/**
* Turns HTML into text browser visible text. Images
* are converted to their alt text and tags are supressed.
* Entities are converted to their visible representation.
* @param string $html HTML to convert.
* @return string Plain text.
* @access public
* @static
*/
function normalise($html)
{
$text = preg_replace('|<img.*?alt\\s*=\\s*"(.*?)".*?>|', ' \\1 ', $html);
$text = preg_replace('|<img.*?alt\\s*=\\s*\'(.*?)\'.*?>|', ' \\1 ', $text);
$text = preg_replace('|<img.*?alt\\s*=\\s*([a-zA-Z_]+).*?>|', ' \\1 ', $text);
$text = preg_replace('|<.*?>|', '', $text);
$text = SimpleSaxParser::decodeHtml($text);
$text = preg_replace('|\\s+|', ' ', $text);
return trim($text);
}
function setUp()
{
$this->_handler =& new MockSimpleSaxParser($this);
$this->_handler->setReturnValue("acceptStartToken", true);
$this->_handler->setReturnValue("acceptEndToken", true);
$this->_handler->setReturnValue("acceptAttributeToken", true);
$this->_handler->setReturnValue("acceptEntityToken", true);
$this->_handler->setReturnValue("acceptTextToken", true);
$this->_handler->setReturnValue("ignore", true);
$this->_lexer =& SimpleSaxParser::createLexer($this->_handler);
}
/**
* Reads the raw content and send events
* into the page to be built.
* @param string $raw Unparsed text.
* @param SimpleSaxParser $parser Event generator.
* @access public
*/
function parse($raw, &$parser)
{
return $parser->parse($raw);
}
/**
* Pattern matches to parse the inside of a tag
* including the attributes and their quoting.
* @param SimpleLexer $lexer Lexer to add patterns to.
* @access private
* @static
*/
function _addInTagTokens(&$lexer) {
$lexer->mapHandler('tag', 'acceptStartToken');
$lexer->addSpecialPattern('\s+', 'tag', 'ignore');
SimpleSaxParser::_addAttributeTokens($lexer);
$lexer->addExitPattern('>', 'tag');
}
/**
* Accessor for plain text of page as a text browser
* would see it.
* @return string Plain text of page.
* @access public
*/
function getText()
{
if (!$this->_text) {
$this->_text = SimpleSaxParser::normalise($this->_raw);
}
return $this->_text;
}
/**
* Accessor for content reduced to visible text. Acts
* like a text mode browser, normalising space and
* reducing images to their alt text.
* @return string Content as plain text.
* @access public
*/
function getText()
{
return SimpleSaxParser::normalise($this->_content);
}
function testHtmlEntityTranslation()
{
$this->assertEqual(SimpleSaxParser::normalise('<>"&'), '<>"&');
}
