From a0052f0176bd079e6a94baec59fea2ec5a8d651e Mon Sep 17 00:00:00 2001 From: friendica Date: Thu, 1 Jan 2015 22:18:27 -0800 Subject: htmlpurifier update - compatibility issue with language library autoloader --- library/HTMLPurifier/Strategy/Composite.php | 13 +- library/HTMLPurifier/Strategy/Core.php | 5 +- library/HTMLPurifier/Strategy/FixNesting.php | 349 ++++++------------- library/HTMLPurifier/Strategy/MakeWellFormed.php | 379 ++++++++++++++------- .../Strategy/RemoveForeignElements.php | 94 +++-- .../HTMLPurifier/Strategy/ValidateAttributes.php | 22 +- 6 files changed, 443 insertions(+), 419 deletions(-) (limited to 'library/HTMLPurifier/Strategy') diff --git a/library/HTMLPurifier/Strategy/Composite.php b/library/HTMLPurifier/Strategy/Composite.php index 816490b79..d7d35ce7d 100644 --- a/library/HTMLPurifier/Strategy/Composite.php +++ b/library/HTMLPurifier/Strategy/Composite.php @@ -8,18 +8,23 @@ abstract class HTMLPurifier_Strategy_Composite extends HTMLPurifier_Strategy /** * List of strategies to run tokens through. + * @type HTMLPurifier_Strategy[] */ protected $strategies = array(); - abstract public function __construct(); - - public function execute($tokens, $config, $context) { + /** + * @param HTMLPurifier_Token[] $tokens + * @param HTMLPurifier_Config $config + * @param HTMLPurifier_Context $context + * @return HTMLPurifier_Token[] + */ + public function execute($tokens, $config, $context) + { foreach ($this->strategies as $strategy) { $tokens = $strategy->execute($tokens, $config, $context); } return $tokens; } - } // vim: et sw=4 sts=4 diff --git a/library/HTMLPurifier/Strategy/Core.php b/library/HTMLPurifier/Strategy/Core.php index d90e15860..4414c17d6 100644 --- a/library/HTMLPurifier/Strategy/Core.php +++ b/library/HTMLPurifier/Strategy/Core.php @@ -5,14 +5,13 @@ */ class HTMLPurifier_Strategy_Core extends HTMLPurifier_Strategy_Composite { - - public function __construct() { + public function __construct() + { $this->strategies[] = new HTMLPurifier_Strategy_RemoveForeignElements(); $this->strategies[] = new HTMLPurifier_Strategy_MakeWellFormed(); $this->strategies[] = new HTMLPurifier_Strategy_FixNesting(); $this->strategies[] = new HTMLPurifier_Strategy_ValidateAttributes(); } - } // vim: et sw=4 sts=4 diff --git a/library/HTMLPurifier/Strategy/FixNesting.php b/library/HTMLPurifier/Strategy/FixNesting.php index f81802391..6fa673db9 100644 --- a/library/HTMLPurifier/Strategy/FixNesting.php +++ b/library/HTMLPurifier/Strategy/FixNesting.php @@ -10,12 +10,12 @@ * document type definitions, such as the chameleon nature of ins/del * tags and global child exclusions. * - * The first major objective of this strategy is to iterate through all the - * nodes (not tokens) of the list of tokens and determine whether or not - * their children conform to the element's definition. If they do not, the - * child definition may optionally supply an amended list of elements that - * is valid or require that the entire node be deleted (and the previous - * node rescanned). + * The first major objective of this strategy is to iterate through all + * the nodes and determine whether or not their children conform to the + * element's definition. If they do not, the child definition may + * optionally supply an amended list of elements that is valid or + * require that the entire node be deleted (and the previous node + * rescanned). * * The second objective is to ensure that explicitly excluded elements of * an element do not appear in its children. Code that accomplishes this @@ -25,24 +25,33 @@ * @note Whether or not unrecognized children are silently dropped or * translated into text depends on the child definitions. * - * @todo Enable nodes to be bubbled out of the structure. + * @todo Enable nodes to be bubbled out of the structure. This is + * easier with our new algorithm. */ class HTMLPurifier_Strategy_FixNesting extends HTMLPurifier_Strategy { - public function execute($tokens, $config, $context) { + /** + * @param HTMLPurifier_Token[] $tokens + * @param HTMLPurifier_Config $config + * @param HTMLPurifier_Context $context + * @return array|HTMLPurifier_Token[] + */ + public function execute($tokens, $config, $context) + { + //####################################################################// // Pre-processing + // O(n) pass to convert to a tree, so that we can efficiently + // refer to substrings + $top_node = HTMLPurifier_Arborize::arborize($tokens, $config, $context); + // get a copy of the HTML definition $definition = $config->getHTMLDefinition(); - // insert implicit "parent" node, will be removed at end. - // DEFINITION CALL - $parent_name = $definition->info_parent; - array_unshift($tokens, new HTMLPurifier_Token_Start($parent_name)); - $tokens[] = new HTMLPurifier_Token_End($parent_name); + $excludes_enabled = !