search/Zend/Search/Lucene/Document/Html.php

   1 <?php
   2 /**
   3  * Zend Framework
   4  *
   5  * LICENSE
   6  *
   7  * This source file is subject to the new BSD license that is bundled
   8  * with this package in the file LICENSE.txt.
   9  * It is also available through the world-wide-web at this URL:
  10  * http://framework.zend.com/license/new-bsd
  11  * If you did not receive a copy of the license and are unable to
  12  * obtain it through the world-wide-web, please send an email
  13  * to license@zend.com so we can send you a copy immediately.
  14  *
  15  * @category   Zend
  16  * @package    Zend_Search_Lucene
  17  * @subpackage Document
  18  * @copyright  Copyright (c) 2005-2007 Zend Technologies USA Inc. (http://www.zend.com)
  19  * @license    http://framework.zend.com/license/new-bsd     New BSD License
  20  */
  21
  22
  23 /** Zend_Search_Lucene_Document */
  24 require_once $CFG->dirroot.'/search/Zend/Search/Lucene/Document.php';
  25
  26
  27 /**
  28  * HTML document.
  29  *
  30  * @category   Zend
  31  * @package    Zend_Search_Lucene
  32  * @subpackage Document
  33  * @copyright  Copyright (c) 2005-2007 Zend Technologies USA Inc. (http://www.zend.com)
  34  * @license    http://framework.zend.com/license/new-bsd     New BSD License
  35  */
  36 class Zend_Search_Lucene_Document_Html extends Zend_Search_Lucene_Document
  37 {
  38     /**
  39      * List of document links
  40      *
  41      * @var array
  42      */
  43     private $_links = array();
  44
  45     /**
  46      * List of document header links
  47      *
  48      * @var array
  49      */
  50     private $_headerLinks = array();
  51
  52     /**
  53      * Stored DOM representation
  54      *
  55      * @var DOMDocument
  56      */
  57     private $_doc;
  58
  59     /**
  60      * Object constructor
  61      *
  62      * @param string  $data
  63      * @param boolean $isFile
  64      * @param boolean $storeContent
  65      */
  66     private function __construct($data, $isFile, $storeContent)
  67     {
  68         $this->_doc = new DOMDocument();
  69         $this->_doc->substituteEntities = true;
  70
  71         if ($isFile) {
  72             @$this->_doc->loadHTMLFile($data);
  73         } else{
  74             @$this->_doc->loadHTML($data);
  75         }
  76
  77         $xpath = new DOMXPath($this->_doc);
  78
  79         $docTitle = '';
  80         $titleNodes = $xpath->query('/html/head/title');
  81         foreach ($titleNodes as $titleNode) {
  82             // title should always have only one entry, but we process all nodeset entries
  83             $docTitle .= $titleNode->nodeValue . ' ';
  84         }
  85         $this->addField(Zend_Search_Lucene_Field::Text('title', $docTitle, $this->_doc->actualEncoding));
  86
  87         $metaNodes = $xpath->query('/html/head/meta[@name]');
  88         foreach ($metaNodes as $metaNode) {
  89             $this->addField(Zend_Search_Lucene_Field::Text($metaNode->getAttribute('name'),
  90                                                            $metaNode->getAttribute('content'),
  91                                                            $this->_doc->actualEncoding));
  92         }
  93
  94         $docBody = '';
  95         $bodyNodes = $xpath->query('/html/body');
  96         foreach ($bodyNodes as $bodyNode) {
  97             // body should always have only one entry, but we process all nodeset entries
  98             $this->_retrieveNodeText($bodyNode, $docBody);
  99         }
 100         if ($storeContent) {
 101             $this->addField(Zend_Search_Lucene_Field::Text('body', $docBody, $this->_doc->actualEncoding));
 102         } else {
 103             $this->addField(Zend_Search_Lucene_Field::UnStored('body', $docBody, $this->_doc->actualEncoding));
 104         }
 105
 106         $linkNodes = $this->_doc->getElementsByTagName('a');
 107         foreach ($linkNodes as $linkNode) {
 108             if (($href = $linkNode->getAttribute('href')) != '') {
 109                 $this->_links[] = $href;
 110             }
 111         }
 112         $this->_links = array_unique($this->_links);
 113
 114         $linkNodes = $xpath->query('/html/head/link');
 115         foreach ($linkNodes as $linkNode) {
 116             if (($href = $linkNode->getAttribute('href')) != '') {
 117                 $this->_headerLinks[] = $href;
 118             }
 119         }
 120         $this->_headerLinks = array_unique($this->_headerLinks);
 121     }
 122
 123     /**
 124      * Get node text
 125      *
 126      * We should exclude scripts, which may be not included into comment tags, CDATA sections,
 127      *
 128      * @param DOMNode $node
 129      * @param string &$text
 130      */
 131     private function _retrieveNodeText(DOMNode $node, &$text)
 132     {
 133         if ($node->nodeType == XML_TEXT_NODE) {
 134             $text .= $node->nodeValue ;
 135             $text .= ' ';
 136         } else if ($node->nodeType == XML_ELEMENT_NODE  &&  $node->nodeName != 'script') {
 137             foreach ($node->childNodes as $childNode) {
 138                 $this->_retrieveNodeText($childNode, $text);
 139             }
 140         }
 141     }
 142
 143     /**
