2249 } elseif ( ! isset( $this->duplicate_attributes[ $comparable_name ] ) ) {
2250 $this->duplicate_attributes[ $comparable_name ] = array( $duplicate_span );
2251 } else {
2252 $this->duplicate_attributes[ $comparable_name ][] = $duplicate_span;
2253 }
2254
2255 return true;
2256 }
2257
2258 /**
2259 * Move the internal cursor past any immediate successive whitespace.
2260 *
2261 * @since 6.2.0
2262 */
2263 private function skip_whitespace(): void {
2264 $this->bytes_already_parsed += strspn( $this->html, " \t\f\r\n", $this->bytes_already_parsed );
2265 }
2266
2267 /**
2268 * Applies attribute updates and cleans up once a tag is fully parsed.
2269 *
2270 * @since 6.2.0
2271 */
2272 private function after_tag(): void {
2273 /*
2274 * There could be lexical updates enqueued for an attribute that
2275 * also exists on the next tag. In order to avoid conflating the
2276 * attributes across the two tags, lexical updates with names
2277 * need to be flushed to raw lexical updates.
2278 */
2279 $this->class_name_updates_to_attributes_updates();
2280
2281 /*
2282 * Purge updates if there are too many. The actual count isn't
2283 * scientific, but a few values from 100 to a few thousand were
2284 * tests to find a practically-useful limit.
2285 *
2286 * If the update queue grows too big, then the Tag Processor
2287 * will spend more time iterating through them and lose the
2288 * efficiency gains of deferring applying them.
2289 */
2290 if ( 1000 < count( $this->lexical_updates ) ) {
2291 $this->get_updated_html();
2292 }
2293
2294 foreach ( $this->lexical_updates as $name => $update ) {
2295 /*
2296 * Any updates appearing after the cursor should be applied
2297 * before proceeding, otherwise they may be overlooked.
2298 */
2299 if ( $update->start >= $this->bytes_already_parsed ) {
2300 $this->get_updated_html();
2301 break;
2302 }
2303
2304 if ( is_int( $name ) ) {
2305 continue;
2306 }
2307
2308 $this->lexical_updates[] = $update;
2309 unset( $this->lexical_updates[ $name ] );
2310 }
2311
2312 $this->token_starts_at = null;
2313 $this->token_length = null;
2314 $this->tag_name_starts_at = null;
2315 $this->tag_name_length = null;
2316 $this->text_starts_at = 0;
2317 $this->text_length = 0;
2318 $this->is_closing_tag = null;
2319 $this->attributes = array();
2320 $this->comment_type = null;
2321 $this->text_node_classification = self::TEXT_IS_GENERIC;
2322 $this->duplicate_attributes = null;
2323 }
2324
2325 /**
2326 * Converts class name updates into tag attributes updates
2327 * (they are accumulated in different data formats for performance).
2328 *
2329 * @since 6.2.0
2330 *
2331 * @see WP_HTML_Tag_Processor::$lexical_updates
2332 * @see WP_HTML_Tag_Processor::$classname_updates
2333 */
2334 private function class_name_updates_to_attributes_updates(): void {
2335 if ( count( $this->classname_updates ) === 0 ) {
2336 return;
2337 }
2338
2339 $existing_class = $this->get_enqueued_attribute_value( 'class' );
2340 if ( null === $existing_class || true === $existing_class ) {
2341 $existing_class = '';
2342 }
2343
2344 if ( false === $existing_class && isset( $this->attributes['class'] ) ) {
2345 $existing_class = WP_HTML_Decoder::decode_attribute(
2346 substr(
2347 $this->html,
2348 $this->attributes['class']->value_starts_at,
2349 $this->attributes['class']->value_length
2350 )
2351 );
2352 }
2353
2354 if ( false === $existing_class ) {
2355 $existing_class = '';
2356 }
2357
2358 /**
2359 * Updated "class" attribute value.
2360 *
2361 * This is incrementally built while scanning through the existing class
2362 * attribute, skipping removed classes on the way, and then appending
2363 * added classes at the end. Only when finished processing will the
2364 * value contain the final new value.
2365
2366 * @var string $class
2367 */
2368 $class = '';
2369
2370 /**
2371 * Tracks the cursor position in the existing
2372 * class attribute value while parsing.
2373 *
2374 * @var int $at
2375 */
2376 $at = 0;
2377
2378 /**
2379 * Indicates if there's any need to modify the existing class attribute.
2380 *
2381 * If a call to `add_class()` and `remove_class()` wouldn't impact
2382 * the `class` attribute value then there's no need to rebuild it.
2383 * For example, when adding a class that's already present or
2384 * removing one that isn't.
2385 *
2386 * This flag enables a performance optimization when none of the enqueued
2387 * class updates would impact the `class` attribute; namely, that the
2388 * processor can continue without modifying the input document, as if
2389 * none of the `add_class()` or `remove_class()` calls had been made.
2390 *
2391 * This flag is set upon the first change that requires a string update.
2392 *
2393 * @var bool $modified
2394 */
2395 $modified = false;
2396
2397 $seen = array();
2398 $to_remove = array();
2399 $is_quirks = self::QUIRKS_MODE === $this->compat_mode;
2400 if ( $is_quirks ) {
2401 foreach ( $this->classname_updates as $updated_name => $action ) {
2402 if ( self::REMOVE_CLASS === $action ) {
2403 $to_remove[] = strtolower( $updated_name );
2404 }
2405 }
2406 } else {
2407 foreach ( $this->classname_updates as $updated_name => $action ) {
2408 if ( self::REMOVE_CLASS === $action ) {
2409 $to_remove[] = $updated_name;
2410 }
2411 }
2412 }
2413
2414 // Remove unwanted classes by only copying the new ones.
2415 $existing_class_length = strlen( $existing_class );
2416 while ( $at < $existing_class_length ) {
2417 // Skip to the first non-whitespace character.
2418 $ws_at = $at;
2419 $ws_length = strspn( $existing_class, " \t\f\r\n", $ws_at );
2420 $at += $ws_length;
2421
2422 // Capture the class name β it's everything until the next whitespace.
2423 $name_length = strcspn( $existing_class, " \t\f\r\n", $at );
2424 if ( 0 === $name_length ) {
2425 // If no more class names are found then that's the end.
2426 break;
2427 }
2428
2429 $name = substr( $existing_class, $at, $name_length );
2430 $comparable_class_name = $is_quirks ? strtolower( $name ) : $name;
2431 $at += $name_length;
2432
2433 // If this class is marked for removal, remove it and move on to the next one.
2434 if ( in_array( $comparable_class_name, $to_remove, true ) ) {
2435 $modified = true;
2436 continue;
2437 }
2438
2439 // If a class has already been seen then skip it; it should not be added twice.
2440 if ( in_array( $comparable_class_name, $seen, true ) ) {
2441 continue;
2442 }
2443
2444 $seen[] = $comparable_class_name;
2445
2446 /*
2447 * Otherwise, append it to the new "class" attribute value.
2448 *
2449 * There are options for handling whitespace between tags.
2450 * Preserving the existing whitespace produces fewer changes
2451 * to the HTML content and should clarify the before/after
2452 * content when debugging the modified output.
2453 *
2454 * This approach contrasts normalizing the inter-class
2455 * whitespace to a single space, which might appear cleaner
2456 * in the output HTML but produce a noisier change.
2457 */
2458 if ( '' !== $class ) {
2459 $class .= substr( $existing_class, $ws_at, $ws_length );
2460 }
2461 $class .= $name;
2462 }
2463
2464 // Add new classes by appending those which haven't already been seen.
2465 foreach ( $this->classname_updates as $name => $operation ) {
2466 $comparable_name = $is_quirks ? strtolower( $name ) : $name;
2467 if ( self::ADD_CLASS === $operation && ! in_array( $comparable_name, $seen, true ) ) {
2468 $modified = true;
2469
2470 $class .= strlen( $class ) > 0 ? ' ' : '';
2471 $class .= $name;
2472 }
2473 }
2474
2475 $this->classname_updates = array();
2476 if ( ! $modified ) {
2477 return;
2478 }
2479
2480 if ( strlen( $class ) > 0 ) {
2481 $this->set_attribute( 'class', $class );
2482 } else {
2483 $this->remove_attribute( 'class' );
2484 }
2485 }
2486
2487 /**
2488 * Applies attribute updates to HTML document.
2489 *
2490 * @since 6.2.0
2491 * @since 6.2.1 Accumulates shift for internal cursor and passed pointer.
2492 * @since 6.3.0 Invalidate any bookmarks whose targets are overwritten.
2493 *
2494 * @param int $shift_this_point Accumulate and return shift for this position.
2495 * @return int How many bytes the given pointer moved in response to the updates.
2496 */
2497 private function apply_attributes_updates( int $shift_this_point ): int {
2498 if ( ! count( $this->lexical_updates ) ) {
2499 return 0;
2500 }
2501
2502 $accumulated_shift_for_given_point = 0;
2503
2504 /*
2505 * Attribute updates can be enqueued in any order but updates
2506 * to the document must occur in lexical order; that is, each
2507 * replacement must be made before all others which follow it
2508 * at later string indices in the input document.
2509 *
2510 * Sorting avoid making out-of-order replacements which
2511 * can lead to mangled output, partially-duplicated
2512 * attributes, and overwritten attributes.
2513 */
2514 usort( $this->lexical_updates, array( self::class, 'sort_start_ascending' ) );
2515
2516 $bytes_already_copied = 0;
2517 $output_buffer = '';
2518 foreach ( $this->lexical_updates as $diff ) {
2519 $shift = strlen( $diff->text ) - $diff->length;
2520
2521 // Adjust the cursor position by however much an update affects it.
2522 if ( $diff->start < $this->bytes_already_parsed ) {
2523 $this->bytes_already_parsed += $shift;
2524 }
2525
2526 // Accumulate shift of the given pointer within this function call.
2527 if ( $diff->start < $shift_this_point ) {
2528 $accumulated_shift_for_given_point += $shift;
2529 }
2530
2531 $output_buffer .= substr( $this->html, $bytes_already_copied, $diff->start - $bytes_already_copied );
2532 $output_buffer .= $diff->text;
2533 $bytes_already_copied = $diff->start + $diff->length;
2534 }
2535
2536 $this->html = $output_buffer . substr( $this->html, $bytes_already_copied );
2537
2538 /*
2539 * Adjust bookmark locations to account for how the text
2540 * replacements adjust offsets in the input document.
2541 */
2542 foreach ( $this->bookmarks as $bookmark_name => $bookmark ) {
2543 $bookmark_end = $bookmark->start + $bookmark->length;
2544
2545 /*
2546 * Each lexical update which appears before the bookmark's endpoints
2547 * might shift the offsets for those endpoints. Loop through each change
2548 * and accumulate the total shift for each bookmark, then apply that
2549 * shift after tallying the full delta.