$config->get('Core.DisableExcludes'); // setup the context variable 'IsInline', for chameleon processing // is 'false' when we are not inline, 'true' when it must always @@ -57,272 +66,116 @@ class HTMLPurifier_Strategy_FixNesting extends HTMLPurifier_Strategy //####################################################################// // Loop initialization - // stack that contains the indexes of all parents, - // $stack[count($stack)-1] being the current parent - $stack = array(); - // stack that contains all elements that are excluded // it is organized by parent elements, similar to $stack, // but it is only populated when an element with exclusions is // processed, i.e. there won't be empty exclusions. - $exclude_stack = array(); + $exclude_stack = array($definition->info_parent_def->excludes); // variable that contains the start token while we are processing // nodes. This enables error reporting to do its job - $start_token = false; - $context->register('CurrentToken', $start_token); + $node = $top_node; + // dummy token + list($token, $d) = $node->toTokenPair(); + $context->register('CurrentNode', $node); + $context->register('CurrentToken', $token); //####################################################################// // Loop - // iterate through all start nodes. Determining the start node - // is complicated so it has been omitted from the loop construct - for ($i = 0, $size = count($tokens) ; $i < $size; ) { - - //################################################################// - // Gather information on children - - // child token accumulator - $child_tokens = array(); - - // scroll to the end of this node, report number, and collect - // all children - for ($j = $i, $depth = 0; ; $j++) { - if ($tokens[$j] instanceof HTMLPurifier_Token_Start) { - $depth++; - // skip token assignment on first iteration, this is the - // token we currently are on - if ($depth == 1) continue; - } elseif ($tokens[$j] instanceof HTMLPurifier_Token_End) { - $depth--; - // skip token assignment on last iteration, this is the - // end token of the token we're currently on - if ($depth == 0) break; - } - $child_tokens[] = $tokens[$j]; - } - - // $i is index of start token - // $j is index of end token - - $start_token = $tokens[$i]; // to make token available via CurrentToken - - //################################################################// - // Gather information on parent - - // calculate parent information - if ($count = count($stack)) { - $parent_index = $stack[$count-1]; - $parent_name = $tokens[$parent_index]->name; - if ($parent_index == 0) { - $parent_def = $definition->info_parent_def; - } else { - $parent_def = $definition->info[$parent_name]; - } - } else { - // processing as if the parent were the "root" node - // unknown info, it won't be used anyway, in the future, - // we may want to enforce one element only (this is - // necessary for HTML Purifier to clean entire documents - $parent_index = $parent_name = $parent_def = null; - } - - // calculate context - if ($is_inline === false) { - // check if conditions make it inline - if (!empty($parent_def) && $parent_def->descendants_are_inline) { - $is_inline = $count - 1; - } - } else { - // check if we're out of inline - if ($count === $is_inline) { - $is_inline = false; - } - } - - //################################################################// - // Determine whether element is explicitly excluded SGML-style - - // determine whether or not element is excluded by checking all - // parent exclusions. The array should not be very large, two - // elements at most. - $excluded = false; - if (!empty($exclude_stack)) { - foreach ($exclude_stack as $lookup) { - if (isset($lookup[$tokens[$i]->name])) { - $excluded = true; - // no need to continue processing - break; - } + // We need to implement a post-order traversal iteratively, to + // avoid running into stack space