MDL-58548 lib: Update htmlpurifier to the latest version.
[moodle.git] / lib / htmlpurifier / HTMLPurifier / Strategy / MakeWellFormed.php
1 <?php
3 /**
4  * Takes tokens makes them well-formed (balance end tags, etc.)
5  *
6  * Specification of the armor attributes this strategy uses:
7  *
8  *      - MakeWellFormed_TagClosedError: This armor field is used to
9  *        suppress tag closed errors for certain tokens [TagClosedSuppress],
10  *        in particular, if a tag was generated automatically by HTML
11  *        Purifier, we may rely on our infrastructure to close it for us
12  *        and shouldn't report an error to the user [TagClosedAuto].
13  */
14 class HTMLPurifier_Strategy_MakeWellFormed extends HTMLPurifier_Strategy
15 {
17     /**
18      * Array stream of tokens being processed.
19      * @type HTMLPurifier_Token[]
20      */
21     protected $tokens;
23     /**
24      * Current token.
25      * @type HTMLPurifier_Token
26      */
27     protected $token;
29     /**
30      * Zipper managing the true state.
31      * @type HTMLPurifier_Zipper
32      */
33     protected $zipper;
35     /**
36      * Current nesting of elements.
37      * @type array
38      */
39     protected $stack;
41     /**
42      * Injectors active in this stream processing.
43      * @type HTMLPurifier_Injector[]
44      */
45     protected $injectors;
47     /**
48      * Current instance of HTMLPurifier_Config.
49      * @type HTMLPurifier_Config
50      */
51     protected $config;
53     /**
54      * Current instance of HTMLPurifier_Context.
55      * @type HTMLPurifier_Context
56      */
57     protected $context;
59     /**
60      * @param HTMLPurifier_Token[] $tokens
61      * @param HTMLPurifier_Config $config
62      * @param HTMLPurifier_Context $context
63      * @return HTMLPurifier_Token[]
64      * @throws HTMLPurifier_Exception
65      */
66     public function execute($tokens, $config, $context)
67     {
68         $definition = $config->getHTMLDefinition();
70         // local variables
71         $generator = new HTMLPurifier_Generator($config, $context);
72         $escape_invalid_tags = $config->get('Core.EscapeInvalidTags');
73         // used for autoclose early abortion
74         $global_parent_allowed_elements = $definition->info_parent_def->child->getAllowedElements($config);
75         $e = $context->get('ErrorCollector', true);
76         $i = false; // injector index
77         list($zipper, $token) = HTMLPurifier_Zipper::fromArray($tokens);
78         if ($token === NULL) {
79             return array();
80         }
81         $reprocess = false; // whether or not to reprocess the same token
82         $stack = array();
84         // member variables
85         $this->stack =& $stack;
86         $this->tokens =& $tokens;
87         $this->token =& $token;
88         $this->zipper =& $zipper;
89         $this->config = $config;
90         $this->context = $context;
92         // context variables
93         $context->register('CurrentNesting', $stack);
94         $context->register('InputZipper', $zipper);
95         $context->register('CurrentToken', $token);
97         // -- begin INJECTOR --
99         $this->injectors = array();
101         $injectors = $config->getBatch('AutoFormat');
102         $def_injectors = $definition->info_injector;
103         $custom_injectors = $injectors['Custom'];
104         unset($injectors['Custom']); // special case
105         foreach ($injectors as $injector => $b) {
106             // XXX: Fix with a legitimate lookup table of enabled filters
107             if (strpos($injector, '.') !== false) {
108                 continue;
109             }
110             $injector = "HTMLPurifier_Injector_$injector";
111             if (!$b) {
112                 continue;
113             }
114             $this->injectors[] = new $injector;
115         }
116         foreach ($def_injectors as $injector) {
117             // assumed to be objects
118             $this->injectors[] = $injector;
119         }
120         foreach ($custom_injectors as $injector) {
121             if (!$injector) {
122                 continue;
123             }
124             if (is_string($injector)) {
125                 $injector = "HTMLPurifier_Injector_$injector";
126                 $injector = new $injector;
127             }
128             $this->injectors[] = $injector;
129         }
131         // give the injectors references to the definition and context
132         // variables for performance reasons
133         foreach ($this->injectors as $ix => $injector) {
134             $error = $injector->prepare($config, $context);
135             if (!$error) {
136                 continue;
137             }
138             array_splice($this->injectors, $ix, 1); // rm the injector
139             trigger_error("Cannot enable {$injector->name} injector because $error is not allowed", E_USER_WARNING);
140         }
142         // -- end INJECTOR --
144         // a note on reprocessing:
145         //      In order to reduce code duplication, whenever some code needs
146         //      to make HTML changes in order to make things "correct", the
147         //      new HTML gets sent through the purifier, regardless of its
148         //      status. This means that if we add a start token, because it
149         //      was totally necessary, we don't have to update nesting; we just
150         //      punt ($reprocess = true; continue;) and it does that for us.