2550 */
2551 $head_delta = 0;
2552 $tail_delta = 0;
2553
2554 foreach ( $this->lexical_updates as $diff ) {
2555 $diff_end = $diff->start + $diff->length;
2556
2557 if ( $bookmark->start < $diff->start && $bookmark_end < $diff->start ) {
2558 break;
2559 }
2560
2561 if ( $bookmark->start >= $diff->start && $bookmark_end < $diff_end ) {
2562 $this->release_bookmark( $bookmark_name );
2563 continue 2;
2564 }
2565
2566 $delta = strlen( $diff->text ) - $diff->length;
2567
2568 if ( $bookmark->start >= $diff->start ) {
2569 $head_delta += $delta;
2570 }
2571
2572 if ( $bookmark_end >= $diff_end ) {
2573 $tail_delta += $delta;
2574 }
2575 }
2576
2577 $bookmark->start += $head_delta;
2578 $bookmark->length += $tail_delta - $head_delta;
2579 }
2580
2581 $this->lexical_updates = array();
2582
2583 return $accumulated_shift_for_given_point;
2584 }
2585
2586 /**
2587 * Checks whether a bookmark with the given name exists.
2588 *
2589 * @since 6.3.0
2590 *
2591 * @param string $bookmark_name Name to identify a bookmark that potentially exists.
2592 * @return bool Whether that bookmark exists.
2593 */
2594 public function has_bookmark( $bookmark_name ): bool {
2595 return array_key_exists( $bookmark_name, $this->bookmarks );
2596 }
2597
2598 /**
2599 * Move the internal cursor in the Tag Processor to a given bookmark's location.
2600 *
2601 * In order to prevent accidental infinite loops, there's a
2602 * maximum limit on the number of times seek() can be called.
2603 *
2604 * @since 6.2.0
2605 *
2606 * @param string $bookmark_name Jump to the place in the document identified by this bookmark name.
2607 * @return bool Whether the internal cursor was successfully moved to the bookmark's location.
2608 */
2609 public function seek( $bookmark_name ): bool {
2610 if ( ! array_key_exists( $bookmark_name, $this->bookmarks ) ) {
2611 _doing_it_wrong(
2612 __METHOD__,
2613 __( 'Unknown bookmark name.' ),
2614 '6.2.0'
2615 );
2616 return false;
2617 }
2618
2619 $existing_bookmark = $this->bookmarks[ $bookmark_name ];
2620
2621 if (
2622 $this->token_starts_at === $existing_bookmark->start &&
2623 $this->token_length === $existing_bookmark->length
2624 ) {
2625 return true;
2626 }
2627
2628 if ( ++$this->seek_count > static::MAX_SEEK_OPS ) {
2629 _doing_it_wrong(
2630 __METHOD__,
2631 __( 'Too many calls to seek() - this can lead to performance issues.' ),
2632 '6.2.0'
2633 );
2634 return false;
2635 }
2636
2637 // Flush out any pending updates to the document.
2638 $this->get_updated_html();
2639
2640 // Point this tag processor before the sought tag opener and consume it.
2641 $this->bytes_already_parsed = $this->bookmarks[ $bookmark_name ]->start;
2642 $this->parser_state = self::STATE_READY;
2643 return $this->next_token();
2644 }
2645
2646 /**
2647 * Compare two WP_HTML_Text_Replacement objects.
2648 *
2649 * @since 6.2.0
2650 *
2651 * @param WP_HTML_Text_Replacement $a First attribute update.
2652 * @param WP_HTML_Text_Replacement $b Second attribute update.
2653 * @return int Comparison value for string order.
2654 */
2655 private static function sort_start_ascending( WP_HTML_Text_Replacement $a, WP_HTML_Text_Replacement $b ): int {
2656 $by_start = $a->start - $b->start;
2657 if ( 0 !== $by_start ) {
2658 return $by_start;
2659 }
2660
2661 $by_text = isset( $a->text, $b->text ) ? strcmp( $a->text, $b->text ) : 0;
2662 if ( 0 !== $by_text ) {
2663 return $by_text;
2664 }
2665
2666 /*
2667 * This code should be unreachable, because it implies the two replacements
2668 * start at the same location and contain the same text.
2669 */
2670 return $a->length - $b->length;
2671 }
2672
2673 /**
2674 * Return the enqueued value for a given attribute, if one exists.
2675 *
2676 * Enqueued updates can take different data types:
2677 * - If an update is enqueued and is boolean, the return will be `true`
2678 * - If an update is otherwise enqueued, the return will be the string value of that update.
2679 * - If an attribute is enqueued to be removed, the return will be `null` to indicate that.
2680 * - If no updates are enqueued, the return will be `false` to differentiate from "removed."
2681 *
2682 * @since 6.2.0
2683 *
2684 * @param string $comparable_name The attribute name in its comparable form.
2685 * @return string|boolean|null Value of enqueued update if present, otherwise false.
2686 */
2687 private function get_enqueued_attribute_value( string $comparable_name ) {
2688 if ( self::STATE_MATCHED_TAG !== $this->parser_state ) {
2689 return false;
2690 }
2691
2692 if ( ! isset( $this->lexical_updates[ $comparable_name ] ) ) {
2693 return false;
2694 }
2695
2696 $enqueued_text = $this->lexical_updates[ $comparable_name ]->text;
2697
2698 // Removed attributes erase the entire span.
2699 if ( '' === $enqueued_text ) {
2700 return null;
2701 }
2702
2703 /*
2704 * Boolean attribute updates are just the attribute name without a corresponding value.
2705 *
2706 * This value might differ from the given comparable name in that there could be leading
2707 * or trailing whitespace, and that the casing follows the name given in `set_attribute`.
2708 *
2709 * Example:
2710 *
2711 * $p->set_attribute( 'data-TEST-id', 'update' );
2712 * 'update' === $p->get_enqueued_attribute_value( 'data-test-id' );
2713 *
2714 * Detect this difference based on the absence of the `=`, which _must_ exist in any
2715 * attribute containing a value, e.g. `<input type="text" enabled />`.
2716 * ΒΉ Β²
2717 * 1. Attribute with a string value.
2718 * 2. Boolean attribute whose value is `true`.
2719 */
2720 $equals_at = strpos( $enqueued_text, '=' );
2721 if ( false === $equals_at ) {
2722 return true;
2723 }
2724
2725 /*
2726 * Finally, a normal update's value will appear after the `=` and
2727 * be double-quoted, as performed incidentally by `set_attribute`.
2728 *
2729 * e.g. `type="text"`
2730 * ΒΉΒ² Β³
2731 * 1. Equals is here.
2732 * 2. Double-quoting starts one after the equals sign.
2733 * 3. Double-quoting ends at the last character in the update.
2734 */
2735 $enqueued_value = substr( $enqueued_text, $equals_at + 2, -1 );
2736 return WP_HTML_Decoder::decode_attribute( $enqueued_value );
2737 }
2738
2739 /**
2740 * Returns the value of a requested attribute from a matched tag opener if that attribute exists.
2741 *
2742 * Example:
2743 *
2744 * $p = new WP_HTML_Tag_Processor( '<div enabled class="test" data-test-id="14">Test</div>' );
2745 * $p->next_tag( array( 'class_name' => 'test' ) ) === true;
2746 * $p->get_attribute( 'data-test-id' ) === '14';
2747 * $p->get_attribute( 'enabled' ) === true;
2748 * $p->get_attribute( 'aria-label' ) === null;
2749 *
2750 * $p->next_tag() === false;
2751 * $p->get_attribute( 'class' ) === null;
2752 *
2753 * @since 6.2.0
2754 *
2755 * @param string $name Name of attribute whose value is requested.
2756 * @return string|true|null Value of attribute or `null` if not available. Boolean attributes return `true`.
2757 */
2758 public function get_attribute( $name ) {
2759 if ( self::STATE_MATCHED_TAG !== $this->parser_state ) {
2760 return null;
2761 }
2762
2763 $comparable = strtolower( $name );
2764
2765 /*
2766 * For every attribute other than `class` it's possible to perform a quick check if
2767 * there's an enqueued lexical update whose value takes priority over what's found in
2768 * the input document.
2769 *
2770 * The `class` attribute is special though because of the exposed helpers `add_class`
2771 * and `remove_class`. These form a builder for the `class` attribute, so an additional
2772 * check for enqueued class changes is required in addition to the check for any enqueued
2773 * attribute values. If any exist, those enqueued class changes must first be flushed out
2774 * into an attribute value update.
2775 */
2776 if ( 'class' === $name ) {
2777 $this->class_name_updates_to_attributes_updates();
2778 }
2779
2780 // Return any enqueued attribute value updates if they exist.
2781 $enqueued_value = $this->get_enqueued_attribute_value( $comparable );
2782 if ( false !== $enqueued_value ) {
2783 return $enqueued_value;
2784 }
2785
2786 if ( ! isset( $this->attributes[ $comparable ] ) ) {
2787 return null;
2788 }
2789
2790 $attribute = $this->attributes[ $comparable ];
2791
2792 /*
2793 * This flag distinguishes an attribute with no value
2794 * from an attribute with an empty string value. For
2795 * unquoted attributes this could look very similar.
2796 * It refers to whether an `=` follows the name.
2797 *
2798 * e.g. <div boolean-attribute empty-attribute=></div>
2799 * ΒΉ Β²
2800 * 1. Attribute `boolean-attribute` is `true`.
2801 * 2. Attribute `empty-attribute` is `""`.
2802 */
2803 if ( true === $attribute->is_true ) {
2804 return true;
2805 }
2806
2807 $raw_value = substr( $this->html, $attribute->value_starts_at, $attribute->value_length );
2808
2809 return WP_HTML_Decoder::decode_attribute( $raw_value );
2810 }
2811
2812 /**
2813 * Gets lowercase names of all attributes matching a given prefix in the current tag.
2814 *
2815 * Note that matching is case-insensitive. This is in accordance with the spec:
2816 *
2817 * > There must never be two or more attributes on
2818 * > the same start tag whose names are an ASCII
2819 * > case-insensitive match for each other.
2820 * - HTML 5 spec
2821 *
2822 * Example:
2823 *
2824 * $p = new WP_HTML_Tag_Processor( '<div data-ENABLED class="test" DATA-test-id="14">Test</div>' );
2825 * $p->next_tag( array( 'class_name' => 'test' ) ) === true;
2826 * $p->get_attribute_names_with_prefix( 'data-' ) === array( 'data-enabled', 'data-test-id' );
2827 *
2828 * $p->next_tag() === false;
2829 * $p->get_attribute_names_with_prefix( 'data-' ) === null;
2830 *
2831 * @since 6.2.0
2832 *
2833 * @see https://html.spec.whatwg.org/multipage/syntax.html#attributes-2:ascii-case-insensitive
2834 *
2835 * @param string $prefix Prefix of requested attribute names.