limits. This is pretty tricky + // to reason about, so we just manually stack-ify the recursive + // variant: + // + // function f($node) { + // foreach ($node->children as $child) { + // f($child); + // } + // validate($node); + // } + // + // Thus, we will represent a stack frame as array($node, + // $is_inline, stack of children) + // e.g. array_reverse($node->children) - already processed + // children. + + $parent_def = $definition->info_parent_def; + $stack = array( + array($top_node, + $parent_def->descendants_are_inline, + $parent_def->excludes, // exclusions + 0) + ); + + while (!empty($stack)) { + list($node, $is_inline, $excludes, $ix) = array_pop($stack); + // recursive call + $go = false; + $def = empty($stack) ? $definition->info_parent_def : $definition->info[$node->name]; + while (isset($node->children[$ix])) { + $child = $node->children[$ix++]; + if ($child instanceof HTMLPurifier_Node_Element) { + $go = true; + $stack[] = array($node, $is_inline, $excludes, $ix); + $stack[] = array($child, + // ToDo: I don't think it matters if it's def or + // child_def, but double check this... + $is_inline || $def->descendants_are_inline, + empty($def->excludes) ? $excludes + : array_merge($excludes, $def->excludes), + 0); + break; } - } - - //################################################################// - // Perform child validation - - if ($excluded) { - // there is an exclusion, remove the entire node - $result = false; - $excludes = array(); // not used, but good to initialize anyway + }; + if ($go) continue; + list($token, $d) = $node->toTokenPair(); + // base case + if ($excludes_enabled && isset($excludes[$node->name])) { + $node->dead = true; + if ($e) $e->send(E_ERROR, 'Strategy_FixNesting: Node excluded'); } else { - // DEFINITION CALL - if ($i === 0) { - // special processing for the first node - $def = $definition->info_parent_def; - } else { - $def = $definition->info[$tokens[$i]->name]; - + // XXX I suppose it would be slightly more efficient to + // avoid the allocation here and have children + // strategies handle it + $children = array(); + foreach ($node->children as $child) { + if (!$child->dead) $children[] = $child; } - - if (!empty($def->child)) { - // have DTD child def validate children - $result = $def->child->validateChildren( - $child_tokens, $config, $context); + $result = $def->child->validateChildren($children, $config, $context); + if ($result === true) { + // nop + $node->children = $children; + } elseif ($result === false) { + $node->dead = true; + if ($e) $e->send(E_ERROR, 'Strategy_FixNesting: Node removed'); } else { - // weird, no child definition, get rid of everything - $result = false; - } - - // determine whether or not this element has any exclusions - $excludes = $def->excludes; - } - - // $result is now a bool or array - - //################################################################// - // Process result by interpreting $result - - if ($result === true || $child_tokens === $result) { - // leave the node as is - - // register start token as a parental node start - $stack[] = $i; - - // register exclusions if there are any - if (!empty($excludes)) $exclude_stack[] = $excludes; - - // move cursor to next possible start node - $i++; - - } elseif($result === false) { - // remove entire node - - if ($e) { - if ($excluded) { - $e->send(E_ERROR, 'Strategy_FixNesting: Node excluded'); - } else { - $e->send(E_ERROR, 'Strategy_FixNesting: Node removed'); - } - } - - // calculate length of inner tokens and current tokens - $length = $j - $i + 1; - - // perform removal - array_splice($tokens, $i, $length); - - // update size - $size -= $length; - - // there is no start token to register, - // current node is now the next possible start node - // unless it turns out that we need to do a double-check - - // this is a rought heuristic that covers 100% of HTML's - // cases and 99% of all other cases. A child definition - // that would be tricked by this would be something like: - // ( | a b c) where it's all or nothing. Fortunately, - // our current implementation claims that that case would - // not allow empty, even if it did - if (!