diff options
Diffstat (limited to 'lib/htmlpurifier/library/HTMLPurifier/Strategy')
6 files changed, 1128 insertions, 0 deletions
diff --git a/lib/htmlpurifier/library/HTMLPurifier/Strategy/Composite.php b/lib/htmlpurifier/library/HTMLPurifier/Strategy/Composite.php new file mode 100644 index 000000000..92aefd33e --- /dev/null +++ b/lib/htmlpurifier/library/HTMLPurifier/Strategy/Composite.php @@ -0,0 +1,23 @@ +<?php + +/** + * Composite strategy that runs multiple strategies on tokens. + */ +abstract class HTMLPurifier_Strategy_Composite extends HTMLPurifier_Strategy +{ + + /** + * List of strategies to run tokens through. + */ + protected $strategies = array(); + + public function execute($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/lib/htmlpurifier/library/HTMLPurifier/Strategy/Core.php b/lib/htmlpurifier/library/HTMLPurifier/Strategy/Core.php new file mode 100644 index 000000000..d90e15860 --- /dev/null +++ b/lib/htmlpurifier/library/HTMLPurifier/Strategy/Core.php @@ -0,0 +1,18 @@ +<?php + +/** + * Core strategy composed of the big four strategies. + */ +class HTMLPurifier_Strategy_Core extends HTMLPurifier_Strategy_Composite +{ + + public function __construct() { + $this->strategies[] = new HTMLPurifier_Strategy_RemoveForeignElements(); + $this->strategies[] = new HTMLPurifier_Strategy_MakeWellFormed(); + $this->strategies[] = new HTMLPurifier_Strategy_FixNesting(); + $this->strategies[] = new HTMLPurifier_Strategy_ValidateAttributes(); + } + +} + +// vim: et sw=4 sts=4 diff --git a/lib/htmlpurifier/library/HTMLPurifier/Strategy/FixNesting.php b/lib/htmlpurifier/library/HTMLPurifier/Strategy/FixNesting.php new file mode 100644 index 000000000..f81802391 --- /dev/null +++ b/lib/htmlpurifier/library/HTMLPurifier/Strategy/FixNesting.php @@ -0,0 +1,328 @@ +<?php + +/** + * Takes a well formed list of tokens and fixes their nesting. + * + * HTML elements dictate which elements are allowed to be their children, + * for example, you can't have a p tag in a span tag. Other elements have + * much more rigorous definitions: tables, for instance, require a specific + * order for their elements. There are also constraints not expressible by + * document type definitions, such as the chameleon nature of ins/del + * tags and global child exclusions. + * + * The first major objective of this strategy is to iterate through all the + * nodes (not tokens) of the list of tokens and determine whether or not + * their children conform to the element's definition. If they do not, the + * child definition may optionally supply an amended list of elements that + * is valid or require that the entire node be deleted (and the previous + * node rescanned). + * + * The second objective is to ensure that explicitly excluded elements of + * an element do not appear in its children. Code that accomplishes this + * task is pervasive through the strategy, though the two are distinct tasks + * and could, theoretically, be seperated (although it's not recommended). + * + * @note Whether or not unrecognized children are silently dropped or + * translated into text depends on the child definitions. + * + * @todo Enable nodes to be bubbled out of the structure. + */ + +class HTMLPurifier_Strategy_FixNesting extends HTMLPurifier_Strategy +{ + + public function execute($tokens, $config, $context) { + //####################################################################// + // Pre-processing + + // 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); + + // setup the context variable 'IsInline', for chameleon processing + // is 'false' when we are not inline, 'true' when it must always + // be inline, and an integer when it is inline for a certain + // branch of the document tree + $is_inline = $definition->info_parent_def->descendants_are_inline; + $context->register('IsInline', $is_inline); + + // setup error collector + $e =& $context->get('ErrorCollector', true); + + //####################################################################// + // 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(); + + // 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', $start_token); + + //####################################################################// + // Loop + + // iterate through all start nodes. Determining the start node + // is complicated so it has been omitted from the loop construct + for ($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])) { + $excluded = true; + // no need to continue processing + break; + } + } + } + + //################################################################// + // Perform child validation + + if ($excluded) { + // there is an exclusion, remove the entire node + $result = false; + $excludes = array(); // not used, but good to initialize anyway + } else { + // DEFINITION CALL + if ($i === 0) { + // special processing for the first node + $def = $definition->info_parent_def; + } else { + $def = $definition->info[$tokens[$i]->name]; + + } + + if (!empty($def->child)) { + // have DTD child def validate children + $result = $def->child->validateChildren( + $child_tokens, $config, $context); + } 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 { + $e->send(E_WARNING, 'Strategy_FixNesting: Node reorganized'); + } + } + + // perform replacement + array_splice($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('CurrentToken'); + + //####################################################################// + // Return + + return $tokens; + + } + +} + +// vim: et sw=4 sts=4 diff --git a/lib/htmlpurifier/library/HTMLPurifier/Strategy/MakeWellFormed.php b/lib/htmlpurifier/library/HTMLPurifier/Strategy/MakeWellFormed.php new file mode 100644 index 000000000..c7aa1bb86 --- /dev/null +++ b/lib/htmlpurifier/library/HTMLPurifier/Strategy/MakeWellFormed.php @@ -0,0 +1,532 @@ +<?php + +/** + * 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. + */ + protected $tokens; + + /** + * Current index in $tokens. + */ + protected $t; + + /** + * Current nesting of elements. + */ + protected $stack; + + /** + * Injectors active in this stream processing. + */ + protected $injectors; + + /** + * Current instance of HTMLPurifier_Config. + */ + protected $config; + + /** + * Current instance of HTMLPurifier_Context. + */ + protected $context; + + public function execute($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 + $global_parent_allowed_elements = array(); + if (isset($definition->info[$definition->info_parent])) { + // may be unset under testing circumstances + $global_parent_allowed_elements = $definition->info[$definition->info_parent]->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 + $stack = array(); + + // member variables + $this->stack =& $stack; + $this->t =& $t; + $this->tokens =& $tokens; + $this->config = $config; + $this->context = $context; + + // context variables + $context->register('CurrentNesting', $stack); + $context->register('InputIndex', $t); + $context->register('InputTokens', $tokens); + $context->register('CurrentToken', $token); + + // -- begin INJECTOR -- + + $this->injectors = array(); + + $injectors = $config->getBatch('AutoFormat'); + $def_injectors = $definition->info_injector; + $custom_injectors = $injectors['Custom']; + unset($injectors['Custom']); // special case + foreach ($injectors as $injector => $b) { + // XXX: Fix with a legitimate lookup table of enabled filters + if (strpos($injector, '.') !== false) continue; + $injector = "HTMLPurifier_Injector_$injector"; + if (!$b) continue; + $this->injectors[] = new $injector; + } + foreach ($def_injectors as $injector) { + // assumed to be objects + $this->injectors[] = $injector; + } + foreach ($custom_injectors as $injector) { + if (!$injector) continue; + if (is_string($injector)) { + $injector = "HTMLPurifier_Injector_$injector"; + $injector = new $injector; + } + $this->injectors[] = $injector; + } + + // give the injectors references to the definition and context + // variables for performance reasons + foreach ($this->injectors as $ix => $injector) { + $error = $injector->prepare($config, $context); + if (!$error) continue; + array_splice($this->injectors, $ix, 1); // rm the injector + trigger_error("Cannot enable {$injector->name} injector because $error is not allowed", E_USER_WARNING); + } + + // -- end INJECTOR -- + + // a note on reprocessing: + // In order to reduce code duplication, whenever some code needs + // to make HTML changes in order to make things "correct", the + // new HTML gets sent through the purifier, regardless of its + // status. This means that if we add a start token, because it + // was totally necessary, we don't have to update nesting; we just + // punt ($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 : $t++ + ) { + + // check for a rewind + if (is_int($i) && $i >= 0) { + // 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]; + // 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; + } + } + $i = false; + } + + // handle case of document end + if (!isset($tokens[$t])) { + // kill processing if stack is empty + if (empty($this->stack)) break; + + // peek + $top_nesting = array_pop($this->stack); + $this->stack[] = $top_nesting; + + // send error [TagClosedSuppress] + if ($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); + + // punt! + $reprocess = true; + continue; + } + + $token = $tokens[$t]; + + //echo '<br>'; printTokens($tokens, $t); printTokens($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); + $reprocess = true; + break; + } + } + // another possibility is a comment + continue; + } + + if (isset($definition->info[$token->name])) { + $type = $definition->info[$token->name]->child->type; + } else { + $type = false; // Type is unknown, treat accordingly + } + + // quick tag checks: anything that's *not* an end tag + $ok = false; + if ($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->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, $token->line, $token->col, $token->armor)); + // punt (since we had to modify the input stream in a non-trivial way) + $reprocess = true; + continue; + } elseif ($token instanceof HTMLPurifier_Token_Empty) { + // real empty token + $ok = true; + } elseif ($token instanceof HTMLPurifier_Token_Start) { + // start tag + + // ...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[] = $parent; + + if (isset($definition->info[$parent->name])) { + $elements = $definition->info[$parent->name]->child->getAllowedElements($config); + $autoclose = !isset($elements[$token->name]); + } else { + $autoclose = false; + } + + if ($autoclose && $definition->info[$token->name]->wrap) { + // Check if an element can be wrapped by another + // element to make it valid in a context (for + // example, <ul><ul> needs a <li> in between) + $wrapname = $definition->info[$token->name]->wrap; + $wrapdef = $definition->info[$wrapname]; + $elements = $wrapdef->child->getAllowedElements($config); + $parent_elements = $definition->info[$parent->name]->child->getAllowedElements($config); + if (isset($elements[$token->name]) && isset($parent_elements[$wrapname])) { + $newtoken = new HTMLPurifier_Token_Start($wrapname); + $this->insertBefore($newtoken); + $reprocess = true; + continue; + } + } + + $carryover = false; + if ($autoclose && $definition->info[$parent->name]->formatting) { + $carryover = true; + } + + if ($autoclose) { + // check if this autoclose is doomed to fail + // (this rechecks $parent, which his harmless) + $autoclose_ok = isset($global_parent_allowed_elements[$token->name]); + if (!$autoclose_ok) { + foreach ($this->stack as $ancestor) { + $elements = $definition->info[$ancestor->name]->child->getAllowedElements($config); + if (isset($elements[$token->name])) { + $autoclose_ok = true; + break; + } + if ($definition->info[$token->name]->wrap) { + $wrapname = $definition->info[$token->name]->wrap; + $wrapdef = $definition->info[$wrapname]; + $wrap_elements = $wrapdef->child->getAllowedElements($config); + if (isset($wrap_elements[$token->name]) && isset($elements[$wrapname])) { + $autoclose_ok = true; + break; + } + } + } + } + if ($autoclose_ok) { + // errors need to be updated + $new_token = new HTMLPurifier_Token_End($parent->name); + $new_token->start = $parent; + if ($carryover) { + $element = clone $parent; + // [TagClosedAuto] + $element->armor['MakeWellFormed_TagClosedError'] = true; + $element->carryover = true; + $this->processToken(array($new_token, $token, $element)); + } else { + $this->insertBefore($new_token); + } + // [TagClosedSuppress] + if ($e && !isset($parent->armor['MakeWellFormed_TagClosedError'])) { + if (!$carryover) { + $e->send(E_NOTICE, 'Strategy_MakeWellFormed: Tag auto closed', $parent); + } else { + $e->send(E_NOTICE, 'Strategy_MakeWellFormed: Tag carryover', $parent); + } + } + } else { + $this->remove(); + } + $reprocess = true; + continue; + } + + } + $ok = true; + } + + if ($ok) { + foreach ($this->injectors as $i => $injector) { + if (isset($token->skip[$i])) continue; + if ($token->rewind !== null && $token->rewind !== $i) continue; + $injector->handleElement($token); + $this->processToken($token, $i); + $reprocess = true; + break; + } + if (!$reprocess) { + // ah, nothing interesting happened; do normal processing + $this->swap($token); + if ($token instanceof HTMLPurifier_Token_Start) { + $this->stack[] = $token; + } elseif ($token instanceof HTMLPurifier_Token_End) { + throw new HTMLPurifier_Exception('Improper handling of end tag in start code; possible error in MakeWellFormed'); + } + } + continue; + } + + // sanity check: we should be dealing with a closing tag + if (!