2836 * @return array|null List of attribute names, or `null` when no tag opener is matched.
2837 */
2838 public function get_attribute_names_with_prefix( $prefix ): ?array {
2839 if (
2840 self::STATE_MATCHED_TAG !== $this->parser_state ||
2841 $this->is_closing_tag
2842 ) {
2843 return null;
2844 }
2845
2846 $comparable = strtolower( $prefix );
2847
2848 $matches = array();
2849 foreach ( array_keys( $this->attributes ) as $attr_name ) {
2850 if ( str_starts_with( $attr_name, $comparable ) ) {
2851 $matches[] = $attr_name;
2852 }
2853 }
2854 return $matches;
2855 }
2856
2857 /**
2858 * Returns the namespace of the matched token.
2859 *
2860 * @since 6.7.0
2861 *
2862 * @return string One of 'html', 'math', or 'svg'.
2863 */
2864 public function get_namespace(): string {
2865 return $this->parsing_namespace;
2866 }
2867
2868 /**
2869 * Returns the uppercase name of the matched tag.
2870 *
2871 * Example:
2872 *
2873 * $p = new WP_HTML_Tag_Processor( '<div class="test">Test</div>' );
2874 * $p->next_tag() === true;
2875 * $p->get_tag() === 'DIV';
2876 *
2877 * $p->next_tag() === false;
2878 * $p->get_tag() === null;
2879 *
2880 * @since 6.2.0
2881 *
2882 * @return string|null Name of currently matched tag in input HTML, or `null` if none found.
2883 */
2884 public function get_tag(): ?string {
2885 if ( null === $this->tag_name_starts_at ) {
2886 return null;
2887 }
2888
2889 $tag_name = substr( $this->html, $this->tag_name_starts_at, $this->tag_name_length );
2890
2891 if ( self::STATE_MATCHED_TAG === $this->parser_state ) {
2892 return strtoupper( $tag_name );
2893 }
2894
2895 if (
2896 self::STATE_COMMENT === $this->parser_state &&
2897 self::COMMENT_AS_PI_NODE_LOOKALIKE === $this->get_comment_type()
2898 ) {
2899 return $tag_name;
2900 }
2901
2902 return null;
2903 }
2904
2905 /**
2906 * Returns the adjusted tag name for a given token, taking into
2907 * account the current parsing context, whether HTML, SVG, or MathML.
2908 *
2909 * @since 6.7.0
2910 *
2911 * @return string|null Name of current tag name.
2912 */
2913 public function get_qualified_tag_name(): ?string {
2914 $tag_name = $this->get_tag();
2915 if ( null === $tag_name ) {
2916 return null;
2917 }
2918
2919 if ( 'html' === $this->get_namespace() ) {
2920 return $tag_name;
2921 }
2922
2923 $lower_tag_name = strtolower( $tag_name );
2924 if ( 'math' === $this->get_namespace() ) {
2925 return $lower_tag_name;
2926 }
2927
2928 if ( 'svg' === $this->get_namespace() ) {
2929 switch ( $lower_tag_name ) {
2930 case 'altglyph':
2931 return 'altGlyph';
2932
2933 case 'altglyphdef':
2934 return 'altGlyphDef';
2935
2936 case 'altglyphitem':
2937 return 'altGlyphItem';
2938
2939 case 'animatecolor':
2940 return 'animateColor';
2941
2942 case 'animatemotion':
2943 return 'animateMotion';
2944
2945 case 'animatetransform':
2946 return 'animateTransform';
2947
2948 case 'clippath':
2949 return 'clipPath';
2950
2951 case 'feblend':
2952 return 'feBlend';
2953
2954 case 'fecolormatrix':
2955 return 'feColorMatrix';
2956
2957 case 'fecomponenttransfer':
2958 return 'feComponentTransfer';
2959
2960 case 'fecomposite':
2961 return 'feComposite';
2962
2963 case 'feconvolvematrix':
2964 return 'feConvolveMatrix';
2965
2966 case 'fediffuselighting':
2967 return 'feDiffuseLighting';
2968
2969 case 'fedisplacementmap':
2970 return 'feDisplacementMap';
2971
2972 case 'fedistantlight':
2973 return 'feDistantLight';
2974
2975 case 'fedropshadow':
2976 return 'feDropShadow';
2977
2978 case 'feflood':
2979 return 'feFlood';
2980
2981 case 'fefunca':
2982 return 'feFuncA';
2983
2984 case 'fefuncb':
2985 return 'feFuncB';
2986
2987 case 'fefuncg':
2988 return 'feFuncG';
2989
2990 case 'fefuncr':
2991 return 'feFuncR';
2992
2993 case 'fegaussianblur':
2994 return 'feGaussianBlur';
2995
2996 case 'feimage':
2997 return 'feImage';
2998
2999 case 'femerge':
3000 return 'feMerge';
3001
3002 case 'femergenode':
3003 return 'feMergeNode';
3004
3005 case 'femorphology':
3006 return 'feMorphology';
3007
3008 case 'feoffset':
3009 return 'feOffset';
3010
3011 case 'fepointlight':
3012 return 'fePointLight';
3013
3014 case 'fespecularlighting':
3015 return 'feSpecularLighting';
3016
3017 case 'fespotlight':
3018 return 'feSpotLight';
3019
3020 case 'fetile':
3021 return 'feTile';
3022
3023 case 'feturbulence':
3024 return 'feTurbulence';
3025
3026 case 'foreignobject':
3027 return 'foreignObject';
3028
3029 case 'glyphref':
3030 return 'glyphRef';
3031
3032 case 'lineargradient':
3033 return 'linearGradient';
3034
3035 case 'radialgradient':
3036 return 'radialGradient';
3037
3038 case 'textpath':
3039 return 'textPath';
3040
3041 default:
3042 return $lower_tag_name;
3043 }
3044 }
3045
3046 // This unnecessary return prevents tools from inaccurately reporting type errors.
3047 return $tag_name;
3048 }
3049
3050 /**
3051 * Returns the adjusted attribute name for a given attribute, taking into
3052 * account the current parsing context, whether HTML, SVG, or MathML.
3053 *
3054 * @since 6.7.0
3055 *
3056 * @param string $attribute_name Which attribute to adjust.
3057 *
3058 * @return string|null
3059 */
3060 public function get_qualified_attribute_name( $attribute_name ): ?string {
3061 if ( self::STATE_MATCHED_TAG !== $this->parser_state ) {
3062 return null;
3063 }
3064
3065 $namespace = $this->get_namespace();
3066 $lower_name = strtolower( $attribute_name );
3067
3068 if ( 'math' === $namespace && 'definitionurl' === $lower_name ) {
3069 return 'definitionURL';
3070 }
3071
3072 if ( 'svg' === $this->get_namespace() ) {
3073 switch ( $lower_name ) {
3074 case 'attributename':
3075 return 'attributeName';
3076
3077 case 'attributetype':
3078 return 'attributeType';
3079
3080 case 'basefrequency':
3081 return 'baseFrequency';
3082
3083 case 'baseprofile':
3084 return 'baseProfile';
3085
3086 case 'calcmode':
3087 return 'calcMode';
3088
3089 case 'clippathunits':
3090 return 'clipPathUnits';
3091
3092 case 'diffuseconstant':
3093 return 'diffuseConstant';
3094
3095 case 'edgemode':
3096 return 'edgeMode';
3097
3098 case 'filterunits':
3099 return 'filterUnits';
3100
3101 case 'glyphref':
3102 return 'glyphRef';
3103
3104 case 'gradienttransform':
3105 return 'gradientTransform';
3106
3107 case 'gradientunits':
3108 return 'gradientUnits';
3109
3110 case 'kernelmatrix':
3111 return 'kernelMatrix';
3112
3113 case 'kernelunitlength':
3114 return 'kernelUnitLength';
3115
3116 case 'keypoints':
3117 return 'keyPoints';
3118
3119 case 'keysplines':
3120 return 'keySplines';
3121
3122 case 'keytimes':
3123 return 'keyTimes';
3124
3125 case 'lengthadjust':
3126 return 'lengthAdjust';
3127
3128 case 'limitingconeangle':
3129 return 'limitingConeAngle';
3130
3131 case 'markerheight':
3132 return 'markerHeight';
3133
3134 case 'markerunits':
3135 return 'markerUnits';
3136
3137 case 'markerwidth':
3138 return 'markerWidth';
3139
3140 case 'maskcontentunits':
3141 return 'maskContentUnits';
3142
3143 case 'maskunits':
3144 return 'maskUnits';
3145
3146 case 'numoctaves':
3147 return 'numOctaves';
3148
3149 case 'pathlength':
3150 return 'pathLength';
3151
3152 case 'patterncontentunits':
3153 return 'patternContentUnits';
3154
3155 case 'patterntransform':
3156 return 'patternTransform';
3157
3158 case 'patternunits':
3159 return 'patternUnits';
3160
3161 case 'pointsatx':
3162 return 'pointsAtX';
3163
3164 case 'pointsaty':
3165 return 'pointsAtY';
3166
3167 case 'pointsatz':
3168 return 'pointsAtZ';
3169
3170 case 'preservealpha':
3171 return 'preserveAlpha';
3172
3173 case 'preserveaspectratio':
3174 return 'preserveAspectRatio';
3175
3176 case 'primitiveunits':
3177 return 'primitiveUnits';
3178
3179 case 'refx':
3180 return 'refX';
3181
3182 case 'refy':
3183 return 'refY';
3184
3185 case 'repeatcount':
3186 return 'repeatCount';
3187
3188 case 'repeatdur':
3189 return 'repeatDur';
3190
3191 case 'requiredextensions':
3192 return 'requiredExtensions';
3193
3194 case 'requiredfeatures':
3195 return 'requiredFeatures';
3196
3197 case 'specularconstant':
3198 return 'specularConstant';
3199
3200 case 'specularexponent':
3201 return 'specularExponent';
3202
3203 case 'spreadmethod':
3204 return 'spreadMethod';
3205
3206 case 'startoffset':
3207 return 'startOffset';
3208
3209 case 'stddeviation':
3210 return 'stdDeviation';
3211
3212 case 'stitchtiles':
3213 return 'stitchTiles';
3214
3215 case 'surfacescale':
3216 return 'surfaceScale';
3217
3218 case 'systemlanguage':
3219 return 'systemLanguage';
3220
3221 case 'tablevalues':
3222 return 'tableValues';
3223
3224 case 'targetx':
3225 return 'targetX';
3226
3227 case 'targety':
3228 return 'targetY';
3229
3230 case 'textlength':
3231 return 'textLength';
3232
3233 case 'viewbox':
3234 return 'viewBox';
3235
3236 case 'viewtarget':
3237 return 'viewTarget';
3238
3239 case 'xchannelselector':
3240 return 'xChannelSelector';
3241
3242 case 'ychannelselector':
3243 return 'yChannelSelector';
3244
3245 case 'zoomandpan':
3246 return 'zoomAndPan';
3247 }
3248 }
3249
3250 if ( 'html' !== $namespace ) {
3251 switch ( $lower_name ) {
3252 case 'xlink:actuate':
3253 return 'xlink actuate';
3254
3255 case 'xlink:arcrole':
3256 return 'xlink arcrole';
3257
3258 case 'xlink:href':
3259 return 'xlink href';
3260
3261 case 'xlink:role':
3262 return 'xlink role';
3263
3264 case 'xlink:show':
3265 return 'xlink show';
3266
3267 case 'xlink:title':
3268 return 'xlink title';
3269
3270 case 'xlink:type':
3271 return 'xlink type';
3272
3273 case 'xml:lang':
3274 return 'xml lang';
3275
3276 case 'xml:space':
3277 return 'xml space';
3278
3279 case 'xmlns':
3280 return 'xmlns';
3281
3282 case 'xmlns:xlink':
3283 return 'xmlns xlink';
3284 }
3285 }
3286
3287 return $attribute_name;
3288 }
3289
3290 /**
3291 * Indicates if the currently matched tag contains the self-closing flag.