$parent_def->child->allow_empty) { - // we need to do a double-check - $i = $parent_index; - array_pop($stack); - } - - // PROJECTED OPTIMIZATION: Process all children elements before - // reprocessing parent node. - - } else { - // replace node with $result - - // calculate length of inner tokens - $length = $j - $i - 1; - - if ($e) { - if (empty($result) && $length) { - $e->send(E_ERROR, 'Strategy_FixNesting: Node contents removed'); - } else { - $e->send(E_WARNING, 'Strategy_FixNesting: Node reorganized'); + $node->children = $result; + if ($e) { + // XXX This will miss mutations of internal nodes. Perhaps defer to the child validators + if (empty($result) && !empty($children)) { + $e->send(E_ERROR, 'Strategy_FixNesting: Node contents removed'); + } else if ($result != $children) { + $e->send(E_WARNING, 'Strategy_FixNesting: Node reorganized'); + } } } - - // perform replacement - array_splice($tokens, $i + 1, $length, $result); - - // update size - $size -= $length; - $size += count($result); - - // register start token as a parental node start - $stack[] = $i; - - // register exclusions if there are any - if (!empty($excludes)) $exclude_stack[] = $excludes; - - // move cursor to next possible start node - $i++; - } - - //################################################################// - // Scroll to next start node - - // We assume, at this point, that $i is the index of the token - // that is the first possible new start point for a node. - - // Test if the token indeed is a start tag, if not, move forward - // and test again. - $size = count($tokens); - while ($i < $size and !$tokens[$i] instanceof HTMLPurifier_Token_Start) { - if ($tokens[$i] instanceof HTMLPurifier_Token_End) { - // pop a token index off the stack if we ended a node - array_pop($stack); - // pop an exclusion lookup off exclusion stack if - // we ended node and that node had exclusions - if ($i == 0 || $i == $size - 1) { - // use specialized var if it's the super-parent - $s_excludes = $definition->info_parent_def->excludes; - } else { - $s_excludes = $definition->info[$tokens[$i]->name]->excludes; - } - if ($s_excludes) { - array_pop($exclude_stack); - } - } - $i++; - } - } //####################################################################// // Post-processing - // remove implicit parent tokens at the beginning and end - array_shift($tokens); - array_pop($tokens); - // remove context variables $context->destroy('IsInline'); + $context->destroy('CurrentNode'); $context->destroy('CurrentToken'); //####################################################################// // Return - return $tokens; - + return HTMLPurifier_Arborize::flatten($node, $config, $context); } - } // vim: et sw=4 sts=4 diff --git a/library/HTMLPurifier/Strategy/MakeWellFormed.php b/library/HTMLPurifier/Strategy/MakeWellFormed.php index c73658400..e389e0011 100644 --- a/library/HTMLPurifier/Strategy/MakeWellFormed.php +++ b/library/HTMLPurifier/Strategy/MakeWellFormed.php @@ -2,66 +2,97 @@ /** * Takes tokens makes them well-formed (balance end tags, etc.) + * + * Specification of the armor attributes this strategy uses: + * + * - MakeWellFormed_TagClosedError: This armor field is used to + * suppress tag closed errors for certain tokens [TagClosedSuppress], + * in particular, if a tag was generated automatically by HTML + * Purifier, we may rely on our infrastructure to close it for us + * and shouldn't report an error to the user [TagClosedAuto]. */ class HTMLPurifier_Strategy_MakeWellFormed extends HTMLPurifier_Strategy { /** * Array stream of tokens being processed. + * @type HTMLPurifier_Token[] */ protected $tokens; /** - * Current index in $tokens. + * Current token. + * @type HTMLPurifier_Token */ - protected $t; + protected $token; + + /** + * Zipper managing the true state. + * @type HTMLPurifier_Zipper + */ + protected $zipper; /** * Current nesting of elements. + * @type array */ protected $stack; /** * Injectors active in this stream processing. + * @type HTMLPurifier_Injector[] */ protected $injectors; /** * Current instance of HTMLPurifier_Config. + * @type HTMLPurifier_Config */ protected $config; /** * Current instance of HTMLPurifier_Context. + * @type HTMLPurifier_Context */ protected $context; - public function execute($tokens, $config, $context) { - + /** + * @param HTMLPurifier_Token[] $tokens + * @param HTMLPurifier_Config $config + * @param HTMLPurifier_Context $context + * @return HTMLPurifier_Token[] + * @throws HTMLPurifier_Exception + */ + public function execute($tokens, $config, $context) + { $definition = $config->getHTMLDefinition(); // local variables $generator = new HTMLPurifier_Generator($config, $context); $escape_invalid_tags = $config->get('Core.EscapeInvalidTags'); + // used for autoclose early abortion + $global_parent_allowed_elements = $definition->info_parent_def->child->getAllowedElements($config); $e = $context->get('ErrorCollector', true); - $t = false; // token index $i = false; // injector index - $token = false; // the current token - $reprocess = false; // whether or not to reprocess the same token + list($zipper, $token) = HTMLPurifier_Zipper::fromArray($tokens); + if ($token === NULL) { + return array(); + } + $reprocess = false; // whether or not to reprocess the same token $stack = array(); // member variables - $this->stack =& $stack; - $this->t =& $t; - $this->tokens =& $tokens; - $this->config = $config; + $this->stack =& $stack; + $this->tokens =& $tokens; + $this->token =& $token; + $this->zipper =& $zipper; + $this->config = $config; $this->context = $context; // context variables $context->register('CurrentNesting', $stack); - $context->register('InputIndex', $t); - $context->register('InputTokens', $tokens); - $context->register('CurrentToken', $token); + $context->register('InputZipper', $zipper); + $context->register('CurrentToken', $token); // -- begin INJECTOR -- @@ -73,9 +104,13 @@ class HTMLPurifier_Strategy_MakeWellFormed extends HTMLPurifier_Strategy unset($injectors['Custom']); // special case foreach ($injectors as $injector => $b) { // XXX: Fix with a legitimate lookup table of enabled filters - if (strpos($injector, '.') !== false) continue; + if (strpos($injector, '.') !== false) { + continue; + } $injector = "HTMLPurifier_Injector_$injector"; - if (!$b) continue; + if (!$b) { + continue; + } $this->injectors[] = new $injector; } foreach ($def_injectors as $injector) { @@ -83,7 +118,9 @@ class HTMLPurifier_Strategy_MakeWellFormed extends HTMLPurifier_Strategy $this->injectors[] = $injector; } foreach ($custom_injectors as $injector) { - if (!$injector) continue; + if (!$injector) { + continue; + } if (is_string($injector)) { $injector = "HTMLPurifier_Injector_$injector"; $injector = new $injector; @@ -95,14 +132,16 @@ class HTMLPurifier_Strategy_MakeWellFormed extends HTMLPurifier_Strategy // variables for performance reasons foreach ($this->injectors as $ix => $injector) { $error = $injector->prepare($config, $context); - if (!$error) continue; + if (!$error) { + continue; + } array_splice($this->injectors, $ix, 1); // rm the injector trigger_error("Cannot enable {$injector->name} injector because $error is not allowed", E_USER_WARNING); } // -- end INJECTOR -- - // a note on punting: + // a note on reprocessing: // In order to reduce code duplication, whenever some code needs // to make HTML changes in order to make things "correct", the // new HTML gets sent through the purifier, regardless of its @@ -111,70 +150,75 @@ class HTMLPurifier_Strategy_MakeWellFormed extends HTMLPurifier_Strategy // punt ($reprocess = true; continue;) and it does that for us. // isset is in loop because $tokens size changes during loop exec - for ( - $t = 0; - $t == 0 || isset($tokens[$t - 1]); - // only increment if we don't need to reprocess - $reprocess ? $reprocess = false : $t++ - ) { + for (;; + // only increment if we don't need to reprocess + $reprocess ? $reprocess = false : $token = $zipper->next($token)) { // check for a rewind - if (is_int($i) && $i >= 0) { + if (is_int($i)) { // possibility: disable rewinding if the current token has a // rewind set on it already. This would offer protection from // infinite loop, but might hinder some advanced rewinding. - $rewind_to = $this->injectors[$i]->getRewind(); - if (is_int($rewind_to) && $rewind_to < $t) { - if ($rewind_to < 0) $rewind_to = 0; - while ($t > $rewind_to) { - $t--; - $prev = $tokens[$t]; + $rewind_offset = $this->injectors[$i]->getRewindOffset(); + if (is_int($rewind_offset)) { + for ($j = 0; $j < $rewind_offset; $j++) { + if (empty($zipper->front)) break; + $token = $zipper->prev($token); // indicate that other injectors should not process this token, // but we need to reprocess it - unset($prev->skip[$i]); - $prev->rewind = $i; - if ($prev instanceof HTMLPurifier_Token_Start) array_pop($this->stack); - elseif ($prev instanceof HTMLPurifier_Token_End) $this->stack[] = $prev->start; + unset($token->skip[$i]); + $token->rewind = $i; + if ($token instanceof HTMLPurifier_Token_Start) { + array_pop($this->stack); + } elseif ($token instanceof HTMLPurifier_Token_End) { + $this->stack[] = $token->start; + } } } $i = false; } // handle case of document end - if (!isset($tokens[$t])) { + if ($token === NULL) { // kill processing if stack is empty - if (empty($this->stack)) break; + if (empty($this->stack)) { + break; + } // peek $top_nesting = array_pop($this->stack); $this->stack[] = $top_nesting; - // send error + // send error [TagClosedSuppress] if ($e && !isset($top_nesting->armor['MakeWellFormed_TagClosedError'])) { $e->send(E_NOTICE, 'Strategy_MakeWellFormed: Tag closed by document end', $top_nesting); } // append, don't splice, since this is the end - $tokens[] = new HTMLPurifier_Token_End($top_nesting->name); + $token = new HTMLPurifier_Token_End($top_nesting->name); // punt! $reprocess = true; continue; } - $token = $tokens[$t]; - - //echo '
'; printTokens($tokens, $t); printTokens($this->stack); + //echo '
'; printZipper($zipper, $token);//printTokens($this->stack); //flush(); // quick-check: if it's not a tag, no need to process if (empty($token->is_tag)) { if ($token instanceof HTMLPurifier_Token_Text) { foreach ($this->injectors as $i => $injector) { - if (isset($token->skip[$i])) continue; - if ($token->rewind !== null && $token->rewind !== $i) continue; - $injector->handleText($token); - $this->processToken($token, $i); + if (isset($token->skip[$i])) { + continue; + } + if ($token->rewind !== null && $token->rewind !== $i) { + continue; + } + // XXX fuckup + $r = $token; + $injector->handleText($r); + $token = $this->processToken($r, $i); $reprocess = true; break; } @@ -193,12 +237,22 @@ class HTMLPurifier_Strategy_MakeWellFormed extends HTMLPurifier_Strategy $ok = false; if ($type === 'empty' && $token instanceof HTMLPurifier_Token_Start) { // claims to be a start tag but is empty - $token = new HTMLPurifier_Token_Empty($token->name, $token->attr); + $token = new HTMLPurifier_Token_Empty( + $token->name, + $token->attr, + $token->line, + $token->col, + $token->armor + ); $ok = true; } elseif ($type && $type !== 'empty' && $token instanceof HTMLPurifier_Token_Empty) { // claims to be empty but really is a start tag - $this->swap(new HTMLPurifier_Token_End($token->name)); - $this->insertBefore(new HTMLPurifier_Token_Start($token->name, $token->attr)); + // NB: this assignment is required + $old_token = $token; + $token = new HTMLPurifier_Token_End($token->name); + $token = $this->insertBefore( + new HTMLPurifier_Token_Start($old_token->name, $old_token->attr, $old_token->line, $old_token->col, $old_token->armor) + ); // punt (since we had to modify the input stream in a non-trivial way) $reprocess = true; continue; @@ -211,55 +265,96 @@ class HTMLPurifier_Strategy_MakeWellFormed extends HTMLPurifier_Strategy // ...unless they also have to close their parent if (!empty($this->stack)) { + // Performance note: you might think that it's rather + // inefficient, recalculating the autoclose information + // for every tag that a token closes (since when we + // do an autoclose, we push a new token into the + // stream and then /process/ that, before + // re-processing this token.) But this is + // necessary, because an injector can make an + // arbitrary transformations to the autoclosing + // tokens we introduce, so things may have changed + // in the meantime. Also, doing the inefficient thing is + // "easy" to reason about (for certain perverse definitions + // of "easy") + $parent = array_pop($this->stack); $this->stack[] = $parent; + $parent_def = null; + $parent_elements = null; + $autoclose = false; if (isset($definition->info[$parent->name])) { - $elements = $definition->info[$parent->name]->child->getAllowedElements($config); - $autoclose = !isset($elements[$token->name]); - } else { - $autoclose = false; + $parent_def = $definition->info[$parent->name]; + $parent_elements = $parent_def->child->getAllowedElements($config); + $autoclose = !isset($parent_elements[$token->name]); } if ($autoclose && $definition->info[$token->name]->wrap) { - // Check if an element can be wrapped by another - // element to make it valid in a context (for + // Check if an element can be wrapped by another + // element to make it valid in a context (for // example,