From 5f0663cad75a5444348d7d0892ba84c7338099be Mon Sep 17 00:00:00 2001 From: "Edward Z. Yang" Date: Sun, 24 Jun 2007 17:44:27 +0000 Subject: [PATCH] Refactor MakeWellFormed/Injector for performance and as little code duplication as possible. Also, make AutoParagraph smarter about root nodes that don't like p tags. git-svn-id: http://htmlpurifier.org/svnroot/htmlpurifier/trunk@1221 48356398-32a2-884e-a903-53898d9a118a --- library/HTMLPurifier/Injector.php | 67 +++++++- .../HTMLPurifier/Injector/AutoParagraph.php | 46 +++-- library/HTMLPurifier/Injector/Linkify.php | 19 +-- .../HTMLPurifier/Strategy/MakeWellFormed.php | 157 ++++++++---------- .../Strategy/MakeWellFormedTest.php | 8 + 5 files changed, 169 insertions(+), 128 deletions(-) diff --git a/library/HTMLPurifier/Injector.php b/library/HTMLPurifier/Injector.php index f8af26dc..95fad34d 100644 --- a/library/HTMLPurifier/Injector.php +++ b/library/HTMLPurifier/Injector.php @@ -8,15 +8,78 @@ class HTMLPurifier_Injector { + /** + * Amount of tokens the injector needs to skip + 1. Because + * the decrement is the first thing that happens, this needs to + * be one greater than the "real" skip count. + */ + var $skip = 1; + + /** + * Instance of HTMLPurifier_HTMLDefinition + */ + var $htmlDefinition; + + /** + * Reference to CurrentNesting variable in Context. This is an array + * list of tokens that we are currently "inside" + */ + var $currentNesting; + + /** + * Reference to InputTokens variable in Context. This is an array + * list of the input tokens that are being processed. + */ + var $inputTokens; + + /** + * Reference to InputIndex variable in Context. This is an integer + * array index for $this->inputTokens that indicates what token + * is currently being processed. + */ + var $inputIndex; + + /** + * Prepares the injector by giving it the config and context objects, + * so that important variables can be extracted and not passed via + * parameter constantly. Remember: always instantiate a new injector + * when handling a set of HTML. + */ + function prepare($config, &$context) { + $this->htmlDefinition = $config->getHTMLDefinition(); + $this->currentNesting =& $context->get('CurrentNesting'); + $this->inputTokens =& $context->get('InputTokens'); + $this->inputIndex =& $context->get('InputIndex'); + } + + /** + * Tests if the context node allows a certain element + * @param $name Name of element to test for + * @return True if element is allowed, false if it is not + */ + function allowsElement($name) { + if (!empty($this->currentNesting)) { + $parent_token = array_pop($this->currentNesting); + $this->currentNesting[] = $parent_token; + $parent = $this->htmlDefinition->info[$parent_token->name]; + } else { + $parent = $this->htmlDefinition->info_parent_def; + } + if (!isset($parent->child->elements[$name]) || isset($parent->excludes[$name])) { + return false; + } + return true; + } + /** * Handler that is called when a text token is processed */ - function handleText(&$token, $config, &$context) {} + function handleText(&$token) {} /** * Handler that is called when a start token is processed */ - function handleStart(&$token, $config, &$context) {} + function handleStart(&$token) {} } diff --git a/library/HTMLPurifier/Injector/AutoParagraph.php b/library/HTMLPurifier/Injector/AutoParagraph.php index 54585c48..9eceb8c8 100644 --- a/library/HTMLPurifier/Injector/AutoParagraph.php +++ b/library/HTMLPurifier/Injector/AutoParagraph.php @@ -5,38 +5,36 @@ require_once 'HTMLPurifier/Injector.php'; /** * Injector that auto paragraphs text in the root node based on * double-spacing. - * @todo Don't assume that root node means paragraphing is alright */ class HTMLPurifier_Injector_AutoParagraph extends HTMLPurifier_Injector { - function handleText(&$token, $config, &$context) { - $current_nesting =& $context->get('CurrentNesting'); + function handleText(&$token) { $text = $token->data; // $token is the focus: if processing is needed, it gets // turned into an array of tokens that will replace the // original token - if (empty($current_nesting)) { - // we're in root node, great time to start a paragraph - // since we're also dealing with a text node + if (empty($this->currentNesting)) { + if (!