3292 *
3293 * No HTML elements ought to have the self-closing flag and for those, the self-closing
3294 * flag will be ignored. For void elements this is benign because they "self close"
3295 * automatically. For non-void HTML elements though problems will appear if someone
3296 * intends to use a self-closing element in place of that element with an empty body.
3297 * For HTML foreign elements and custom elements the self-closing flag determines if
3298 * they self-close or not.
3299 *
3300 * This function does not determine if a tag is self-closing,
3301 * but only if the self-closing flag is present in the syntax.
3302 *
3303 * @since 6.3.0
3304 *
3305 * @return bool Whether the currently matched tag contains the self-closing flag.
3306 */
3307 public function has_self_closing_flag(): bool {
3308 if ( self::STATE_MATCHED_TAG !== $this->parser_state ) {
3309 return false;
3310 }
3311
3312 /*
3313 * The self-closing flag is the solidus at the _end_ of the tag, not the beginning.
3314 *
3315 * Example:
3316 *
3317 * <figure />
3318 * ^ this appears one character before the end of the closing ">".
3319 */
3320 return '/' === $this->html[ $this->token_starts_at + $this->token_length - 2 ];
3321 }
3322
3323 /**
3324 * Indicates if the current tag token is a tag closer.
3325 *
3326 * Example:
3327 *
3328 * $p = new WP_HTML_Tag_Processor( '<div></div>' );
3329 * $p->next_tag( array( 'tag_name' => 'div', 'tag_closers' => 'visit' ) );
3330 * $p->is_tag_closer() === false;
3331 *
3332 * $p->next_tag( array( 'tag_name' => 'div', 'tag_closers' => 'visit' ) );
3333 * $p->is_tag_closer() === true;
3334 *
3335 * @since 6.2.0
3336 * @since 6.7.0 Reports all BR tags as opening tags.
3337 *
3338 * @return bool Whether the current tag is a tag closer.
3339 */
3340 public function is_tag_closer(): bool {
3341 return (
3342 self::STATE_MATCHED_TAG === $this->parser_state &&
3343 $this->is_closing_tag &&
3344
3345 /*
3346 * The BR tag can only exist as an opening tag. If something like `</br>`
3347 * appears then the HTML parser will treat it as an opening tag with no
3348 * attributes. The BR tag is unique in this way.
3349 *
3350 * @see https://html.spec.whatwg.org/#parsing-main-inbody
3351 */
3352 'BR' !== $this->get_tag()
3353 );
3354 }
3355
3356 /**
3357 * Indicates the kind of matched token, if any.
3358 *
3359 * This differs from `get_token_name()` in that it always
3360 * returns a static string indicating the type, whereas
3361 * `get_token_name()` may return values derived from the
3362 * token itself, such as a tag name or processing
3363 * instruction tag.
3364 *
3365 * Possible values:
3366 * - `#tag` when matched on a tag.
3367 * - `#text` when matched on a text node.
3368 * - `#cdata-section` when matched on a CDATA node.
3369 * - `#comment` when matched on a comment.
3370 * - `#doctype` when matched on a DOCTYPE declaration.
3371 * - `#presumptuous-tag` when matched on an empty tag closer.
3372 * - `#funky-comment` when matched on a funky comment.
3373 *
3374 * @since 6.5.0
3375 *
3376 * @return string|null What kind of token is matched, or null.
3377 */
3378 public function get_token_type(): ?string {
3379 switch ( $this->parser_state ) {
3380 case self::STATE_MATCHED_TAG:
3381 return '#tag';
3382
3383 case self::STATE_DOCTYPE:
3384 return '#doctype';
3385
3386 default:
3387 return $this->get_token_name();
3388 }
3389 }
3390
3391 /**
3392 * Returns the node name represented by the token.
3393 *
3394 * This matches the DOM API value `nodeName`. Some values
3395 * are static, such as `#text` for a text node, while others
3396 * are dynamically generated from the token itself.
3397 *
3398 * Dynamic names:
3399 * - Uppercase tag name for tag matches.
3400 * - `html` for DOCTYPE declarations.
3401 *
3402 * Note that if the Tag Processor is not matched on a token
3403 * then this function will return `null`, either because it
3404 * hasn't yet found a token or because it reached the end
3405 * of the document without matching a token.
3406 *
3407 * @since 6.5.0
3408 *
3409 * @return string|null Name of the matched token.
3410 */
3411 public function get_token_name(): ?string {
3412 switch ( $this->parser_state ) {
3413 case self::STATE_MATCHED_TAG:
3414 return $this->get_tag();
3415
3416 case self::STATE_TEXT_NODE:
3417 return '#text';
3418
3419 case self::STATE_CDATA_NODE:
3420 return '#cdata-section';
3421
3422 case self::STATE_COMMENT:
3423 return '#comment';
3424
3425 case self::STATE_DOCTYPE:
3426 return 'html';
3427
3428 case self::STATE_PRESUMPTUOUS_TAG:
3429 return '#presumptuous-tag';
3430
3431 case self::STATE_FUNKY_COMMENT:
3432 return '#funky-comment';
3433 }
3434
3435 return null;
3436 }
3437
3438 /**
3439 * Indicates what kind of comment produced the comment node.
3440 *
3441 * Because there are different kinds of HTML syntax which produce
3442 * comments, the Tag Processor tracks and exposes this as a type
3443 * for the comment. Nominally only regular HTML comments exist as
3444 * they are commonly known, but a number of unrelated syntax errors
3445 * also produce comments.
3446 *
3447 * @see self::COMMENT_AS_ABRUPTLY_CLOSED_COMMENT
3448 * @see self::COMMENT_AS_CDATA_LOOKALIKE
3449 * @see self::COMMENT_AS_INVALID_HTML
3450 * @see self::COMMENT_AS_HTML_COMMENT
3451 * @see self::COMMENT_AS_PI_NODE_LOOKALIKE
3452 *
3453 * @since 6.5.0
3454 *
3455 * @return string|null
3456 */
3457 public function get_comment_type(): ?string {
3458 if ( self::STATE_COMMENT !== $this->parser_state ) {
3459 return null;
3460 }
3461
3462 return $this->comment_type;
3463 }
3464
3465 /**
3466 * Returns the text of a matched comment or null if not on a comment type node.
3467 *
3468 * This method returns the entire text content of a comment node as it
3469 * would appear in the browser.
3470 *
3471 * This differs from {@see ::get_modifiable_text()} in that certain comment
3472 * types in the HTML API cannot allow their entire comment text content to
3473 * be modified. Namely, "bogus comments" of the form `<?not allowed in html>`
3474 * will create a comment whose text content starts with `?`. Note that if
3475 * that character were modified, it would be possible to change the node
3476 * type.
3477 *
3478 * @since 6.7.0
3479 *
3480 * @return string|null The comment text as it would appear in the browser or null
3481 * if not on a comment type node.
3482 */
3483 public function get_full_comment_text(): ?string {
3484 if ( self::STATE_FUNKY_COMMENT === $this->parser_state ) {
3485 return $this->get_modifiable_text();
3486 }
3487
3488 if ( self::STATE_COMMENT !== $this->parser_state ) {
3489 return null;
3490 }
3491
3492 switch ( $this->get_comment_type() ) {
3493 case self::COMMENT_AS_HTML_COMMENT:
3494 case self::COMMENT_AS_ABRUPTLY_CLOSED_COMMENT:
3495 return $this->get_modifiable_text();
3496
3497 case self::COMMENT_AS_CDATA_LOOKALIKE:
3498 return "[CDATA[{$this->get_modifiable_text()}]]";
3499
3500 case self::COMMENT_AS_PI_NODE_LOOKALIKE:
3501 return "?{$this->get_tag()}{$this->get_modifiable_text()}?";
3502
3503 /*
3504 * This represents "bogus comments state" from HTML tokenization.
3505 * This can be entered by `<?` or `<!`, where `?` is included in
3506 * the comment text but `!` is not.
3507 */
3508 case self::COMMENT_AS_INVALID_HTML:
3509 $preceding_character = $this->html[ $this->text_starts_at - 1 ];
3510 $comment_start = '?' === $preceding_character ? '?' : '';
3511 return "{$comment_start}{$this->get_modifiable_text()}";
3512 }
3513
3514 return null;
3515 }
3516
3517 /**
3518 * Subdivides a matched text node, splitting NULL byte sequences and decoded whitespace as
3519 * distinct nodes prefixes.
3520 *
3521 * Note that once anything that's neither a NULL byte nor decoded whitespace is
3522 * encountered, then the remainder of the text node is left intact as generic text.
3523 *
3524 * - The HTML Processor uses this to apply distinct rules for different kinds of text.
3525 * - Inter-element whitespace can be detected and skipped with this method.
3526 *
3527 * Text nodes aren't eagerly subdivided because there's no need to split them unless
3528 * decisions are being made on NULL byte sequences or whitespace-only text.