152         // isset is in loop because $tokens size changes during loop exec
153         for (;;
154              // only increment if we don't need to reprocess
155              $reprocess ? $reprocess = false : $token = $zipper->next($token)) {
157             // check for a rewind
158             if (is_int($i)) {
159                 // possibility: disable rewinding if the current token has a
160                 // rewind set on it already. This would offer protection from
161                 // infinite loop, but might hinder some advanced rewinding.
162                 $rewind_offset = $this->injectors[$i]->getRewindOffset();
163                 if (is_int($rewind_offset)) {
164                     for ($j = 0; $j < $rewind_offset; $j++) {
165                         if (empty($zipper->front)) break;
166                         $token = $zipper->prev($token);
167                         // indicate that other injectors should not process this token,
168                         // but we need to reprocess it.  See Note [Injector skips]
169                         unset($token->skip[$i]);
170                         $token->rewind = $i;
171                         if ($token instanceof HTMLPurifier_Token_Start) {
172                             array_pop($this->stack);
173                         } elseif ($token instanceof HTMLPurifier_Token_End) {
174                             $this->stack[] = $token->start;
175                         }
176                     }
177                 }
178                 $i = false;
179             }
181             // handle case of document end
182             if ($token === NULL) {
183                 // kill processing if stack is empty
184                 if (empty($this->stack)) {
185                     break;
186                 }
188                 // peek
189                 $top_nesting = array_pop($this->stack);
190                 $this->stack[] = $top_nesting;
192                 // send error [TagClosedSuppress]
193                 if ($e && !isset($top_nesting->armor['MakeWellFormed_TagClosedError'])) {
194                     $e->send(E_NOTICE, 'Strategy_MakeWellFormed: Tag closed by document end', $top_nesting);
195                 }
197                 // append, don't splice, since this is the end
198                 $token = new HTMLPurifier_Token_End($top_nesting->name);
200                 // punt!
