public function run()
{
if ($this->debugMode) {
echo "Restricting crawl to {$this->domain}\n";
}
//loop across available items in the queue of pages to crawl
while (!$this->queue->isEmpty()) {
if (isset($this->limit) && $this->counter >= $this->limit) {
break;
}
$this->counter++;
//get a new url to crawl
$url = $this->queue->pop();
if ($this->debugMode) {
echo "Queue Length: " . $this->queue->queueLength() . "\n";
echo "Crawling " . $url . "\n";
}
//set the url into the http client
$this->client->setUri($url);
//make the request to the remote server
$this->currentResponse = $this->client->request();
//don't bother trying to parse this if it's not text
if (stripos($this->currentResponse->getHeader('Content-type'), 'text') === false) {
continue;
}
//search for <a> tags in the document
$body = $this->currentResponse->getBody();
$linksQuery = new Zend_Dom_Query($body);
$links = $linksQuery->query('a');
if ($this->debugMode) {
echo "\tFound " . count($links) . " links...\n";
}
foreach ($links as $link) {
//get the href of the link and find out if it links to the current host
$href = $link->getAttribute('href');
$urlparts = parse_url($href);
if ($this->stayOnDomain && isset($urlparts["host"]) && $urlparts["host"] != $this->domain) {
continue;
}
//if it's an absolute link without a domain or a scheme, attempt to fix it
if (!isset($urlparts["host"])) {
$href = 'http://' . $this->domain . $href;
//this is a really naive way of doing this!
}
//push this link into the queue to be crawled
$this->queue->push($href);
}
//for each page that we see, run every registered task across it
foreach ($this->tasks as $task) {
$task->task($this->currentResponse, $this->client);
}
}
//after we're done with everything, call the shutdown hook on all the tasks
$this->shutdownTasks();
}
/**
* Sends a request and returns a response
*
* @param CartRecover_Request $request
* @return Cart_Recover_Response
*/
public function sendRequest(CartRecover_Request $request)
{
$this->client->setUri($request->getUri());
$this->client->setParameterGet($request->getParams());
$this->client->setMethod($request->getMethod());
$this->client->setHeaders('Accept', 'application/json');
$this->response = $this->client->request();
if ($this->response->getHeader('Content-Type') != 'application/json') {
throw new CartRecover_Exception_UnexpectedValueException("Unknown response format.");
}
$body = json_decode($this->response->getBody(), true);
$response = new CartRecover_Response();
$response->setRawResponse($this->response->asString());
$response->setBody($body);
$response->setHeaders($this->response->getHeaders());
$response->setStatus($this->response->getMessage(), $this->response->getStatus());
return $response;
}
protected function createResponse(\Zend_Http_Response $response)
{
$headers = array($response->getHeader('Set-Cookie'));
$cookies = array();
foreach ($headers as $header) {
if (!trim($header)) {
continue;
}
$parts = explode(';', $header);
$value = array_shift($parts);
list($name, $value) = explode('=', trim($value));
$cookies[$name] = array('value' => $value);
foreach ($parts as $part) {
list($key, $value) = explode('=', trim($part));
$cookies[$name][$key] = $value;
}
}
return new Response($response->getBody(), $response->getStatus(), $response->getHeaders(), $cookies);
}
/**
* Test that headers are properly parsed when passed to the constructor as
* an indexed array with no whitespace after the ':' sign
*
* @link http://framework.zend.com/issues/browse/ZF-10277
* @group ZF-10277
*/
public function testConstructorWithHeadersIndexedArrayNoWhitespace()
{
$response = new Zend_Http_Response(200, array('content-type:text/plain', 'x-foo:bar:baz'));
$this->assertEquals('text/plain', $response->getHeader('content-type'));
$this->assertEquals('bar:baz', $response->getHeader('x-foo'));
}
protected function _previewHtml($uri, Zend_Http_Response $response)
{
$body = $response->getBody();
$body = trim($body);
if (preg_match('/charset=([a-zA-Z0-9-_]+)/i', $response->getHeader('content-type'), $matches) || preg_match('/charset=([a-zA-Z0-9-_]+)/i', $response->getBody(), $matches)) {
$this->view->charset = $charset = trim($matches[1]);
} else {
$this->view->charset = $charset = 'UTF-8';
}
// if( function_exists('mb_convert_encoding') ) {