3529 *
3530 * Example:
3531 *
3532 * $processor = new WP_HTML_Tag_Processor( "\x00Apples & Oranges" );
3533 * true === $processor->next_token(); // Text is "Apples & Oranges".
3534 * true === $processor->subdivide_text_appropriately(); // Text is "".
3535 * true === $processor->next_token(); // Text is "Apples & Oranges".
3536 * false === $processor->subdivide_text_appropriately();
3537 *
3538 * $processor = new WP_HTML_Tag_Processor( " \r\n\tMore" );
3539 * true === $processor->next_token(); // Text is "β€ β€βMore".
3540 * true === $processor->subdivide_text_appropriately(); // Text is "β€ β€β".
3541 * true === $processor->next_token(); // Text is "More".
3542 * false === $processor->subdivide_text_appropriately();
3543 *
3544 * @since 6.7.0
3545 *
3546 * @return bool Whether the text node was subdivided.
3547 */
3548 public function subdivide_text_appropriately(): bool {
3549 if ( self::STATE_TEXT_NODE !== $this->parser_state ) {
3550 return false;
3551 }
3552
3553 $this->text_node_classification = self::TEXT_IS_GENERIC;
3554
3555 /*
3556 * NULL bytes are treated categorically different than numeric character
3557 * references whose number is zero. `�` is not the same as `"\x00"`.
3558 */
3559 $leading_nulls = strspn( $this->html, "\x00", $this->text_starts_at, $this->text_length );
3560 if ( $leading_nulls > 0 ) {
3561 $this->token_length = $leading_nulls;
3562 $this->text_length = $leading_nulls;
3563 $this->bytes_already_parsed = $this->token_starts_at + $leading_nulls;
3564 $this->text_node_classification = self::TEXT_IS_NULL_SEQUENCE;
3565 return true;
3566 }
3567
3568 /*
3569 * Start a decoding loop to determine the point at which the
3570 * text subdivides. This entails raw whitespace bytes and any
3571 * character reference that decodes to the same.
3572 */
3573 $at = $this->text_starts_at;
3574 $end = $this->text_starts_at + $this->text_length;
3575 while ( $at < $end ) {
3576 $skipped = strspn( $this->html, " \t\f\r\n", $at, $end - $at );
3577 $at += $skipped;
3578
3579 if ( $at < $end && '&' === $this->html[ $at ] ) {
3580 $matched_byte_length = null;
3581 $replacement = WP_HTML_Decoder::read_character_reference( 'data', $this->html, $at, $matched_byte_length );
3582 if ( isset( $replacement ) && 1 === strspn( $replacement, " \t\f\r\n" ) ) {
3583 $at += $matched_byte_length;
3584 continue;
3585 }
3586 }
3587
3588 break;
3589 }
3590
3591 if ( $at > $this->text_starts_at ) {
3592 $new_length = $at - $this->text_starts_at;
3593 $this->text_length = $new_length;
3594 $this->token_length = $new_length;
3595 $this->bytes_already_parsed = $at;
3596 $this->text_node_classification = self::TEXT_IS_WHITESPACE;
3597 return true;
3598 }
3599
3600 return false;
3601 }
3602
3603 /**
3604 * Returns the modifiable text for a matched token, or an empty string.
3605 *
3606 * Modifiable text is text content that may be read and changed without
3607 * changing the HTML structure of the document around it. This includes
3608 * the contents of `#text` nodes in the HTML as well as the inner
3609 * contents of HTML comments, Processing Instructions, and others, even
3610 * though these nodes aren't part of a parsed DOM tree. They also contain
3611 * the contents of SCRIPT and STYLE tags, of TEXTAREA tags, and of any
3612 * other section in an HTML document which cannot contain HTML markup (DATA).
3613 *
3614 * If a token has no modifiable text then an empty string is returned to
3615 * avoid needless crashing or type errors. An empty string does not mean
3616 * that a token has modifiable text, and a token with modifiable text may
3617 * have an empty string (e.g. a comment with no contents).
3618 *
3619 * Limitations:
3620 *
3621 * - This function will not strip the leading newline appropriately
3622 * after seeking into a LISTING or PRE element. To ensure that the
3623 * newline is treated properly, seek to the LISTING or PRE opening
3624 * tag instead of to the first text node inside the element.
3625 *
3626 * @since 6.5.0
3627 * @since 6.7.0 Replaces NULL bytes (U+0000) and newlines appropriately.
3628 *
3629 * @return string
3630 */
3631 public function get_modifiable_text(): string {
3632 $has_enqueued_update = isset( $this->lexical_updates['modifiable text'] );
3633
3634 if ( ! $has_enqueued_update && ( null === $this->text_starts_at || 0 === $this->text_length ) ) {
3635 return '';
3636 }
3637
3638 $text = $has_enqueued_update
3639 ? $this->lexical_updates['modifiable text']->text
3640 : substr( $this->html, $this->text_starts_at, $this->text_length );
3641
3642 /*
3643 * Pre-processing the input stream would normally happen before
3644 * any parsing is done, but deferring it means it's possible to
3645 * skip in most cases. When getting the modifiable text, however
3646 * it's important to apply the pre-processing steps, which is
3647 * normalizing newlines.
3648 *
3649 * @see https://html.spec.whatwg.org/#preprocessing-the-input-stream
3650 * @see https://infra.spec.whatwg.org/#normalize-newlines
3651 */
3652 $text = str_replace( "\r\n", "\n", $text );
3653 $text = str_replace( "\r", "\n", $text );
3654
3655 // Comment data is not decoded.
3656 if (
3657 self::STATE_CDATA_NODE === $this->parser_state ||
3658 self::STATE_COMMENT === $this->parser_state ||
3659 self::STATE_DOCTYPE === $this->parser_state ||
3660 self::STATE_FUNKY_COMMENT === $this->parser_state
3661 ) {
3662 return str_replace( "\x00", "\u{FFFD}", $text );
3663 }
3664
3665 $tag_name = $this->get_token_name();
3666 if (
3667 // Script data is not decoded.
3668 'SCRIPT' === $tag_name ||
3669
3670 // RAWTEXT data is not decoded.
3671 'IFRAME' === $tag_name ||
3672 'NOEMBED' === $tag_name ||
3673 'NOFRAMES' === $tag_name ||
3674 'STYLE' === $tag_name ||
3675 'XMP' === $tag_name
3676 ) {
3677 return str_replace( "\x00", "\u{FFFD}", $text );
3678 }
3679
3680 $decoded = WP_HTML_Decoder::decode_text_node( $text );
3681
3682 /*
3683 * Skip the first line feed after LISTING, PRE, and TEXTAREA opening tags.
3684 *
3685 * Note that this first newline may come in the form of a character
3686 * reference, such as `
`, and so it's important to perform
3687 * this transformation only after decoding the raw text content.
3688 */
3689 if (
3690 ( "\n" === ( $decoded[0] ?? '' ) ) &&
3691 ( ( $this->skip_newline_at === $this->token_starts_at && '#text' === $tag_name ) || 'TEXTAREA' === $tag_name )
3692 ) {
3693 $decoded = substr( $decoded, 1 );
3694 }
3695
3696 /*
3697 * Only in normative text nodes does the NULL byte (U+0000) get removed.
3698 * In all other contexts it's replaced by the replacement character (U+FFFD)
3699 * for security reasons (to avoid joining together strings that were safe
3700 * when separated, but not when joined).
3701 *
3702 * @todo Inside HTML integration points and MathML integration points, the
3703 * text is processed according to the insertion mode, not according
3704 * to the foreign content rules. This should strip the NULL bytes.
3705 */
3706 return ( '#text' === $tag_name && 'html' === $this->get_namespace() )
3707 ? str_replace( "\x00", '', $decoded )
3708 : str_replace( "\x00", "\u{FFFD}", $decoded );
3709 }
3710
3711 /**
3712 * Sets the modifiable text for the matched token, if matched.
3713 *
3714 * Modifiable text is text content that may be read and changed without
3715 * changing the HTML structure of the document around it. This includes
3716 * the contents of `#text` nodes in the HTML as well as the inner
3717 * contents of HTML comments, Processing Instructions, and others, even
3718 * though these nodes aren't part of a parsed DOM tree. They also contain
3719 * the contents of SCRIPT and STYLE tags, of TEXTAREA tags, and of any
3720 * other section in an HTML document which cannot contain HTML markup (DATA).
3721 *
3722 * Not all modifiable text may be set by this method, and not all content
3723 * may be set as modifiable text. In the case that this fails it will return
3724 * `false` indicating as much. For instance, it will not allow inserting the
3725 * string `</script` into a SCRIPT element, because the rules for escaping
3726 * that safely are complicated. Similarly, it will not allow setting content
3727 * into a comment which would prematurely terminate the comment.
3728 *
3729 * Example:
3730 *
3731 * // Add a preface to all STYLE contents.
3732 * while ( $processor->next_tag( 'STYLE' ) ) {
3733 * $style = $processor->get_modifiable_text();
3734 * $processor->set_modifiable_text( "// Made with love on the World Wide Web\n{$style}" );
3735 * }
3736 *
3737 * // Replace smiley text with Emoji smilies.
3738 * while ( $processor->next_token() ) {
3739 * if ( '#text' !== $processor->get_token_name() ) {
3740 * continue;
3741 * }
3742 *
3743 * $chunk = $processor->get_modifiable_text();
3744 * if ( ! str_contains( $chunk, ':)' ) ) {
3745 * continue;
3746 * }
3747 *
3748 * $processor->set_modifiable_text( str_replace( ':)', 'π', $chunk ) );
3749 * }
3750 *
3751 * This function handles all necessary HTML encoding. Provide normal, unescaped string values.
3752 * The HTML API will encode the strings appropriately so that the browser will interpret them
3753 * as the intended value.
3754 *
3755 * Example:
3756 *
3757 * // Renders as βEggs & Milkβ in a browser, encoded as `<p>Eggs & Milk</p>`.
3758 * $processor->set_modifiable_text( 'Eggs & Milk' );
3759 *
3760 * // Renders as βEggs & Milkβ in a browser, encoded as `<p>Eggs &amp; Milk</p>`.
3761 * $processor->set_modifiable_text( 'Eggs & Milk' );
3762 *
3763 * @since 6.7.0
3764 * @since 6.9.0 Escapes all character references instead of trying to avoid double-escaping.
3765 *
3766 * @param string $plaintext_content New text content to represent in the matched token.
3767 * @return bool Whether the text was able to update.
