/**
* Handle document - fetch links out of the document and add them to the
* queue
*
* @param Spizer_Document_Html $doc
*/
public function handle(Spizer_Document $doc)
{
// If need, set the match domain according to the first URL
if (!isset($this->_config['domain']) && $this->_config['samedomain']) {
$this->_config['domain'] = $this->_engine->getBaseUri()->getHost();
}
// Add document URL to the list of visited pages
$baseUrl = (string) $doc->getUrl();
if (!in_array($baseUrl, $this->_targets)) {
$this->_targets[] = $baseUrl;
}
// Silently skip all non-HTML documents
if (!$doc instanceof Spizer_Document_Html) {
return;
}
// Fetch links out of the document
$links = array();
if ($this->_config['followhref']) {
$links = array_merge($links, $doc->getLinks());
}
if ($this->_config['followimg']) {
$links = array_merge($links, $doc->getImages());
}
if ($this->_config['followlink']) {
$links = array_merge($links, $doc->getHeaderLinks());
}
if ($this->_config['followscript']) {
$links = array_merge($links, $doc->getScriptLinks());
}
if ($this->_config['followframes']) {
$links = array_merge($links, $doc->getFrameLinks());
}
// Iterate over all document links
foreach ($links as $link) {
// Try to parse URL - if we fail, skip this link (should not happen normally)
if (!($parts = @parse_url($link))) {
continue;
}
// Skip non-http schemes
if (isset($parts['scheme']) && ($parts['scheme'] != 'http' && $parts['scheme'] != 'https')) {
continue;
}
// Full URI
if (isset($parts['host'])) {
if (preg_match('/' . preg_quote($this->_config['domain']) . '$/', $parts['host'])) {
$this->_addToQueue($link, $baseUrl);
}
// Partial URI
} elseif (isset($parts['path'])) {
try {
$linkUri = clone $doc->getUrl();
$linkUri->setQuery(isset($parts['query']) ? $parts['query'] : null);
$linkUri->getFragment(isset($parts['fragment']) ? $parts['fragment'] : null);
// Full absolute path
if (substr_compare($parts['path'], '/', 0, 1) == 0) {
$linkUri->setPath($parts['path']);
// Relative path
} else {
$basePath = $doc->getUrl()->getPath();
$pos = strrpos($basePath, '/');
if ($pos === false) {
$linkUri->setPath('/' . $parts['path']);
} else {
$linkUri->setPath(substr($basePath, 0, $pos + 1) . $parts['path']);
}
}
$this->_addToQueue($linkUri, $baseUrl);
// If any of the URL parts is invalid, an exception will be caught here
} catch (Zend_Uri_Exception $e) {
$this->_log(array('link' => $link, 'message' => 'Unable to parse link URL: ' . $e->getMessage()));
}
}
}
}