201                 $reprocess = true;
202                 continue;
203             }
205             //echo '<br>'; printZipper($zipper, $token);//printTokens($this->stack);
206             //flush();
208             // quick-check: if it's not a tag, no need to process
209             if (empty($token->is_tag)) {
210                 if ($token instanceof HTMLPurifier_Token_Text) {
211                     foreach ($this->injectors as $i => $injector) {
212                         if (isset($token->skip[$i])) {
213                             // See Note [Injector skips]
214                             continue;
215                         }
216                         if ($token->rewind !== null && $token->rewind !== $i) {
217                             continue;
218                         }
219                         // XXX fuckup
220                         $r = $token;
221                         $injector->handleText($r);
222                         $token = $this->processToken($r, $i);
223                         $reprocess = true;
224                         break;
225                     }
226                 }
227                 // another possibility is a comment
228                 continue;
229             }
231             if (isset($definition->info[$token->name])) {
232                 $type = $definition->info[$token->name]->child->type;
233             } else {
234                 $type = false; // Type is unknown, treat accordingly
235             }
237             // quick tag checks: anything that's *not* an end tag
238             $ok = false;
239             if ($type === 'empty' && $token instanceof HTMLPurifier_Token_Start) {
240                 // claims to be a start tag but is empty
241                 $token = new HTMLPurifier_Token_Empty(
242                     $token->name,
243                     $token->attr,
244                     $token->line,
245                     $token->col,
246                     $token->armor
247                 );
248                 $ok = true;
249             } elseif ($type && $type !== 'empty' && $token instanceof HTMLPurifier_Token_Empty) {
250                 // claims to be empty but really is a start tag
251                 // NB: this assignment is required
252                 $old_token = $token;
253                 $token = new HTMLPurifier_Token_End($token->name);
254                 $token = $this->insertBefore(
255                     new HTMLPurifier_Token_Start($old_token->name, $old_token->attr, $old_token->line, $old_token->col, $old_token->armor)
256                 );
257                 // punt (since we had to modify the input stream in a non-trivial way)
258                 $reprocess = true;
259                 continue;
260             } elseif ($token instanceof HTMLPurifier_Token_Empty) {
261                 // real empty token
262                 $ok = true;
263             } elseif ($token instanceof HTMLPurifier_Token_Start) {
264                 // start tag
266                 // ...unless they also have to close their parent
267                 if (!empty($this->stack)) {
269                     // Performance note: you might think that it's rather
270                     // inefficient, recalculating the autoclose information
271                     // for every tag that a token closes (since when we
272                     // do an autoclose, we push a new token into the
273                     // stream and then /process/ that, before
274                     // re-processing this token.)  But this is
275                     // necessary, because an injector can make an
276                     // arbitrary transformations to the autoclosing
277                     // tokens we introduce, so things may have changed
278                     // in the meantime.  Also, doing the inefficient thing is
279                     // "easy" to reason about (for certain perverse definitions
280                     // of "easy")
282                     $parent = array_pop($this->stack);
283                     $this->stack[] = $parent;
285                     $parent_def = null;
286                     $parent_elements = null;
287                     $autoclose = false;
288                     if (isset($definition->info[$parent->name])) {
289                         $parent_def = $definition->info[$parent->name];
290                         $parent_elements = $parent_def->child->getAllowedElements($config);
291                         $autoclose = !isset($parent_elements[$token->name]);
292                     }
294                     if ($autoclose && $definition->info[$token->name]->wrap) {
295                         // Check if an element can be wrapped by another
296                         // element to make it valid in a context (for
297                         // example, <ul><ul> needs a <li> in between)
298                         $wrapname = $definition->info[$token->name]->wrap;
299                         $wrapdef = $definition->info[$wrapname];
300                         $elements = $wrapdef->child->getAllowedElements($config);
301                         if (isset($elements[$token->name]) && isset($parent_elements[$wrapname])) {
302                             $newtoken = new HTMLPurifier_Token_Start($wrapname);
303                             $token = $this->insertBefore($newtoken);
304                             $reprocess = true;
305                             continue;
306                         }
307                     }
309                     $carryover = false;
310                     if ($autoclose && $parent_def->formatting) {
311                         $carryover = true;
312                     }
314                     if ($autoclose) {
315                         // check if this autoclose is doomed to fail
316                         // (this rechecks $parent, which his harmless)
317                         $autoclose_ok = isset($global_parent_allowed_elements[$token->name]);
318                         if (!$autoclose_ok) {
319                             foreach ($this->stack as $ancestor) {
320                                 $elements = $definition->info[$ancestor->name]->child->getAllowedElements($config);
321                                 if (isset($elements[$token->name])) {