3768 */
3769 public function set_modifiable_text( string $plaintext_content ): bool {
3770 if ( self::STATE_TEXT_NODE === $this->parser_state ) {
3771 $this->lexical_updates['modifiable text'] = new WP_HTML_Text_Replacement(
3772 $this->text_starts_at,
3773 $this->text_length,
3774 strtr(
3775 $plaintext_content,
3776 array(
3777 '<' => '<',
3778 '>' => '>',
3779 '&' => '&',
3780 '"' => '"',
3781 "'" => ''',
3782 )
3783 )
3784 );
3785
3786 return true;
3787 }
3788
3789 // Comment data is not encoded.
3790 if (
3791 self::STATE_COMMENT === $this->parser_state &&
3792 self::COMMENT_AS_HTML_COMMENT === $this->comment_type
3793 ) {
3794 // Check if the text could close the comment.
3795 if ( 1 === preg_match( '/--!?>/', $plaintext_content ) ) {
3796 return false;
3797 }
3798
3799 $this->lexical_updates['modifiable text'] = new WP_HTML_Text_Replacement(
3800 $this->text_starts_at,
3801 $this->text_length,
3802 $plaintext_content
3803 );
3804
3805 return true;
3806 }
3807
3808 if ( self::STATE_MATCHED_TAG !== $this->parser_state ) {
3809 return false;
3810 }
3811
3812 switch ( $this->get_tag() ) {
3813 case 'SCRIPT':
3814 /**
3815 * This is over-protective, but ensures the update doesn't break
3816 * the HTML structure of the SCRIPT element.
3817 *
3818 * More thorough analysis could track the HTML tokenizer states
3819 * and to ensure that the SCRIPT element closes at the expected
3820 * SCRIPT close tag as is done in {@see ::skip_script_data()}.
3821 *
3822 * A SCRIPT element could be closed prematurely by contents
3823 * like `</script>`. A SCRIPT element could be prevented from
3824 * closing by contents like `<!--<script>`.
3825 *
3826 * The following strings are essential for dangerous content,
3827 * although they are insufficient on their own. This trade-off
3828 * prevents dangerous scripts from being sent to the browser.
3829 * It is also unlikely to produce HTML that may confuse more
3830 * basic HTML tooling.
3831 */
3832 if (
3833 false !== stripos( $plaintext_content, '</script' ) ||
3834 false !== stripos( $plaintext_content, '<script' )
3835 ) {
3836 return false;
3837 }
3838
3839 $this->lexical_updates['modifiable text'] = new WP_HTML_Text_Replacement(
3840 $this->text_starts_at,
3841 $this->text_length,
3842 $plaintext_content
3843 );
3844
3845 return true;
3846
3847 case 'STYLE':
3848 $plaintext_content = preg_replace_callback(
3849 '~</(?P<TAG_NAME>style)~i',
3850 static function ( $tag_match ) {
3851 return "\\3c\\2f{$tag_match['TAG_NAME']}";
3852 },
3853 $plaintext_content
3854 );
3855
3856 $this->lexical_updates['modifiable text'] = new WP_HTML_Text_Replacement(
3857 $this->text_starts_at,
3858 $this->text_length,
3859 $plaintext_content
3860 );
3861
3862 return true;
3863
3864 case 'TEXTAREA':
3865 case 'TITLE':
3866 $plaintext_content = preg_replace_callback(
3867 "~</(?P<TAG_NAME>{$this->get_tag()})~i",
3868 static function ( $tag_match ) {
3869 return "</{$tag_match['TAG_NAME']}";
3870 },
3871 $plaintext_content
3872 );
3873
3874 /*
3875 * These don't _need_ to be escaped, but since they are decoded it's
3876 * safe to leave them escaped and this can prevent other code from
3877 * naively detecting tags within the contents.
3878 *
3879 * @todo It would be useful to prefix a multiline replacement text
3880 * with a newline, but not necessary. This is for aesthetics.
3881 */
3882 $this->lexical_updates['modifiable text'] = new WP_HTML_Text_Replacement(
3883 $this->text_starts_at,
3884 $this->text_length,
3885 $plaintext_content
3886 );
3887
3888 return true;
3889 }
3890
3891 return false;
3892 }
3893
3894 /**
3895 * Updates or creates a new attribute on the currently matched tag with the passed value.
3896 *
3897 * This function handles all necessary HTML encoding. Provide normal, unescaped string values.
3898 * The HTML API will encode the strings appropriately so that the browser will interpret them
3899 * as the intended value.
3900 *
3901 * Example:
3902 *
3903 * // Renders βEggs & Milkβ in a browser, encoded as `<abbr title="Eggs & Milk">`.
3904 * $processor->set_attribute( 'title', 'Eggs & Milk' );
3905 *
3906 * // Renders βEggs & Milkβ in a browser, encoded as `<abbr title="Eggs &amp; Milk">`.
3907 * $processor->set_attribute( 'title', 'Eggs & Milk' );
3908 *
3909 * // Renders `true` as `<abbr title>`.
3910 * $processor->set_attribute( 'title', true );
3911 *
3912 * // Renders without the attribute for `false` as `<abbr>`.
3913 * $processor->set_attribute( 'title', false );
3914 *
3915 * Special handling is provided for boolean attribute values:
3916 * - When `true` is passed as the value, then only the attribute name is added to the tag.
3917 * - When `false` is passed, the attribute gets removed if it existed before.
3918 *
3919 * @since 6.2.0
3920 * @since 6.2.1 Fix: Only create a single update for multiple calls with case-variant attribute names.
3921 * @since 6.9.0 Escapes all character references instead of trying to avoid double-escaping.
3922 *
3923 * @param string $name The attribute name to target.
3924 * @param string|bool $value The new attribute value.
3925 * @return bool Whether an attribute value was set.
3926 */
3927 public function set_attribute( $name, $value ): bool {
3928 if (
3929 self::STATE_MATCHED_TAG !== $this->parser_state ||
3930 $this->is_closing_tag
3931 ) {
3932 return false;
3933 }
3934
3935 $name_length = strlen( $name );
3936
3937 /**
3938 * WordPress rejects more characters than are strictly forbidden
3939 * in HTML5. This is to prevent additional security risks deeper
3940 * in the WordPress and plugin stack. Specifically the following
3941 * are not allowed to be set as part of an HTML attribute name:
3942 *
3943 * - greater-than β>β
3944 * - ampersand β&β
3945 *
3946 * @see https://html.spec.whatwg.org/#attributes-2
3947 */
3948 if (
3949 0 === $name_length ||
3950 // Syntax-like characters.
3951 strcspn( $name, '"\'>&</ =' ) !== $name_length ||
3952 // Control characters.
3953 strcspn(
3954 $name,
3955 "\x00\x01\x02\x03\x04\x05\x06\x07\x08\x09\x0A\x0B\x0C\x0D\x0E\x0F" .
3956 "\x10\x11\x12\x13\x14\x15\x16\x17\x18\x19\x1A\x1B\x1C\x1D\x1E\x1F"
3957 ) !== $name_length ||
3958 // Unicode noncharacters.
3959 wp_has_noncharacters( $name )
3960 ) {
3961 _doing_it_wrong(
3962 __METHOD__,
3963 __( 'Invalid attribute name.' ),
3964 '6.2.0'
3965 );
3966
3967 return false;
3968 }
3969
3970 /*
3971 * > The values "true" and "false" are not allowed on boolean attributes.
3972 * > To represent a false value, the attribute has to be omitted altogether.
3973 * - HTML5 spec, https://html.spec.whatwg.org/#boolean-attributes
3974 */
3975 if ( false === $value ) {
3976 return $this->remove_attribute( $name );
3977 }
3978
3979 if ( true === $value ) {
3980 $updated_attribute = $name;
3981 } else {
3982 $comparable_name = strtolower( $name );
3983
3984 /**
3985 * Escape attribute values appropriately.
3986 *
3987 * @see https://html.spec.whatwg.org/#attributes-3
3988 */
3989 $escaped_new_value = in_array( $comparable_name, wp_kses_uri_attributes(), true )
3990 ? esc_url( $value )
3991 : strtr(
3992 $value,
3993 array(
3994 '<' => '<',
3995 '>' => '>',
3996 '&' => '&',
3997 '"' => '"',
3998 "'" => ''',
3999 )
4000 );
4001
4002 // If the escaping functions wiped out the update, reject it and indicate it was rejected.
4003 if ( '' === $escaped_new_value && '' !== $value ) {
4004 return false;
4005 }
4006
4007 $updated_attribute = "{$name}=\"{$escaped_new_value}\"";
4008 }
4009
4010 /*
4011 * > There must never be two or more attributes on
4012 * > the same start tag whose names are an ASCII
4013 * > case-insensitive match for each other.
4014 * - HTML 5 spec
4015 *
4016 * @see https://html.spec.whatwg.org/multipage/syntax.html#attributes-2:ascii-case-insensitive
4017 */
4018 $comparable_name = strtolower( $name );
4019
4020 if ( isset( $this->attributes[ $comparable_name ] ) ) {
4021 /*
4022 * Update an existing attribute.
4023 *
4024 * Example β set attribute id to "new" in <div id="initial_id" />:
4025 *
4026 * <div id="initial_id"/>
4027 * ^-------------^
4028 * start end
4029 * replacement: `id="new"`
4030 *
4031 * Result: <div id="new"/>
4032 */
4033 $existing_attribute = $this->attributes[ $comparable_name ];
4034 $this->lexical_updates[ $comparable_name ] = new WP_HTML_Text_Replacement(
4035 $existing_attribute->start,
4036 $existing_attribute->length,
4037 $updated_attribute
4038 );
4039 } else {
4040 /*
4041 * Create a new attribute at the tag's name end.
4042 *
4043 * Example β add attribute id="new" to <div />:
4044 *
4045 * <div/>
4046 * ^
4047 * start and end
4048 * replacement: ` id="new"`
4049 *
4050 * Result: <div id="new"/>
4051 */
4052 $this->lexical_updates[ $comparable_name ] = new WP_HTML_Text_Replacement(
4053 $this->tag_name_starts_at + $this->tag_name_length,
4054 0,
4055 ' ' . $updated_attribute
4056 );
4057 }
4058
4059 /*
4060 * Any calls to update the `class` attribute directly should wipe out any
4061 * enqueued class changes from `add_class` and `remove_class`.
4062 */
4063 if ( 'class' === $comparable_name && ! empty( $this->classname_updates ) ) {
4064 $this->classname_updates = array();
4065 }
4066
4067 return true;
4068 }
4069
4070 /**
4071 * Remove an attribute from the currently-matched tag.
4072 *
4073 * @since 6.2.0
4074 *
4075 * @param string $name The attribute name to remove.
4076 * @return bool Whether an attribute was removed.