$token instanceof HTMLPurifier_Token_End) { + throw new HTMLPurifier_Exception('Unaccounted for tag token in input stream, bug in HTML Purifier'); + } + + // make sure that we have something open + if (empty($this->stack)) { + if ($escape_invalid_tags) { + if ($e) $e->send(E_WARNING, 'Strategy_MakeWellFormed: Unnecessary end tag to text'); + $this->swap(new HTMLPurifier_Token_Text( + $generator->generateFromToken($token) + )); + } else { + $this->remove(); + if ($e) $e->send(E_WARNING, 'Strategy_MakeWellFormed: Unnecessary end tag removed'); + } + $reprocess = true; + continue; + } + + // first, check for the simplest case: everything closes neatly. + // Eventually, everything passes through here; if there are problems + // we modify the input stream accordingly and then punt, so that + // the tokens get processed again. + $current_parent = array_pop($this->stack); + if ($current_parent->name == $token->name) { + $token->start = $current_parent; + foreach ($this->injectors as $i => $injector) { + if (isset($token->skip[$i])) continue; + if ($token->rewind !== null && $token->rewind !== $i) continue; + $injector->handleEnd($token); + $this->processToken($token, $i); + $this->stack[] = $current_parent; + $reprocess = true; + break; + } + continue; + } + + // okay, so we're trying to close the wrong tag + + // undo the pop previous pop + $this->stack[] = $current_parent; + + // scroll back the entire nest, trying to find our tag. + // (feature could be to specify how far you'd like to go) + $size = count($this->stack); + // -2 because -1 is the last element, but we already checked that + $skipped_tags = false; + for ($j = $size - 2; $j >= 0; $j--) { + if ($this->stack[$j]->name == $token->name) { + $skipped_tags = array_slice($this->stack, $j); + break; + } + } + + // we didn't find the tag, so remove + if ($skipped_tags === false) { + if ($escape_invalid_tags) { + $this->swap(new HTMLPurifier_Token_Text( + $generator->generateFromToken($token) + )); + if ($e) $e->send(E_WARNING, 'Strategy_MakeWellFormed: Stray end tag to text'); + } else { + $this->remove(); + if ($e) $e->send(E_WARNING, 'Strategy_MakeWellFormed: Stray end tag removed'); + } + $reprocess = true; + continue; + } + + // do errors, in REVERSE $j order: a,b,c with </a></b></c> + $c = count($skipped_tags); + if ($e) { + for ($j = $c - 1; $j > 0; $j--) { + // notice we exclude $j == 0, i.e. the current ending tag, from + // the errors... [TagClosedSuppress] + if (!isset($skipped_tags[$j]->armor['MakeWellFormed_TagClosedError'])) { + $e->send(E_NOTICE, 'Strategy_MakeWellFormed: Tag closed by element end', $skipped_tags[$j]); + } + } + } + + // insert tags, in FORWARD $j order: c,b,a with </a></b></c> + $replace = array($token); + for ($j = 1; $j < $c; $j++) { + // ...as well as from the insertions + $new_token = new HTMLPurifier_Token_End($skipped_tags[$j]->name); + $new_token->start = $skipped_tags[$j]; + array_unshift($replace, $new_token); + if (isset($definition->info[$new_token->name]) && $definition->info[$new_token->name]->formatting) { + // [TagClosedAuto] + $element = clone $skipped_tags[$j]; + $element->carryover = true; + $element->armor['MakeWellFormed_TagClosedError'] = true; + $replace[] = $element; + } + } + $this->processToken($replace); + $reprocess = true; + continue; + } + + $context->destroy('CurrentNesting'); + $context->destroy('InputTokens'); + $context->destroy('InputIndex'); + $context->destroy('CurrentToken'); + + unset($this->injectors, $this->stack, $this->tokens, $this->t); + return $tokens; + } + + /** + * Processes arbitrary token values for complicated substitution patterns. + * In general: + * + * If $token is an array, it is a list of tokens to substitute for the + * current token. These tokens then get individually processed. If there + * is a leading integer in the list, that integer determines how many + * tokens from the stream should be removed. + * + * If $token is a regular token, it is swapped with the current token. + * + * If $token is false, the current token is deleted. + * + * If $token is an integer, that number of tokens (with the first token + * being the current one) will be deleted. + * + * @param $token Token substitution value + * @param $injector Injector that performed the substitution; default is if + * this is not an injector related operation. + */ + protected function processToken($token, $injector = -1) { + + // normalize forms of token + if (is_object($token)) $token = array(1, $token); + if (is_int($token)) $token = array($token); + if ($token === false) $token = array(1); + if (!is_array($token)) throw new HTMLPurifier_Exception('Invalid token type from injector'); + if (!is_int($token[0])) array_unshift($token, 1); + if ($token[0] === 0) throw new HTMLPurifier_Exception('Deleting zero tokens is not valid'); + + // $token is now an array with the following form: + // array(number nodes to delete, new node 1, new node 2, ...) + + $delete = array_shift($token); + $old = array_splice($this->tokens, $this->t, $delete, $token); + + if ($injector > -1) { + // determine appropriate skips + $oldskip = isset($old[0]) ? $old[0]->skip : array(); + foreach ($token as $object) { + $object->skip = $oldskip; + $object->skip[$injector] = true; + } + } + + } + + /** + * Inserts a token before the current token. Cursor now points to + * this token. You must reprocess after this. + */ + private function insertBefore($token) { + array_splice($this->tokens, $this->t, 0, array($token)); + } + + /** + * Removes current token. Cursor now points to new token occupying previously + * occupied space. You must reprocess after this. + */ + private function remove() { + array_splice($this->tokens, $this->t, 1); + } + + /** + * Swap current token with new token. Cursor points to new token (no + * change). You must reprocess after this. + */ + private function swap($token) { + $this->tokens[$this->t] = $token; + } + +} + +// vim: et sw=4 sts=4 diff --git a/lib/htmlpurifier/library/HTMLPurifier/Strategy/RemoveForeignElements.php b/lib/htmlpurifier/library/HTMLPurifier/Strategy/RemoveForeignElements.php new file mode 100644 index 000000000..bccaf14d3 --- /dev/null +++ b/lib/htmlpurifier/library/HTMLPurifier/Strategy/RemoveForeignElements.php @@ -0,0 +1,188 @@ +<?php + +/** + * Removes all unrecognized tags from the list of tokens. + * + * This strategy iterates through all the tokens and removes unrecognized + * tokens. If a token is not recognized but a TagTransform is defined for + * that element, the element will be transformed accordingly. + */ + +class HTMLPurifier_Strategy_RemoveForeignElements extends HTMLPurifier_Strategy +{ + + public function execute($tokens, $config, $context) { + $definition = $config->getHTMLDefinition(); + $generator = new HTMLPurifier_Generator($config, $context); + $result = array(); + + $escape_invalid_tags = $config->get('Core.EscapeInvalidTags'); + $remove_invalid_img = $config->get('Core.RemoveInvalidImg'); + + // currently only used to determine if comments should be kept + $trusted = $config->get('HTML.Trusted'); + $comment_lookup = $config->get('HTML.AllowedComments'); + $comment_regexp = $config->get('HTML.AllowedCommentsRegexp'); + $check_comments = $comment_lookup !== array() || $comment_regexp !== null; + + $remove_script_contents = $config->get('Core.RemoveScriptContents'); + $hidden_elements = $config->get('Core.HiddenElements'); + + // remove script contents compatibility + if ($remove_script_contents === true) { + $hidden_elements['script'] = true; + } elseif ($remove_script_contents === false && isset($hidden_elements['script'])) { + unset($hidden_elements['script']); + } + + $attr_validator = new HTMLPurifier_AttrValidator(); + + // removes tokens until it reaches a closing tag with its value + $remove_until = false; + + // converts comments into text tokens when this is equal to a tag name + $textify_comments = false; + + $token = false; + $context->register('CurrentToken', $token); + + $e = false; + if ($config->get('Core.CollectErrors')) { + $e =& $context->get('ErrorCollector'); + } + + foreach($tokens as $token) { + if ($remove_until) { + if (empty($token->is_tag) || $token->name !== $remove_until) { + continue; + } + } + if (!empty( $token->is_tag )) { + // DEFINITION CALL + + // before any processing, try to transform the element + if ( + isset($definition->info_tag_transform[$token->name]) + ) { + $original_name = $token->name; + // there is a transformation for this tag + // DEFINITION CALL + $token = $definition-> + info_tag_transform[$token->name]-> + transform($token, $config, $context); + if ($e) $e->send(E_NOTICE, 'Strategy_RemoveForeignElements: Tag transform', $original_name); + } + + if (isset($definition->info[$token->name])) { + + // mostly everything's good, but + // we need to make sure required attributes are in order + if ( + ($token instanceof HTMLPurifier_Token_Start || $token instanceof HTMLPurifier_Token_Empty) && + $definition->info[$token->name]->required_attr && + ($token->name != 'img' || $remove_invalid_img) // ensure config option still works + ) { + $attr_validator->validateToken($token, $config, $context); + $ok = true; + foreach ($definition->info[$token->name]->required_attr as $name) { + if (!isset($token->attr[$name])) { + $ok = false; + break; + } + } + if (!