// $body = mb_convert_encoding($body, 'HTML-ENTITIES', $charset);
// }
// Get DOM
if (class_exists('DOMDocument')) {
$dom = new Zend_Dom_Query($body);
} else {
$dom = null;
// Maybe add b/c later
}
$title = null;
if ($dom) {
$titleList = $dom->query('title');
if (count($titleList) > 0) {
$title = trim($titleList->current()->textContent);
$title = substr($title, 0, 255);
}
}
$this->view->title = $title;
$description = null;
if ($dom) {
$descriptionList = $dom->queryXpath("//meta[@name='description']");
// Why are they using caps? -_-
if (count($descriptionList) == 0) {
$descriptionList = $dom->queryXpath("//meta[@name='Description']");
}
if (count($descriptionList) > 0) {
$description = trim($descriptionList->current()->getAttribute('content'));
$description = substr($description, 0, 255);
}
}
$this->view->description = $description;
$thumb = null;
if ($dom) {
$thumbList = $dom->queryXpath("//link[@rel='image_src']");
if (count($thumbList) > 0) {
$thumb = $thumbList->current()->getAttribute('href');
}
}
$this->view->thumb = $thumb;
$medium = null;
if ($dom) {
$mediumList = $dom->queryXpath("//meta[@name='medium']");
if (count($mediumList) > 0) {
$medium = $mediumList->current()->getAttribute('content');
}
}
$this->view->medium = $medium;
// Get baseUrl and baseHref to parse . paths
$baseUrlInfo = parse_url($uri);
$baseUrl = null;
$baseHostUrl = null;
if ($dom) {
$baseUrlList = $dom->query('base');
if ($baseUrlList && count($baseUrlList) > 0 && $baseUrlList->current()->getAttribute('href')) {
$baseUrl = $baseUrlList->current()->getAttribute('href');
$baseUrlInfo = parse_url($baseUrl);
$baseHostUrl = $baseUrlInfo['scheme'] . '://' . $baseUrlInfo['host'] . '/';
}
}
if (!$baseUrl) {
$baseHostUrl = $baseUrlInfo['scheme'] . '://' . $baseUrlInfo['host'] . '/';
if (empty($baseUrlInfo['path'])) {
$baseUrl = $baseHostUrl;
} else {
$baseUrl = explode('/', $baseUrlInfo['path']);
array_pop($baseUrl);
$baseUrl = join('/', $baseUrl);
$baseUrl = trim($baseUrl, '/');
$baseUrl = $baseUrlInfo['scheme'] . '://' . $baseUrlInfo['host'] . '/' . $baseUrl . '/';
}
}
$images = array();
if ($thumb) {
$images[] = $thumb;
}
if ($dom) {
$imageQuery = $dom->query('img');
foreach ($imageQuery as $image) {
$src = $image->getAttribute('src');
// Ignore images that don't have a src
if (!$src || false === ($srcInfo = @parse_url($src))) {
continue;
}
$ext = ltrim(strrchr($src, '.'), '.');
// Detect absolute url
if (strpos($src, '/') === 0) {
// If relative to root, add host
$src = $baseHostUrl . ltrim($src, '/');
} else {
if (strpos($src, './') === 0) {
// If relative to current path, add baseUrl
$src = $baseUrl . substr($src, 2);
} else {
if (!empty($srcInfo['scheme']) && !empty($srcInfo['host'])) {
// Contians host and scheme, do nothing
} else {
if (empty($srcInfo['scheme']) && empty($srcInfo['host'])) {
// if not contains scheme or host, add base
$src = $baseUrl . ltrim($src, '/');
} else {
if (empty($srcInfo['scheme']) && !empty($srcInfo['host'])) {
// if contains host, but not scheme, add scheme?
$src = $baseUrlInfo['scheme'] . ltrim($src, '/');
} else {
// Just add base
$src = $baseUrl . ltrim($src, '/');
}
}
}
}
}
// Ignore images that don't come from the same domain
//if( strpos($src, $srcInfo['host']) === false ) {
// @todo should we do this? disabled for now
//continue;
//}
// Ignore images that don't end in an image extension
if (!in_array($ext, array('jpg', 'jpeg', 'gif', 'png'))) {
// @todo should we do this? disabled for now
//continue;
}
if (!in_array($src, $images)) {
$images[] = $src;
}
}
}
// Unique
$images = array_values(array_unique($images));
// Truncate if greater than 20
if (count($images) > 30) {
array_splice($images, 30, count($images));
}
$this->view->imageCount = count($images);
$this->view->images = $images;
}
/**
* Try to find the document's language by first looking for Content-Language in Http headers than in html
* attribute and last in content-language meta tag
* @param Zend_Http_Response $response
* @return string
*/
protected function getLanguageFromResponse($response)
{
$l = $response->getHeader("Content-Language");
if (empty($l)) {
//try html lang attribute
$languages = array();