4077 */
4078 public function remove_attribute( $name ): bool {
4079 if (
4080 self::STATE_MATCHED_TAG !== $this->parser_state ||
4081 $this->is_closing_tag
4082 ) {
4083 return false;
4084 }
4085
4086 /*
4087 * > There must never be two or more attributes on
4088 * > the same start tag whose names are an ASCII
4089 * > case-insensitive match for each other.
4090 * - HTML 5 spec
4091 *
4092 * @see https://html.spec.whatwg.org/multipage/syntax.html#attributes-2:ascii-case-insensitive
4093 */
4094 $name = strtolower( $name );
4095
4096 /*
4097 * Any calls to update the `class` attribute directly should wipe out any
4098 * enqueued class changes from `add_class` and `remove_class`.
4099 */
4100 if ( 'class' === $name && count( $this->classname_updates ) !== 0 ) {
4101 $this->classname_updates = array();
4102 }
4103
4104 /*
4105 * If updating an attribute that didn't exist in the input
4106 * document, then remove the enqueued update and move on.
4107 *
4108 * For example, this might occur when calling `remove_attribute()`
4109 * after calling `set_attribute()` for the same attribute
4110 * and when that attribute wasn't originally present.
4111 */
4112 if ( ! isset( $this->attributes[ $name ] ) ) {
4113 if ( isset( $this->lexical_updates[ $name ] ) ) {
4114 unset( $this->lexical_updates[ $name ] );
4115 }
4116 return false;
4117 }
4118
4119 /*
4120 * Removes an existing tag attribute.
4121 *
4122 * Example β remove the attribute id from <div id="main"/>:
4123 * <div id="initial_id"/>
4124 * ^-------------^
4125 * start end
4126 * replacement: ``
4127 *
4128 * Result: <div />
4129 */
4130 $this->lexical_updates[ $name ] = new WP_HTML_Text_Replacement(
4131 $this->attributes[ $name ]->start,
4132 $this->attributes[ $name ]->length,
4133 ''
4134 );
4135
4136 // Removes any duplicated attributes if they were also present.
4137 foreach ( $this->duplicate_attributes[ $name ] ?? array() as $attribute_token ) {
4138 $this->lexical_updates[] = new WP_HTML_Text_Replacement(
4139 $attribute_token->start,
4140 $attribute_token->length,
4141 ''
4142 );
4143 }
4144
4145 return true;
4146 }
4147
4148 /**
4149 * Adds a new class name to the currently matched tag.
4150 *
4151 * @since 6.2.0
4152 *
4153 * @param string $class_name The class name to add.
4154 * @return bool Whether the class was set to be added.
4155 */
4156 public function add_class( $class_name ): bool {
4157 if (
4158 self::STATE_MATCHED_TAG !== $this->parser_state ||
4159 $this->is_closing_tag
4160 ) {
4161 return false;
4162 }
4163
4164 if ( self::QUIRKS_MODE !== $this->compat_mode ) {
4165 $this->classname_updates[ $class_name ] = self::ADD_CLASS;
4166 return true;
4167 }
4168
4169 /*
4170 * Because class names are matched ASCII-case-insensitively in quirks mode,
4171 * this needs to see if a case variant of the given class name is already
4172 * enqueued and update that existing entry, if so. This picks the casing of
4173 * the first-provided class name for all lexical variations.
4174 */
4175 $class_name_length = strlen( $class_name );
4176 foreach ( $this->classname_updates as $updated_name => $action ) {
4177 if (
4178 strlen( $updated_name ) === $class_name_length &&
4179 0 === substr_compare( $updated_name, $class_name, 0, $class_name_length, true )
4180 ) {
4181 $this->classname_updates[ $updated_name ] = self::ADD_CLASS;
4182 return true;
4183 }
4184 }
4185
4186 $this->classname_updates[ $class_name ] = self::ADD_CLASS;
4187 return true;
4188 }
4189
4190 /**
4191 * Removes a class name from the currently matched tag.
4192 *
4193 * @since 6.2.0
4194 *
4195 * @param string $class_name The class name to remove.
4196 * @return bool Whether the class was set to be removed.
4197 */
4198 public function remove_class( $class_name ): bool {
4199 if (
4200 self::STATE_MATCHED_TAG !== $this->parser_state ||
4201 $this->is_closing_tag
4202 ) {
4203 return false;
4204 }
4205
4206 if ( self::QUIRKS_MODE !== $this->compat_mode ) {
4207 $this->classname_updates[ $class_name ] = self::REMOVE_CLASS;
4208 return true;
4209 }
4210
4211 /*
4212 * Because class names are matched ASCII-case-insensitively in quirks mode,
4213 * this needs to see if a case variant of the given class name is already
4214 * enqueued and update that existing entry, if so. This picks the casing of
4215 * the first-provided class name for all lexical variations.
4216 */
4217 $class_name_length = strlen( $class_name );
4218 foreach ( $this->classname_updates as $updated_name => $action ) {
4219 if (
4220 strlen( $updated_name ) === $class_name_length &&
4221 0 === substr_compare( $updated_name, $class_name, 0, $class_name_length, true )
4222 ) {
4223 $this->classname_updates[ $updated_name ] = self::REMOVE_CLASS;
4224 return true;
4225 }
4226 }
4227
4228 $this->classname_updates[ $class_name ] = self::REMOVE_CLASS;
4229 return true;
4230 }
4231
4232 /**
4233 * Returns the string representation of the HTML Tag Processor.
4234 *
4235 * @since 6.2.0
4236 *
4237 * @see WP_HTML_Tag_Processor::get_updated_html()
4238 *
4239 * @return string The processed HTML.
4240 */
4241 public function __toString(): string {
4242 return $this->get_updated_html();
4243 }
4244
4245 /**
4246 * Returns the string representation of the HTML Tag Processor.
4247 *
4248 * @since 6.2.0
4249 * @since 6.2.1 Shifts the internal cursor corresponding to the applied updates.
4250 * @since 6.4.0 No longer calls subclass method `next_tag()` after updating HTML.
4251 *
4252 * @return string The processed HTML.
4253 */
4254 public function get_updated_html(): string {
4255 $requires_no_updating = 0 === count( $this->classname_updates ) && 0 === count( $this->lexical_updates );
4256
4257 /*
4258 * When there is nothing more to update and nothing has already been
4259 * updated, return the original document and avoid a string copy.
4260 */
4261 if ( $requires_no_updating ) {
4262 return $this->html;
4263 }
4264
4265 /*
4266 * Keep track of the position right before the current tag. This will
4267 * be necessary for reparsing the current tag after updating the HTML.
4268 */
4269 $before_current_tag = $this->token_starts_at ?? 0;
4270
4271 /*
4272 * 1. Apply the enqueued edits and update all the pointers to reflect those changes.
4273 */
4274 $this->class_name_updates_to_attributes_updates();
4275 $before_current_tag += $this->apply_attributes_updates( $before_current_tag );
4276
4277 /*
4278 * 2. Rewind to before the current tag and reparse to get updated attributes.
4279 *
4280 * At this point the internal cursor points to the end of the tag name.
4281 * Rewind before the tag name starts so that it's as if the cursor didn't
4282 * move; a call to `next_tag()` will reparse the recently-updated attributes
4283 * and additional calls to modify the attributes will apply at this same
4284 * location, but in order to avoid issues with subclasses that might add
4285 * behaviors to `next_tag()`, the internal methods should be called here
4286 * instead.
4287 *
4288 * It's important to note that in this specific place there will be no change
4289 * because the processor was already at a tag when this was called and it's
4290 * rewinding only to the beginning of this very tag before reprocessing it
4291 * and its attributes.
4292 *
4293 * <p>Previous HTML<em>More HTML</em></p>
4294 * β β back up by the length of the tag name plus the opening <
4295 * ββββ back up by strlen("em") + 1 ==> 3
4296 */
4297 $this->bytes_already_parsed = $before_current_tag;
4298 $this->base_class_next_token();
4299
4300 return $this->html;
4301 }
4302
4303 /**
4304 * Parses tag query input into internal search criteria.
4305 *
4306 * @since 6.2.0
4307 *
4308 * @param array|string|null $query {
4309 * Optional. Which tag name to find, having which class, etc. Default is to find any tag.
4310 *
4311 * @type string|null $tag_name Which tag to find, or `null` for "any tag."
4312 * @type int|null $match_offset Find the Nth tag matching all search criteria.
4313 * 1 for "first" tag, 3 for "third," etc.
4314 * Defaults to first tag.
4315 * @type string|null $class_name Tag must contain this class name to match.
4316 * @type string $tag_closers "visit" or "skip": whether to stop on tag closers, e.g. </div>.
4317 * }
4318 */
4319 private function parse_query( $query ) {
4320 if ( null !== $query && $query === $this->last_query ) {
4321 return;
4322 }
4323
4324 $this->last_query = $query;
4325 $this->sought_tag_name = null;
4326 $this->sought_class_name = null;
4327 $this->sought_match_offset = 1;
4328 $this->stop_on_tag_closers = false;
4329
4330 // A single string value means "find the tag of this name".
4331 if ( is_string( $query ) ) {
4332 $this->sought_tag_name = $query;
4333 return;
4334 }
4335
4336 // An empty query parameter applies no restrictions on the search.
4337 if ( null === $query ) {
4338 return;
4339 }
4340
4341 // If not using the string interface, an associative array is required.
4342 if ( ! is_array( $query ) ) {
4343 _doing_it_wrong(
4344 __METHOD__,
4345 __( 'The query argument must be an array or a tag name.' ),
4346 '6.2.0'
4347 );
4348 return;
4349 }
4350
4351 if ( isset( $query['tag_name'] ) && is_string( $query['tag_name'] ) ) {
4352 $this->sought_tag_name = $query['tag_name'];
4353 }
4354
4355 if ( isset( $query['class_name'] ) && is_string( $query['class_name'] ) ) {
4356 $this->sought_class_name = $query['class_name'];
4357 }
4358
4359 if ( isset( $query['match_offset'] ) && is_int( $query['match_offset'] ) && 0 < $query['match_offset'] ) {
4360 $this->sought_match_offset = $query['match_offset'];
4361 }
4362
4363 if ( isset( $query['tag_closers'] ) ) {
4364 $this->stop_on_tag_closers = 'visit' === $query['tag_closers'];
4365 }
4366 }
4367
4368
4369 /**
4370 * Checks whether a given tag and its attributes match the search criteria.
4371 *
4372 * @since 6.2.0
4373 *
4374 * @return bool Whether the given tag and its attribute match the search criteria.
4375 */
4376 private function matches(): bool {
4377 if ( $this->is_closing_tag && ! $this->stop_on_tag_closers ) {
4378 return false;
4379 }
4380
4381 // Does the tag name match the requested tag name in a case-insensitive manner?