$ok) { + if ($e) $e->send(E_ERROR, 'Strategy_RemoveForeignElements: Missing required attribute', $name); + continue; + } + $token->armor['ValidateAttributes'] = true; + } + + if (isset($hidden_elements[$token->name]) && $token instanceof HTMLPurifier_Token_Start) { + $textify_comments = $token->name; + } elseif ($token->name === $textify_comments && $token instanceof HTMLPurifier_Token_End) { + $textify_comments = false; + } + + } elseif ($escape_invalid_tags) { + // invalid tag, generate HTML representation and insert in + if ($e) $e->send(E_WARNING, 'Strategy_RemoveForeignElements: Foreign element to text'); + $token = new HTMLPurifier_Token_Text( + $generator->generateFromToken($token) + ); + } else { + // check if we need to destroy all of the tag's children + // CAN BE GENERICIZED + if (isset($hidden_elements[$token->name])) { + if ($token instanceof HTMLPurifier_Token_Start) { + $remove_until = $token->name; + } elseif ($token instanceof HTMLPurifier_Token_Empty) { + // do nothing: we're still looking + } else { + $remove_until = false; + } + if ($e) $e->send(E_ERROR, 'Strategy_RemoveForeignElements: Foreign meta element removed'); + } else { + if ($e) $e->send(E_ERROR, 'Strategy_RemoveForeignElements: Foreign element removed'); + } + continue; + } + } elseif ($token instanceof HTMLPurifier_Token_Comment) { + // textify comments in script tags when they are allowed + if ($textify_comments !== false) { + $data = $token->data; + $token = new HTMLPurifier_Token_Text($data); + } elseif ($trusted || $check_comments) { + // always cleanup comments + $trailing_hyphen = false; + if ($e) { + // perform check whether or not there's a trailing hyphen + if (substr($token->data, -1) == '-') { + $trailing_hyphen = true; + } + } + $token->data = rtrim($token->data, '-'); + $found_double_hyphen = false; + while (strpos($token->data, '--') !== false) { + $found_double_hyphen = true; + $token->data = str_replace('--', '-', $token->data); + } + if ($trusted || !empty($comment_lookup[trim($token->data)]) || ($comment_regexp !== NULL && preg_match($comment_regexp, trim($token->data)))) { + // OK good + if ($e) { + if ($trailing_hyphen) { + $e->send(E_NOTICE, 'Strategy_RemoveForeignElements: Trailing hyphen in comment removed'); + } + if ($found_double_hyphen) { + $e->send(E_NOTICE, 'Strategy_RemoveForeignElements: Hyphens in comment collapsed'); + } + } + } else { + if ($e) { + $e->send(E_NOTICE, 'Strategy_RemoveForeignElements: Comment removed'); + } + continue; + } + } else { + // strip comments + if ($e) $e->send(E_NOTICE, 'Strategy_RemoveForeignElements: Comment removed'); + continue; + } + } elseif ($token instanceof HTMLPurifier_Token_Text) { + } else { + continue; + } + $result[] = $token; + } + if ($remove_until && $e) { + // we removed tokens until the end, throw error + $e->send(E_ERROR, 'Strategy_RemoveForeignElements: Token removed to end', $remove_until); + } + + $context->destroy('CurrentToken'); + + return $result; + } + +} + +// vim: et sw=4 sts=4 diff --git a/lib/htmlpurifier/library/HTMLPurifier/Strategy/ValidateAttributes.php b/lib/htmlpurifier/library/HTMLPurifier/Strategy/ValidateAttributes.php new file mode 100644 index 000000000..c3328a9d4 --- /dev/null +++ b/lib/htmlpurifier/library/HTMLPurifier/Strategy/ValidateAttributes.php @@ -0,0 +1,39 @@ +<?php + +/** + * Validate all attributes in the tokens. + */ + +class HTMLPurifier_Strategy_ValidateAttributes extends HTMLPurifier_Strategy +{ + + public function execute($tokens, $config, $context) { + + // setup validator + $validator = new HTMLPurifier_AttrValidator(); + + $token = false; + $context->register('CurrentToken', $token); + + foreach ($tokens as $key => $token) { + + // only process tokens that have attributes, + // namely start and empty tags + if (!$token instanceof HTMLPurifier_Token_Start && !$token instanceof HTMLPurifier_Token_Empty) continue; + + // skip tokens that are armored + if (!empty($token->armor['ValidateAttributes'])) continue; + + // note that we have no facilities here for removing tokens + $validator->validateToken($token, $config, $context); + + $tokens[$key] = $token; // for PHP 4 + } + $context->destroy('CurrentToken'); + + return $tokens; + } + +} + +// vim: et sw=4 sts=4 |