 144      * Get document HREF links
 145      *
 146      * @return array
 147      */
 148     public function getLinks()
 149     {
 150         return $this->_links;
 151     }
 152
 153     /**
 154      * Get document header links
 155      *
 156      * @return array
 157      */
 158     public function getHeaderLinks()
 159     {
 160         return $this->_headerLinks;
 161     }
 162
 163     /**
 164      * Load HTML document from a string
 165      *
 166      * @param string $data
 167      * @param boolean $storeContent
 168      * @return Zend_Search_Lucene_Document_Html
 169      */
 170     public static function loadHTML($data, $storeContent = false)
 171     {
 172         return new Zend_Search_Lucene_Document_Html($data, false, $storeContent);
 173     }
 174
 175     /**
 176      * Load HTML document from a file
 177      *
 178      * @param string $file
 179      * @param boolean $storeContent
 180      * @return Zend_Search_Lucene_Document_Html
 181      */
 182     public static function loadHTMLFile($file, $storeContent = false)
 183     {
 184         return new Zend_Search_Lucene_Document_Html($file, true, $storeContent);
 185     }
 186
 187
 188     /**
 189      * Highlight text in text node
 190      *
 191      * @param DOMText $node
 192      * @param array   $wordsToHighlight
 193      * @param string  $color
 194      */
 195     public function _highlightTextNode(DOMText $node, $wordsToHighlight, $color)
 196     {
 197         $analyzer = Zend_Search_Lucene_Analysis_Analyzer::getDefault();
 198         $analyzer->setInput($node->nodeValue, $this->_doc->encoding);
 199
 200         $matchedTokens = array();
 201
 202         while (($token = $analyzer->nextToken()) !== null) {
 203             if (isset($wordsToHighlight[$token->getTermText()])) {
 204                 $matchedTokens[] = $token;
 205             }
 206         }
 207
 208         if (count($matchedTokens) == 0) {
 209             return;
 210         }
 211
 212         $matchedTokens = array_reverse($matchedTokens);
 213
 214         foreach ($matchedTokens as $token) {
 215             // Cut text after matched token
 216             $node->splitText($token->getEndOffset());
 217
 218             // Cut matched node
 219             $matchedWordNode = $node->splitText($token->getStartOffset());
 220
 221             $highlightedNode = $this->_doc->createElement('b', $matchedWordNode->nodeValue);
 222             $highlightedNode->setAttribute('style', 'color:black;background-color:' . $color);
 223
 224             $node->parentNode->replaceChild($highlightedNode, $matchedWordNode);
 225         }
 226     }
 227
 228
 229     /**
 230      * highlight words in content of the specified node
 231      *
 232      * @param DOMNode $contextNode
 233      * @param array $wordsToHighlight
 234      * @param string $color
 235      */
 236     public function _highlightNode(DOMNode $contextNode, $wordsToHighlight, $color)
 237     {
 238         $textNodes = array();
 239
 240         if (!$contextNode->hasChildNodes()) {
 241             return;
 242         }
 243
 244         foreach ($contextNode->childNodes as $childNode) {
 245             if ($childNode->nodeType == XML_TEXT_NODE) {
 246                 // process node later to leave childNodes structure untouched
 247                 $textNodes[] = $childNode;
 248             } else {
 249                 // Skip script nodes
 250                 if ($childNode->nodeName != 'script') {
 251                     $this->_highlightNode($childNode, $wordsToHighlight, $color);
 252                 }
 253             }
 254         }
 255
 256         foreach ($textNodes as $textNode) {
 257             $this->_highlightTextNode($textNode, $wordsToHighlight, $color);
 258         }
 259     }
 260
 261
 262
 263     /**
 264      * Highlight text with specified color
 265      *
 266      * @param string|array $words
 267      * @param string $color
 268      * @return string
 269      */
 270     public function highlight($words, $color = '#66ffff')
 271     {
 272         if (!is_array($words)) {
 273             $words = array($words);
 274         }
 275         $wordsToHighlight = array();
 276
 277         $analyzer = Zend_Search_Lucene_Analysis_Analyzer::getDefault();
 278         foreach ($words as $wordString) {
 279             $wordsToHighlight = array_merge($wordsToHighlight, $analyzer->tokenize($wordString));
 280         }
 281
 282         if (count($wordsToHighlight) == 0) {
 283             return $this->_doc->saveHTML();
 284         }
 285
 286         $wordsToHighlightFlipped = array();
 287         foreach ($wordsToHighlight as $id => $token) {
 288             $wordsToHighlightFlipped[$token->getTermText()] = $id;
 289         }
 290
 291         $xpath = new DOMXPath($this->_doc);
 292
 293         $matchedNodes = $xpath->query("/html/body/*");
 294         foreach ($matchedNodes as $matchedNode) {
 295             $this->_highlightNode($matchedNode, $wordsToHighlightFlipped, $color);
 296         }
 297
 298     }
 299
 300     /**
 301      * Get HTML
 302      *
 303      * @return string
 304      */
 305     public function getHTML()
 306     {
 307         return $this->_doc->saveHTML();
 308     }
 309 }
 310