$this->allowsElement('p')) return; + // we're in root node, and the root node allows paragraphs + // start a paragraph since we just hit some text $token = array(new HTMLPurifier_Token_Start('p')); - $this->_splitText($text, $token, $config, $context); - } elseif ($current_nesting[count($current_nesting)-1]->name == 'p') { + $this->_splitText($text, $token); + } elseif ($this->currentNesting[count($this->currentNesting)-1]->name == 'p') { // we're not in root node but we're in a paragraph, so don't // add a paragraph start tag but still perform processing $token = array(); - $this->_splitText($text, $token, $config, $context); + $this->_splitText($text, $token); } } - function handleStart(&$token, $config, &$context) { + function handleStart(&$token) { // check if we're inside a tag already, if so, don't add // paragraph tags - $current_nesting = $context->get('CurrentNesting'); - if (!empty($current_nesting)) return; + if (!empty($this->currentNesting)) return; // check if the start tag counts as a "block" element - if (!$this->_isInline($token, $config)) return; + if (!$this->_isInline($token)) return; // append a paragraph tag before the token $token = array(new HTMLPurifier_Token_Start('p'), $token); @@ -53,7 +51,7 @@ class HTMLPurifier_Injector_AutoParagraph extends HTMLPurifier_Injector * @param $context Instance of HTMLPurifier_Context * @private */ - function _splitText($data, &$result, $config, &$context) { + function _splitText($data, &$result) { $raw_paragraphs = explode(PHP_EOL . PHP_EOL, $data); // remove empty paragraphs @@ -78,7 +76,7 @@ class HTMLPurifier_Injector_AutoParagraph extends HTMLPurifier_Injector // check the outside to determine whether or not the // end paragraph tag should be removed - if ($this->_removeParagraphEnd($config, $context)) { + if ($this->_removeParagraphEnd()) { array_pop($result); } @@ -89,19 +87,15 @@ class HTMLPurifier_Injector_AutoParagraph extends HTMLPurifier_Injector * Returns boolean whether or not to remove the paragraph end tag * that was automatically added. The paragraph end tag should be * removed unless the next token is a paragraph or block element. - * @param $config Instance of HTMLPurifier_Config - * @param $context Instance of HTMLPurifier_Context * @private */ - function _removeParagraphEnd($config, &$context) { - $tokens = $context->get('InputTokens'); - $i = $context->get('InputIndex'); + function _removeParagraphEnd() { + $tokens =& $this->inputTokens; $remove_paragraph_end = true; // Start of the checks one after the current token's index - for ($i++; isset($tokens[$i]); $i++) { + for ($i = $this->inputIndex + 1; isset($tokens[$i]); $i++) { if ($tokens[$i]->type == 'start' || $tokens[$i]->type == 'empty') { - $definition = $config->getHTMLDefinition(); - $remove_paragraph_end = $this->_isInline($tokens[$i], $config); + $remove_paragraph_end = $this->_isInline($tokens[$i]); break; } // check if we can abort early (whitespace means we carry-on!) @@ -114,10 +108,10 @@ class HTMLPurifier_Injector_AutoParagraph extends HTMLPurifier_Injector /** * Returns true if passed token is inline (and, ergo, allowed in * paragraph tags) + * @private */ - function _isInline($token, $config) { - $definition = $config->getHTMLDefinition(); - return isset($definition->info['p']->child->elements[$token->name]); + function _isInline($token) { + return isset($this->htmlDefinition->info['p']->child->elements[$token->name]); } } diff --git a/library/HTMLPurifier/Injector/Linkify.php b/library/HTMLPurifier/Injector/Linkify.php index 9ef162f0..76bdf140 100644 --- a/library/HTMLPurifier/Injector/Linkify.php +++ b/library/HTMLPurifier/Injector/Linkify.php @@ -9,28 +9,17 @@ class HTMLPurifier_Injector_Linkify extends HTMLPurifier_Injector { function handleText(&$token, $config, &$context) { - $current_nesting =& $context->get('CurrentNesting'); - // this snippet could be factored out - $definition = $config->getHTMLDefinition(); - if (!empty($current_nesting)) { - $parent_token = array_pop($current_nesting); - $current_nesting[] = $parent_token; - $parent = $definition->info[$parent_token->name]; - } else { - $parent = $definition->info_parent_def; - } - if (!isset($parent->child->elements['a']) || isset($parent->excludes['a'])) { - // parent element does not allow link elements, don't bother - return; - } + if (!$this->allowsElement('a')) return; + if (strpos($token->data, '://') === false) { // our really quick heuristic failed, abort // this may not work so well if we want to match things like - // "google.com" + // "google.com", but then again, most people don't return; } // there is/are URL(s). Let's split the string: + // Note: this regex is extremely permissive $bits = preg_split('#((?:https?|ftp)://[^\s\'"<>()]+)#S', $token->data, -1, PREG_SPLIT_DELIM_CAPTURE); $token = array(); diff --git a/library/HTMLPurifier/Strategy/MakeWellFormed.php b/library/HTMLPurifier/Strategy/MakeWellFormed.php index 3029f62e..c90bb84a 100644 --- a/library/HTMLPurifier/Strategy/MakeWellFormed.php +++ b/library/HTMLPurifier/Strategy/MakeWellFormed.php @@ -33,38 +33,49 @@ HTMLPurifier_ConfigSchema::define( class HTMLPurifier_Strategy_MakeWellFormed extends HTMLPurifier_Strategy { + /** + * Locally shared variable references + * @private + */ + var $inputTokens, $inputIndex, $outputTokens, $currentNesting, + $currentInjector, $injectors; + function execute($tokens, $config, &$context) { $definition = $config->getHTMLDefinition(); + + // CurrentNesting + $this->currentNesting = array(); + $context->register('CurrentNesting', $this->currentNesting); + + // InputIndex + $this->inputIndex = false; + $context->register('InputIndex', $this->inputIndex); + + // InputTokens + $context->register('InputTokens', $tokens); + $this->inputTokens =& $tokens; + + // OutputTokens + $result = array(); + $this->outputTokens =& $result; + + // %Core.EscapeInvalidTags + $escape_invalid_tags = $config->get('Core', 'EscapeInvalidTags'); $generator = new HTMLPurifier_Generator(); - $current_nesting = array(); - $context->register('CurrentNesting', $current_nesting); - - $tokens_index = null; - $context->register('InputIndex', $tokens_index); - $context->register('InputTokens', $tokens); - - $result = array(); - $context->register('OutputTokens', $result); - - $escape_invalid_tags = $config->get('Core', 'EscapeInvalidTags'); - // -- begin INJECTOR -- - // factor this stuff out to its own class - $injector = array(); - $injector_skip = array(); + $this->injectors = array(); + // we need a generic way of adding injectors, and also its own + // configuration namespace if ($config->get('Core', 'AutoParagraph')) { - $injector[] = new HTMLPurifier_Injector_AutoParagraph(); - // decrement happens first, so set to one so we start at zero - $injector_skip[] = 1; + $this->injectors[] = new HTMLPurifier_Injector_AutoParagraph(); } if ($config->get('Core', 'AutoLinkify')) { - $injector[] = new HTMLPurifier_Injector_Linkify(); - $injector_skip[] = 1; + $this->injectors[] = new HTMLPurifier_Injector_Linkify(); } // array index of the injector that resulted in an array @@ -72,64 +83,55 @@ class HTMLPurifier_Strategy_MakeWellFormed extends HTMLPurifier_Strategy // injectors are affected by the added tokens and which are // not (namely, the ones after the current injector are not // affected) - $current_injector = false; + $this->currentInjector = false; - $context->register('Injector', $injector); - $context->register('CurrentInjector', $current_injector); - - // number of tokens to skip + 1 - // before processing, this gets