322                                     $autoclose_ok = true;
323                                     break;
324                                 }
325                                 if ($definition->info[$token->name]->wrap) {
326                                     $wrapname = $definition->info[$token->name]->wrap;
327                                     $wrapdef = $definition->info[$wrapname];
328                                     $wrap_elements = $wrapdef->child->getAllowedElements($config);
329                                     if (isset($wrap_elements[$token->name]) && isset($elements[$wrapname])) {
330                                         $autoclose_ok = true;
331                                         break;
332                                     }
333                                 }
334                             }
335                         }
336                         if ($autoclose_ok) {
337                             // errors need to be updated
338                             $new_token = new HTMLPurifier_Token_End($parent->name);
339                             $new_token->start = $parent;
340                             // [TagClosedSuppress]
341                             if ($e && !isset($parent->armor['MakeWellFormed_TagClosedError'])) {
342                                 if (!$carryover) {
343                                     $e->send(E_NOTICE, 'Strategy_MakeWellFormed: Tag auto closed', $parent);
344                                 } else {
345                                     $e->send(E_NOTICE, 'Strategy_MakeWellFormed: Tag carryover', $parent);
346                                 }
347                             }
348                             if ($carryover) {
349                                 $element = clone $parent;
350                                 // [TagClosedAuto]
351                                 $element->armor['MakeWellFormed_TagClosedError'] = true;
352                                 $element->carryover = true;
353                                 $token = $this->processToken(array($new_token, $token, $element));
354                             } else {
355                                 $token = $this->insertBefore($new_token);
356                             }
357                         } else {
358                             $token = $this->remove();
359                         }
360                         $reprocess = true;
361                         continue;
362                     }
364                 }
365                 $ok = true;
366             }
368             if ($ok) {
369                 foreach ($this->injectors as $i => $injector) {
370                     if (isset($token->skip[$i])) {
371                         // See Note [Injector skips]
372                         continue;
373                     }
374                     if ($token->rewind !== null && $token->rewind !== $i) {
375                         continue;
376                     }
377                     $r = $token;
378                     $injector->handleElement($r);
379                     $token = $this->processToken($r, $i);
380                     $reprocess = true;
381                     break;
382                 }
383                 if (!$reprocess) {
384                     // ah, nothing interesting happened; do normal processing
385                     if ($token instanceof HTMLPurifier_Token_Start) {
386                         $this->stack[] = $token;
387                     } elseif ($token instanceof HTMLPurifier_Token_End) {
388                         throw new HTMLPurifier_Exception(
389                             'Improper handling of end tag in start code; possible error in MakeWellFormed'
390                         );
391                     }
392                 }
393                 continue;
394             }
396             // sanity check: we should be dealing with a closing tag
397             if (!$token instanceof HTMLPurifier_Token_End) {
398                 throw new HTMLPurifier_Exception('Unaccounted for tag token in input stream, bug in HTML Purifier');
399             }
401             // make sure that we have something open
402             if (empty($this->stack)) {
403                 if ($escape_invalid_tags) {
404                     if ($e) {
405                         $e->send(E_WARNING, 'Strategy_MakeWellFormed: Unnecessary end tag to text');
406                     }
407                     $token = new HTMLPurifier_Token_Text($generator->generateFromToken($token));
408                 } else {
409                     if ($e) {
410                         $e->send(E_WARNING, 'Strategy_MakeWellFormed: Unnecessary end tag removed');
411                     }
412                     $token = $this->remove();
413                 }
414                 $reprocess = true;
415                 continue;
416             }
418             // first, check for the simplest case: everything closes neatly.
419             // Eventually, everything passes through here; if there are problems
420             // we modify the input stream accordingly and then punt, so that
421             // the tokens get processed again.
422             $current_parent = array_pop($this->stack);
423             if ($current_parent->name == $token->name) {
424                 $token->start = $current_parent;
425                 foreach ($this->injectors as $i => $injector) {
426                     if (isset($token->skip[$i])) {
427                         // See Note [Injector skips]
428                         continue;
429                     }
430                     if ($token->rewind !== null && $token->rewind !== $i) {
431                         continue;
432                     }
433                     $r = $token;
434                     $injector->handleEnd($r);
435                     $token = $this->processToken($r, $i);
436                     $this->stack[] = $current_parent;
437                     $reprocess = true;
438                     break;
439                 }
440                 continue;
441             }
443             // okay, so we're trying to close the wrong tag
445             // undo the pop previous pop
446             $this->stack[] = $current_parent;
448             // scroll back the entire nest, trying to find our tag.