4382 if (
4383 isset( $this->sought_tag_name ) &&
4384 (
4385 strlen( $this->sought_tag_name ) !== $this->tag_name_length ||
4386 0 !== substr_compare( $this->html, $this->sought_tag_name, $this->tag_name_starts_at, $this->tag_name_length, true )
4387 )
4388 ) {
4389 return false;
4390 }
4391
4392 if ( null !== $this->sought_class_name && ! $this->has_class( $this->sought_class_name ) ) {
4393 return false;
4394 }
4395
4396 return true;
4397 }
4398
4399 /**
4400 * Gets DOCTYPE declaration info from a DOCTYPE token.
4401 *
4402 * DOCTYPE tokens may appear in many places in an HTML document. In most places, they are
4403 * simply ignored. The main parsing functions find the basic shape of DOCTYPE tokens but
4404 * do not perform detailed parsing.
4405 *
4406 * This method can be called to perform a full parse of the DOCTYPE token and retrieve
4407 * its information.
4408 *
4409 * @return WP_HTML_Doctype_Info|null The DOCTYPE declaration information or `null` if not
4410 * currently at a DOCTYPE node.
4411 */
4412 public function get_doctype_info(): ?WP_HTML_Doctype_Info {
4413 if ( self::STATE_DOCTYPE !== $this->parser_state ) {
4414 return null;
4415 }
4416
4417 return WP_HTML_Doctype_Info::from_doctype_token( substr( $this->html, $this->token_starts_at, $this->token_length ) );
4418 }
4419
4420 /**
4421 * Parser Ready State.
4422 *
4423 * Indicates that the parser is ready to run and waiting for a state transition.
4424 * It may not have started yet, or it may have just finished parsing a token and
4425 * is ready to find the next one.
4426 *
4427 * @since 6.5.0
4428 *
4429 * @access private
4430 */
4431 const STATE_READY = 'STATE_READY';
4432
4433 /**
4434 * Parser Complete State.
4435 *
4436 * Indicates that the parser has reached the end of the document and there is
4437 * nothing left to scan. It finished parsing the last token completely.
4438 *
4439 * @since 6.5.0
4440 *
4441 * @access private
4442 */
4443 const STATE_COMPLETE = 'STATE_COMPLETE';
4444
4445 /**
4446 * Parser Incomplete Input State.
4447 *
4448 * Indicates that the parser has reached the end of the document before finishing
4449 * a token. It started parsing a token but there is a possibility that the input
4450 * HTML document was truncated in the middle of a token.
4451 *
4452 * The parser is reset at the start of the incomplete token and has paused. There
4453 * is nothing more than can be scanned unless provided a more complete document.
4454 *
4455 * @since 6.5.0
4456 *
4457 * @access private
4458 */
4459 const STATE_INCOMPLETE_INPUT = 'STATE_INCOMPLETE_INPUT';
4460
4461 /**
4462 * Parser Matched Tag State.
4463 *
4464 * Indicates that the parser has found an HTML tag and it's possible to get
4465 * the tag name and read or modify its attributes (if it's not a closing tag).
4466 *
4467 * @since 6.5.0
4468 *
4469 * @access private
4470 */
4471 const STATE_MATCHED_TAG = 'STATE_MATCHED_TAG';
4472
4473 /**
4474 * Parser Text Node State.
4475 *
4476 * Indicates that the parser has found a text node and it's possible
4477 * to read and modify that text.
4478 *
4479 * @since 6.5.0
4480 *
4481 * @access private
4482 */
4483 const STATE_TEXT_NODE = 'STATE_TEXT_NODE';
4484
4485 /**
4486 * Parser CDATA Node State.
4487 *
4488 * Indicates that the parser has found a CDATA node and it's possible
4489 * to read and modify its modifiable text. Note that in HTML there are
4490 * no CDATA nodes outside of foreign content (SVG and MathML). Outside
4491 * of foreign content, they are treated as HTML comments.
4492 *
4493 * @since 6.5.0
4494 *
4495 * @access private
4496 */
4497 const STATE_CDATA_NODE = 'STATE_CDATA_NODE';
4498
4499 /**
4500 * Indicates that the parser has found an HTML comment and it's
4501 * possible to read and modify its modifiable text.
4502 *
4503 * @since 6.5.0
4504 *
4505 * @access private
4506 */
4507 const STATE_COMMENT = 'STATE_COMMENT';
4508
4509 /**
4510 * Indicates that the parser has found a DOCTYPE node and it's
4511 * possible to read its DOCTYPE information via `get_doctype_info()`.
4512 *
4513 * @since 6.5.0
4514 *
4515 * @access private
4516 */
4517 const STATE_DOCTYPE = 'STATE_DOCTYPE';
4518
4519 /**
4520 * Indicates that the parser has found an empty tag closer `</>`.
4521 *
4522 * Note that in HTML there are no empty tag closers, and they
4523 * are ignored. Nonetheless, the Tag Processor still
4524 * recognizes them as they appear in the HTML stream.
4525 *
4526 * These were historically discussed as a "presumptuous tag
4527 * closer," which would close the nearest open tag, but were
4528 * dismissed in favor of explicitly-closing tags.
4529 *
4530 * @since 6.5.0
4531 *
4532 * @access private
4533 */
4534 const STATE_PRESUMPTUOUS_TAG = 'STATE_PRESUMPTUOUS_TAG';
4535
4536 /**
4537 * Indicates that the parser has found a "funky comment"
4538 * and it's possible to read and modify its modifiable text.
4539 *
4540 * Example:
4541 *
4542 * </%url>
4543 * </{"wp-bit":"query/post-author"}>
4544 * </2>
4545 *
4546 * Funky comments are tag closers with invalid tag names. Note
4547 * that in HTML these are turn into bogus comments. Nonetheless,
4548 * the Tag Processor recognizes them in a stream of HTML and
4549 * exposes them for inspection and modification.
4550 *
4551 * @since 6.5.0
4552 *
4553 * @access private
4554 */
4555 const STATE_FUNKY_COMMENT = 'STATE_WP_FUNKY';
4556
4557 /**
4558 * Indicates that a comment was created when encountering abruptly-closed HTML comment.
4559 *
4560 * Example:
4561 *
4562 * <!-->
4563 * <!--->
4564 *
4565 * @since 6.5.0
4566 */
4567 const COMMENT_AS_ABRUPTLY_CLOSED_COMMENT = 'COMMENT_AS_ABRUPTLY_CLOSED_COMMENT';
4568
4569 /**
4570 * Indicates that a comment would be parsed as a CDATA node,
4571 * were HTML to allow CDATA nodes outside of foreign content.
4572 *
4573 * Example:
4574 *
4575 * <![CDATA[This is a CDATA node.]]>
4576 *
4577 * This is an HTML comment, but it looks like a CDATA node.
4578 *
4579 * @since 6.5.0
4580 */
4581 const COMMENT_AS_CDATA_LOOKALIKE = 'COMMENT_AS_CDATA_LOOKALIKE';
4582
4583 /**
4584 * Indicates that a comment was created when encountering
4585 * normative HTML comment syntax.
4586 *
4587 * Example:
4588 *
4589 * <!-- this is a comment -->
4590 *
4591 * @since 6.5.0
4592 */
4593 const COMMENT_AS_HTML_COMMENT = 'COMMENT_AS_HTML_COMMENT';
4594
4595 /**
4596 * Indicates that a comment would be parsed as a Processing
4597 * Instruction node, were they to exist within HTML.
4598 *
4599 * Example:
4600 *
4601 * <?wp __( 'Like' ) ?>
4602 *
4603 * This is an HTML comment, but it looks like a CDATA node.
4604 *
4605 * @since 6.5.0
4606 */
4607 const COMMENT_AS_PI_NODE_LOOKALIKE = 'COMMENT_AS_PI_NODE_LOOKALIKE';
4608
4609 /**
4610 * Indicates that a comment was created when encountering invalid
4611 * HTML input, a so-called "bogus comment."
4612 *
4613 * Example:
4614 *
4615 * <?nothing special>
4616 * <!{nothing special}>
4617 *
4618 * @since 6.5.0
4619 */
4620 const COMMENT_AS_INVALID_HTML = 'COMMENT_AS_INVALID_HTML';
4621
4622 /**
4623 * No-quirks mode document compatibility mode.
4624 *
4625 * > In no-quirks mode, the behavior is (hopefully) the desired behavior
4626 * > described by the modern HTML and CSS specifications.
4627 *
4628 * @see self::$compat_mode
4629 * @see https://developer.mozilla.org/en-US/docs/Web/HTML/Quirks_Mode_and_Standards_Mode
4630 *
4631 * @since 6.7.0
4632 *
4633 * @var string
4634 */
4635 const NO_QUIRKS_MODE = 'no-quirks-mode';
4636
4637 /**
4638 * Quirks mode document compatibility mode.
4639 *
4640 * > In quirks mode, layout emulates behavior in Navigator 4 and Internet
4641 * > Explorer 5. This is essential in order to support websites that were
4642 * > built before the widespread adoption of web standards.
4643 *
4644 * @see self::$compat_mode
4645 * @see https://developer.mozilla.org/en-US/docs/Web/HTML/Quirks_Mode_and_Standards_Mode
4646 *
4647 * @since 6.7.0
4648 *
4649 * @var string
4650 */
4651 const QUIRKS_MODE = 'quirks-mode';
4652
4653 /**
4654 * Indicates that a span of text may contain any combination of significant
4655 * kinds of characters: NULL bytes, whitespace, and others.
4656 *
4657 * @see self::$text_node_classification
4658 * @see self::subdivide_text_appropriately
4659 *
4660 * @since 6.7.0
4661 */
4662 const TEXT_IS_GENERIC = 'TEXT_IS_GENERIC';
4663
4664 /**
4665 * Indicates that a span of text comprises a sequence only of NULL bytes.
4666 *
4667 * @see self::$text_node_classification
4668 * @see self::subdivide_text_appropriately
4669 *
4670 * @since 6.7.0
4671 */
4672 const TEXT_IS_NULL_SEQUENCE = 'TEXT_IS_NULL_SEQUENCE';
4673
4674 /**
4675 * Indicates that a span of decoded text comprises only whitespace.
4676 *
4677 * @see self::$text_node_classification
4678 * @see self::subdivide_text_appropriately
4679 *
4680 * @since 6.7.0
4681 */
4682 const TEXT_IS_WHITESPACE = 'TEXT_IS_WHITESPACE';
4683
4684 /**
4685 * Wakeup magic method.
4686 *
4687 * @since 6.9.2
4688 */
4689 public function __wakeup() {
4690 throw new \LogicException( __CLASS__ . ' should never be unserialized' );
4691 }
4692}
4693