[Sommaire] [Imprimer]

   1  <?php
   2  
   3  require_once  'Horde/String.php';
   4  
   5  /**
   6   * Takes HTML and converts it to formatted, plain text.
   7   *
   8   * Parameters:
   9   * <pre>
  10   * charset -- The charset to use for html_entity_decode() calls.
  11   * width   -- The wrapping width.
  12   * wrap    -- Whether to wrap the text or not.
  13   * </pre>
  14   *
  15   * $Horde: framework/Text_Filter/Filter/html2text.php,v 1.4.10.12 2006/01/01 21:28:38 jan Exp $
  16   *
  17   * Copyright 2003-2006 Jon Abernathy <jon@chuggnutt.com>
  18   * Original source: http://www.chuggnutt.com/html2text.php
  19   * Copyright 2004-2006 Jan Schneider <jan@horde.org>
  20   *
  21   * See the enclosed file COPYING for license information (LGPL). If you did
  22   * not receive this file, see http://www.fsf.org/copyleft/lgpl.html.
  23   *
  24   * @author  Jon Abernathy <jon@chuggnutt.com>
  25   * @author  Jan Schneider <jan@horde.org>
  26   * @since   Horde 3.0
  27   * @package Horde_Text
  28   */
  29  class Text_Filter_html2text extends Text_Filter {
  30  
  31      /**
  32       * Filter parameters.
  33       *
  34       * @var array
  35       */
  36      var $_params = array('charset' => null,
  37                           'width' => 70,
  38                           'wrap' => true);
  39  
  40      /**
  41       * Executes any code necessaray before applying the filter patterns.
  42       *
  43       * @param string $text  The text before the filtering.
  44       *
  45       * @return string  The modified text.
  46       */
  47      function preProcess($text)
  48      {
  49          global $_html2text_state;
  50  
  51          if (is_null($this->_params['charset'])) {
  52              $this->_params['charset'] = isset($GLOBALS['_HORDE_STRING_CHARSET']) ? $GLOBALS['_HORDE_STRING_CHARSET'] : 'ISO-8859-1';
  53          }
  54  
  55          $_html2text_state['linkList'] = '';
  56          $_html2text_state['linkCount'] = 0;
  57  
  58          return trim($text);
  59      }
  60  
  61      /**
  62       * Returns a hash with replace patterns.
  63       *
  64       * @return array  Patterns hash.
  65       */
  66      function getPatterns()
  67      {
  68          $regexp = array(
  69              // Non-legal carriage return.
  70              '/\r/' => '',
  71  
  72              // Leading and trailing whitespace.
  73              '/^\s*(.*?)\s*$/m' => '\1',
  74  
  75              // Normalize <br>.
  76              '/<br[^>]*>([^\n])/i' => "<br>\n\\1",
  77  
  78              // Newlines and tabs.
  79              '/[\n\t]+/' => ' ',
  80  
  81              // <script>s -- which strip_tags() supposedly has problems with.
  82              '/<script[^>]*>.*?<\/script>/i' => '',
  83  
  84              // <style>s -- which strip_tags() supposedly has problems with.
  85              '/<style[^>]*>.*?<\/style>/i' => '',
  86  
  87              // Comments -- which strip_tags() might have a problem with.
  88              // //'/<!-- .* -->/' => '',
  89  
  90              // h1 - h3
  91              '/<h[123][^>]*>(.+?)<\/h[123]> ?/ie' => 'strtoupper("\n\n" . \'\1\' . "\n\n")',
  92  
  93              // h4 - h6
  94              '/<h[456][^>]*>(.+?)<\/h[456]> ?/ie' => 'ucwords("\n\n" . \'\1\' . "\n\n")',
  95  
  96              // <p>
  97              '/<p[^>]*> ?/i' => "\n\n  ",
  98  
  99              // <br>
 100              '/<br[^>]*> ?/i' => "\n",
 101  
 102              // <b>
 103              '/<b[^>]*>(.+?)<\/b>/ie' => 'strtoupper(\'\1\')',
 104  
 105              // <strong>
 106              '/<strong[^>]*>(.+?)<\/strong>/ie' => 'strtoupper(\'\1\')',
 107              '/<span\\s+style="font-weight:\\s*bold.*">(.+?)<\/span>/ie' => 'strtoupper(\'\1\')',
 108  
 109              // <i>
 110              '/<i[^>]*>(.+?)<\/i>/i' => '/\\1/',
 111  
 112              // <em>
 113              '/<em[^>]*>(.+?)<\/em>/i' => '/\\1/',
 114  
 115              // <u>
 116              '/<u[^>]*>(.+?)<\/u>/i' => '_\\1_',
 117  
 118              // <ul>/<ol> and </ul>/</ol>
 119              '/(<(u|o)l[^>]*>| ?<\/(u|o)l>) ?/i' => "\n\n",
 120  
 121              // <li>
 122              '/ ?<li[^>]*>/i' => "\n  * ",
 123  
 124              // <a href="">