preg_match_all('@<html[\\n|\\r\\n]*.*?[\\n|\\r\\n]*lang="(?P<language>\\S+)"[\\n|\\r\\n]*.*?[\\n|\\r\\n]*>@si', $response->getBody(), $languages);
if ($languages['language']) {
$l = str_replace(array("_", "-"), "", $languages['language'][0]);
}
}
if (empty($l)) {
//try meta tag
$languages = array();
preg_match_all('@<meta\\shttp-equiv="content-language"\\scontent="(?P<language>\\S+)"\\s\\/>@si', $response->getBody(), $languages);
if ($languages['language']) {
//for lucene index remove "_" - this causes tokenization
$l = str_replace("_", "", $languages['language'][0]);
}
}
return $l;
}
/**
* @throws OpenSocial_Rest_Exception
* @param $serviceType
* @param Zend_Http_Response $response
* @return array
*/
protected function _mapResponseToModels($serviceType, Zend_Http_Response $response)
{
if (strpos($response->getHeader('Content-Type'), 'application/json') !== 0) {
throw new OpenSocial_Rest_Exception("Unknown Content-Type for response:<br /> " . var_export($response, true) . ' with body: ' . var_export($response->getBody(), true) . ' for request: ' . $this->_httpClient->getLastRequest());
}
$modelClass = 'OpenSocial_Model_' . $this->_getModelTypeForService($serviceType);
if (!class_exists($modelClass, true)) {
throw new OpenSocial_Rest_Exception("Model class {$modelClass} not found for service {$serviceType}!");
}
/**
* @var OpenSocial_Rest_Mapper_Interface $mapper
*/
$mapper = new OpenSocial_Rest_Mapper_Json($modelClass);
return $mapper->map($response->getBody());
}
/**
* Parse an HTTP response, adding all the cookies set in that response
* to the cookie jar.
*
* @param Zend_Http_Response $response
* @param Zend_Uri_Http|string $ref_uri Requested URI
*/
public function addCookiesFromResponse($response, $ref_uri)
{
if (!$response instanceof Sabel_Http_Response) {
$message = __METHOD__ . "() \$response is expected to be a Response object.";
throw new Sabel_Exception_Runtime($message);
}
$cookie_hdrs = $response->getHeader("Set-Cookie");
if (is_array($cookie_hdrs)) {
foreach ($cookie_hdrs as $cookie) {
$this->addCookie($cookie, $ref_uri);
}
} elseif (is_string($cookie_hdrs)) {
$this->addCookie($cookie_hdrs, $ref_uri);
}
}
/**
* Parse an HTTP response, adding all the cookies set in that response
* to the cookie jar.
*
* @param Zend_Http_Response $response
* @param Zend_Uri_Http|string $ref_uri Requested URI
*/
public function addCookiesFromResponse($response, $ref_uri)
{
$cookie_hdrs = $response->getHeader('Set-Cookie');
if (is_array($cookie_hdrs)) {
foreach ($cookie_hdrs as $cookie) {
$this->addCookie($cookie, $ref_uri);
}
} elseif (is_string($cookie_hdrs)) {
$this->addCookie($cookie_hdrs, $ref_uri);
}
}
protected function _previewHtml($uri, Zend_Http_Response $response)
{
$arr_return = array();
$body = $response->getBody();
$body = trim($body);
if (preg_match('/charset=([a-zA-Z0-9-_]+)/i', $response->getHeader('content-type'), $matches) || preg_match('/charset=([a-zA-Z0-9-_]+)/i', $response->getBody(), $matches)) {
$charset = trim($matches[1]);
} else {
$charset = 'UTF-8';
}
if (function_exists('mb_convert_encoding')) {
$body = mb_convert_encoding($body, 'HTML-ENTITIES', $charset);
}
// Get DOM
if (class_exists('DOMDocument')) {
$dom = new Zend_Dom_Query($body);
} else {
$dom = null;
// Maybe add b/c later
}
$title = null;
if ($dom) {
$titleList = $dom->query('title');
if (count($titleList) > 0) {
$title = trim($titleList->current()->textContent);
$title = substr($title, 0, 255);
}
}
$arr_return['title'] = $title;
$description = null;
if ($dom) {
$descriptionList = $dom->queryXpath("//meta[@name='description']");
// Why are they using caps? -_-
if (count($descriptionList) == 0) {
$descriptionList = $dom->queryXpath("//meta[@name='Description']");
}
if (count($descriptionList) > 0) {
$description = trim($descriptionList->current()->getAttribute('content'));
$description = substr($description, 0, 255);
}
}
$arr_return['description'] = $description;
$thumb = null;
if ($dom) {
$mediumList = $dom->queryXpath("//meta[@property='og:image']");
if (count($mediumList) > 0) {
$thumb = $mediumList->current()->getAttribute('content');
}
if (!$thumb) {
$thumbList = $dom->queryXpath("//link[@rel='image_src']");
if (count($thumbList) > 0) {
$thumb = $thumbList->current()->getAttribute('href');
}
}
}
$arr_return['thumb'] = $thumb;
return $arr_return;
}