449             // (feature could be to specify how far you'd like to go)
450             $size = count($this->stack);
451             // -2 because -1 is the last element, but we already checked that
452             $skipped_tags = false;
453             for ($j = $size - 2; $j >= 0; $j--) {
454                 if ($this->stack[$j]->name == $token->name) {
455                     $skipped_tags = array_slice($this->stack, $j);
456                     break;
457                 }
458             }
460             // we didn't find the tag, so remove
461             if ($skipped_tags === false) {
462                 if ($escape_invalid_tags) {
463                     if ($e) {
464                         $e->send(E_WARNING, 'Strategy_MakeWellFormed: Stray end tag to text');
465                     }
466                     $token = new HTMLPurifier_Token_Text($generator->generateFromToken($token));
467                 } else {
468                     if ($e) {
469                         $e->send(E_WARNING, 'Strategy_MakeWellFormed: Stray end tag removed');
470                     }
471                     $token = $this->remove();
472                 }
473                 $reprocess = true;
474                 continue;
475             }
477             // do errors, in REVERSE $j order: a,b,c with </a></b></c>
478             $c = count($skipped_tags);
479             if ($e) {
480                 for ($j = $c - 1; $j > 0; $j--) {
481                     // notice we exclude $j == 0, i.e. the current ending tag, from
482                     // the errors... [TagClosedSuppress]
483                     if (!isset($skipped_tags[$j]->armor['MakeWellFormed_TagClosedError'])) {
484                         $e->send(E_NOTICE, 'Strategy_MakeWellFormed: Tag closed by element end', $skipped_tags[$j]);
485                     }
486                 }
487             }
489             // insert tags, in FORWARD $j order: c,b,a with </a></b></c>
490             $replace = array($token);
491             for ($j = 1; $j < $c; $j++) {
492                 // ...as well as from the insertions
493                 $new_token = new HTMLPurifier_Token_End($skipped_tags[$j]->name);
494                 $new_token->start = $skipped_tags[$j];
495                 array_unshift($replace, $new_token);
496                 if (isset($definition->info[$new_token->name]) && $definition->info[$new_token->name]->formatting) {
497                     // [TagClosedAuto]
498                     $element = clone $skipped_tags[$j];
499                     $element->carryover = true;
500                     $element->armor['MakeWellFormed_TagClosedError'] = true;
501                     $replace[] = $element;
502                 }
503             }
504             $token = $this->processToken($replace);
505             $reprocess = true;
506             continue;
507         }
509         $context->destroy('CurrentToken');
510         $context->destroy('CurrentNesting');
511         $context->destroy('InputZipper');
513         unset($this->injectors, $this->stack, $this->tokens);
514         return $zipper->toArray($token);
515     }
517     /**
518      * Processes arbitrary token values for complicated substitution patterns.
519      * In general:
520      *
521      * If $token is an array, it is a list of tokens to substitute for the
522      * current token. These tokens then get individually processed. If there
523      * is a leading integer in the list, that integer determines how many
524      * tokens from the stream should be removed.
525      *
526      * If $token is a regular token, it is swapped with the current token.
527      *
528      * If $token is false, the current token is deleted.
529      *
530      * If $token is an integer, that number of tokens (with the first token
531      * being the current one) will be deleted.
532      *
533      * @param HTMLPurifier_Token|array|int|bool $token Token substitution value
534      * @param HTMLPurifier_Injector|int $injector Injector that performed the substitution; default is if
535      *        this is not an injector related operation.
536      * @throws HTMLPurifier_Exception
537      */
538     protected function processToken($token, $injector = -1)
539     {
540         // Zend OpCache miscompiles $token = array($token), so
541         // avoid this pattern.  See: https://github.com/ezyang/htmlpurifier/issues/108
543         // normalize forms of token
544         if (is_object($token)) {
545             $tmp = $token;
546             $token = array(1, $tmp);
547         }
548         if (is_int($token)) {
549             $tmp = $token;
550             $token = array($tmp);
551         }
552         if ($token === false) {
553             $token = array(1);
554         }
555         if (!is_array($token)) {
556             throw new HTMLPurifier_Exception('Invalid token type from injector');
557         }
558         if (!is_int($token[0])) {
559             array_unshift($token, 1);
560         }
561         if ($token[0] === 0) {
562             throw new HTMLPurifier_Exception('Deleting zero tokens is not valid');
563         }
565         // $token is now an array with the following form:
566         // array(number nodes to delete, new node 1, new node 2, ...)