 125              '/<a href="([^"]+)"[^>]*>(.+?)<\/a>/ie' => 'Text_Filter_html2text::_buildLinkList($GLOBALS["_html2text_state"]["linkCount"], \'\1\', \'\2\')',
 126  
 127              // <hr>
 128              '/<hr[^>]*> ?/i' => "\n-------------------------\n",
 129  
 130              // <table> and </table>
 131              '/(<table[^>]*>| ?<\/table>) ?/i' => "\n\n",
 132  
 133              // <tr>
 134              '/ ?<tr[^>]*> ?/i' => "\n\t",
 135  
 136              // <td> and </td>
 137              '/ ?<td[^>]*>(.+?)<\/td> ?/i' => '\1' . "\t\t",
 138              '/\t\t<\/tr>/i' => '',
 139  
 140              // entities
 141              '/&nbsp;/i' => ' ',
 142              '/&trade;/i' => '(tm)',
 143              '/&#(\d+);/e' => 'String::convertCharset(Text_Filter_html2text::_int2utf8(\'\1\'), "UTF-8", "' . $this->_params['charset'] . '")',
 144  
 145              // Some mailers (e.g. Hotmail) use the following div tag as a way
 146              // to define a block of text.
 147              '/<div class=rte>(.+?)<\/div> ?/i' => '\1' . "\n"
 148          );
 149  
 150          return array('regexp' => $regexp);
 151      }
 152  
 153      /**
 154       * Executes any code necessaray after applying the filter
 155       * patterns.
 156       *
 157       * @param string $text  The text after the filtering.
 158       *
 159       * @return string  The modified text.
 160       */
 161      function postProcess($text)
 162      {
 163          global $_html2text_state;
 164  
 165          /* Strip any other HTML tags. */
 166          $text = strip_tags($text);
 167  
 168          /* Convert html entities. */
 169          $text = @html_entity_decode($text, ENT_QUOTES, $this->_params['charset']);
 170  
 171          /* Bring down number of empty lines to 2 max. */
 172          $text = preg_replace("/\n[[:space:]]+\n/", "\n\n", $text);
 173          $text = preg_replace("/[\n]{3,}/", "\n\n", $text);
 174  
 175          /* Wrap the text to a readable format. */
 176          if ($this->_params['wrap']) {
 177              $text = wordwrap($text, $this->_params['width']);
 178          }
 179  
 180          /* Add link list. */
 181          if (!empty($_html2text_state['linkList'])) {
 182              $text .= "\n\n" . _("Links") . ":\n" . str_repeat('-', String::length(_("Links")) + 1) . "\n" . $_html2text_state['linkList'];
 183          }
 184  
 185          return $text;
 186      }
 187  
 188      /**
 189       * Returns the UTF-8 character sequence of a Unicode value.
 190       *
 191       * @param integer $num  A Unicode value.
 192       *
 193       * @return string  The UTF-8 string.
 194       */
 195      function _int2utf8($num)
 196      {
 197          if ($num < 128) {
 198              return chr($num);
 199          }
 200          if ($num < 2048) {
 201              return chr(($num >> 6) + 192) . chr(($num & 63) + 128);
 202          }
 203          if ($num < 65536) {
 204              return chr(($num >> 12) + 224) . chr((($num >> 6) & 63) + 128) .
 205                  chr(($num & 63) + 128);
 206          }
 207          if ($num < 2097152) {
 208              return chr(($num >> 18) + 240) . chr((($num >> 12) & 63) + 128) .
 209                  chr((($num >> 6) & 63) + 128) . chr(($num & 63) + 128);
 210          }
 211          return '';
 212      }
 213  
 214      /**
 215       * Helper function called by preg_replace() on link replacement.
 216       *
 217       * Maintains an internal list of links to be displayed at the end
 218       * of the text, with numeric indices to the original point in the
 219       * text they appeared.
 220       *
 221       * @access private
 222       *
 223       * @param integer $link_count  Counter tracking current link number.
 224       * @param string  $link        URL of the link.
 225       * @param string  $display     Part of the text to associate number with.
 226       *
 227       * @return string  The link replacement.
 228       */
 229      function _buildLinkList($link_count, $link, $display)
 230      {
 231          global $_html2text_state;
 232  
 233          $_html2text_state['linkCount']++;
 234          $_html2text_state['linkList'] .= '[' . $_html2text_state['linkCount'] . "] $link\n";
 235          return $display . '[' . $_html2text_state['linkCount'] . ']';
 236      }
 237  
 238  }