decremented: if it equals zero, - // it means the injector is active and is processing tokens, if - // it is greater than zero, then it is inactive, presumably having - // been the source of the tokens - $context->register('InjectorSkip', $injector_skip); + // give the injectors references to the definition and context + // variables for performance reasons + foreach ($this->injectors as $i => $x) { + $this->injectors[$i]->prepare($config, $context); + } // -- end INJECTOR -- - for ($tokens_index = 0; isset($tokens[$tokens_index]); $tokens_index++) { + for ($this->inputIndex = 0; isset($tokens[$this->inputIndex]); $this->inputIndex++) { // if all goes well, this token will be passed through unharmed - $token = $tokens[$tokens_index]; + $token = $tokens[$this->inputIndex]; - foreach ($injector as $i => $x) { - if ($injector_skip[$i] > 0) $injector_skip[$i]--; + foreach ($this->injectors as $i => $x) { + if ($x->skip > 0) $this->injectors[$i]->skip--; } // quick-check: if it's not a tag, no need to process if (empty( $token->is_tag )) { - - // duplicated with handleStart if ($token->type === 'text') { - foreach ($injector as $i => $x) { - if (!$injector_skip[$i]) { - $x->handleText($token, $config, $context); - } + // injector handler code; duplicated for performance reasons + foreach ($this->injectors as $i => $x) { + if (!$x->skip) $x->handleText($token, $config, $context); if (is_array($token)) { - $current_injector = $i; + $this->currentInjector = $i; break; } } } - $this->processToken($token, $config, $context); continue; } $info = $definition->info[$token->name]->child; + // quick checks: // test if it claims to be a start tag but is empty if ($info->type == 'empty' && $token->type == 'start') { $result[] = new HTMLPurifier_Token_Empty($token->name, $token->attr); continue; } - // test if it claims to be empty but really is a start tag if ($info->type != 'empty' && $token->type == 'empty' ) { $result[] = new HTMLPurifier_Token_Start($token->name, $token->attr); $result[] = new HTMLPurifier_Token_End($token->name); continue; } - // automatically insert empty tags if ($token->type == 'empty') { $result[] = $token; @@ -140,9 +142,9 @@ class HTMLPurifier_Strategy_MakeWellFormed extends HTMLPurifier_Strategy if ($token->type == 'start') { // ...unless they also have to close their parent - if (!empty($current_nesting)) { + if (!empty($this->currentNesting)) { - $parent = array_pop($current_nesting); + $parent = array_pop($this->currentNesting); $parent_info = $definition->info[$parent->name]; // this can be replaced with a more general algorithm: @@ -152,20 +154,18 @@ class HTMLPurifier_Strategy_MakeWellFormed extends HTMLPurifier_Strategy // close the parent, then append the token $result[] = new HTMLPurifier_Token_End($parent->name); $result[] = $token; - $current_nesting[] = $token; + $this->currentNesting[] = $token; continue; } - $current_nesting[] = $parent; // undo the pop + $this->currentNesting[] = $parent; // undo the pop } - // injectors - foreach ($injector as $i => $x) { - if (!$injector_skip[$i]) { - $x->handleStart($token, $config, $context); - } + // injector handler code; duplicated for performance reasons + foreach ($this->injectors as $i => $x) { + if (!$x->skip[$i]) $x->handleStart($token, $config, $context); if (is_array($token)) { - $current_injector = $i; + $this->currentInjector = $i; break; } } @@ -178,7 +178,7 @@ class HTMLPurifier_Strategy_MakeWellFormed extends HTMLPurifier_Strategy if ($token->type != 'end') continue; // make sure that we have something open - if (empty($current_nesting)) { + if (empty($this->currentNesting)) { if ($escape_invalid_tags) { $result[] = new HTMLPurifier_Token_Text( $generator->generateFromToken($token, $config, $context) @@ -188,7 +188,7 @@ class HTMLPurifier_Strategy_MakeWellFormed extends HTMLPurifier_Strategy } // first, check for the simplest case: everything closes neatly - $current_parent = array_pop($current_nesting); + $current_parent = array_pop($this->currentNesting); if ($current_parent->name == $token->name) { $result[] = $token; continue; @@ -197,17 +197,17 @@ class HTMLPurifier_Strategy_MakeWellFormed extends HTMLPurifier_Strategy // okay, so we're trying to close the wrong tag // undo the pop previous pop - $current_nesting[] = $current_parent; + $this->currentNesting[] = $current_parent; // scroll back the entire nest, trying to find our tag. // (feature could be to specify how far you'd like to go) - $size = count($current_nesting); + $size = count($this->currentNesting); // -2 because -1 is the last element, but we already checked that $skipped_tags = false; for ($i = $size - 2; $i >= 0; $i--) { - if ($current_nesting[$i]->name == $token->name) { + if ($this->currentNesting[$i]->name == $token->name) { // current nesting is modified - $skipped_tags = array_splice($current_nesting, $i); + $skipped_tags = array_splice($this->currentNesting, $i); break; } } @@ -234,22 +234,20 @@ class HTMLPurifier_Strategy_MakeWellFormed extends HTMLPurifier_Strategy // we're at the end now, fix all still unclosed tags // not using processToken() because at this point we don't // care about current nesting - if (!empty($current_nesting)) { - $size = count($current_nesting); + if (!empty($this->currentNesting)) { + $size = count($this->currentNesting); for ($i = $size - 1; $i >= 0; $i--) { $result[] = - new HTMLPurifier_Token_End($current_nesting[$i]->name); + new HTMLPurifier_Token_End($this->currentNesting[$i]->name); } } $context->destroy('CurrentNesting'); $context->destroy('InputTokens'); $context->destroy('InputIndex'); - $context->destroy('OutputTokens'); - $context->destroy('Injector'); - $context->destroy('CurrentInjector'); - $context->destroy('InjectorSkip'); + unset($this->outputTokens, $this->injectors, $this->currentInjector, + $this->currentNesting, $this->inputTokens, $this->inputIndex); return $result; } @@ -259,33 +257,22 @@ class HTMLPurifier_Strategy_MakeWellFormed extends HTMLPurifier_Strategy // the original token was overloaded by an injector, time // to some fancy acrobatics - $tokens =& $context->get('InputTokens'); - $tokens_index =& $context->get('InputIndex'); - // $tokens_index is decremented so that the entire set gets + // $this->inputIndex is decremented so that the entire set gets // re-processed - array_splice($tokens, $tokens_index--, 1, $token); + array_splice($this->inputTokens, $this->inputIndex--, 1, $token); // adjust the injector skips based on the array substitution - $injector_skip =& $context->get('InjectorSkip'); - $current_injector =& $context->get('CurrentInjector'); - $offset = count($token) + 1; - for ($i = 0; $i <= $current_injector; $i++) { - $injector_skip[$i] += $offset; + for ($i = 0; $i <= $this->currentInjector; $i++) { + $this->injectors[$i]->skip += $offset; } - } elseif ($token) { // regular case - $result =& $context->get('OutputTokens'); - $current_nesting =& $context->get('CurrentNesting'); - $result[] = $token; + $this->outputTokens[] = $token; if ($token->type == 'start') { - $current_nesting[] = $token; + $this->currentNesting[] = $token; } elseif ($token->type == 'end') { - // theoretical: this code doesn't get run because performing - // the calculations inline is more efficient, and - // end tokens (currently) do not cause a handler invocation - array_pop($current_nesting); + array_pop($this->currentNesting); // not actually used } } } diff --git a/tests/HTMLPurifier/Strategy/MakeWellFormedTest.php b/tests/HTMLPurifier/Strategy/MakeWellFormedTest.php index 92d19f58..a6025027 100644 --- a/tests/HTMLPurifier/Strategy/MakeWellFormedTest.php +++ b/tests/HTMLPurifier/Strategy/MakeWellFormedTest.php @@ -175,6 +175,14 @@ Par '

Par

' ); + $this->assertResult( +'Par + +Par2', + true, + array('Core.AutoParagraph' => true, 'HTML.Parent' => 'span') + ); + } function testLinkify() {