568         $delete = array_shift($token);
569         list($old, $r) = $this->zipper->splice($this->token, $delete, $token);
571         if ($injector > -1) {
572             // See Note [Injector skips]
573             // Determine appropriate skips.  Here's what the code does:
574             //  *If* we deleted one or more tokens, copy the skips
575             //  of those tokens into the skips of the new tokens (in $token).
576             //  Also, mark the newly inserted tokens as having come from
577             //  $injector.
578             $oldskip = isset($old[0]) ? $old[0]->skip : array();
579             foreach ($token as $object) {
580                 $object->skip = $oldskip;
581                 $object->skip[$injector] = true;
582             }
583         }
585         return $r;
587     }
589     /**
590      * Inserts a token before the current token. Cursor now points to
591      * this token.  You must reprocess after this.
592      * @param HTMLPurifier_Token $token
593      */
594     private function insertBefore($token)
595     {
596         // NB not $this->zipper->insertBefore(), due to positioning
597         // differences
598         $splice = $this->zipper->splice($this->token, 0, array($token));
600         return $splice[1];
601     }
603     /**
604      * Removes current token. Cursor now points to new token occupying previously
605      * occupied space.  You must reprocess after this.
606      */
607     private function remove()
608     {
609         return $this->zipper->delete();
610     }
613 // Note [Injector skips]
614 // ~~~~~~~~~~~~~~~~~~~~~
615 // When I originally designed this class, the idea behind the 'skip'
616 // property of HTMLPurifier_Token was to help avoid infinite loops
617 // in injector processing.  For example, suppose you wrote an injector
618 // that bolded swear words.  Naively, you might write it so that
619 // whenever you saw ****, you replaced it with <strong>****</strong>.
620 //
621 // When this happens, we will reprocess all of the tokens with the
622 // other injectors.  Now there is an opportunity for infinite loop:
623 // if we rerun the swear-word injector on these tokens, we might
624 // see **** and then reprocess again to get
625 // <strong><strong>****</strong></strong> ad infinitum.
626 //
627 // Thus, the idea of a skip is that once we process a token with
628 // an injector, we mark all of those tokens as having "come from"
629 // the injector, and we never run the injector again on these
630 // tokens.
631 //
632 // There were two more complications, however:
633 //
634 //  - With HTMLPurifier_Injector_RemoveEmpty, we noticed that if
635 //    you had <b><i></i></b>, after you removed the <i></i>, you
636 //    really would like this injector to go back and reprocess
637 //    the <b> tag, discovering that it is now empty and can be
638 //    removed.  So we reintroduced the possibility of infinite looping
639 //    by adding a "rewind" function, which let you go back to an
640 //    earlier point in the token stream and reprocess it with injectors.
641 //    Needless to say, we need to UN-skip the token so it gets
642 //    reprocessed.
643 //
644 //  - Suppose that you successfuly process a token, replace it with
645 //    one with your skip mark, but now another injector wants to
646 //    process the skipped token with another token.  Should you continue
647 //    to skip that new token, or reprocess it?  If you reprocess,
648 //    you can end up with an infinite loop where one injector converts
649 //    <a> to <b>, and then another injector converts it back.  So
650 //    we inherit the skips, but for some reason, I thought that we
651 //    should inherit the skip from the first token of the token
652 //    that we deleted.  Why?  Well, it seems to work OK.
653 //
654 // If I were to redesign this functionality, I would absolutely not
655 // go about doing it this way: the semantics are just not very well
656 // defined, and in any case you probably wanted to operate on trees,
657 // not token streams.
659 // vim: et sw=4 sts=4