/**
* Check whether a URL is available, and update the status of the URL in the database
* @param $last_check array of the data from the last check for the URL
* @param bool $force true if the check should be forced to happen, even if it's not yet scheduled
* @return array|bool
*/
public function check($last_check, $force = false)
{
$url = $last_check['url'];
$id = isset($last_check['id']) ? $last_check['id'] : md5($url);
//TODO: Unify ID generation
/* Make sure we're still scheduled to check the $url */
$next_check_timestamp = isset($last_check['next_check']) ? $last_check['next_check'] : 0;
if (!$force && $next_check_timestamp > time()) {
return false;
}
$date = new DateTime();
if (!AmberRobots::robots_allowed($url)) {
/* If blocked by robots.txt, schedule next check for 6 months out */
$next = $date->add(new DateInterval("P6M"))->getTimestamp();
$status = isset($last_check['status']) ? $last_check['status'] : NULL;
error_log(join(":", array(__FILE__, __METHOD__, "Blocked by robots.txt", $url)));
$message = "Blocked by robots.txt";
} else {
$fetch_result = AmberNetworkUtils::open_url($url, array(CURLOPT_FAILONERROR => FALSE));
$status = $this->is_up($fetch_result);
$next = $this->next_check_date(isset($last_check['status']) ? $last_check['status'] : NULL, isset($last_check['last_checked']) ? $last_check['last_checked'] : NULL, isset($last_check['next_check']) ? $last_check['next_check'] : NULL, $status);
}
$now = new DateTime();
$result = array('id' => $id, 'url' => $url, 'last_checked' => $now->getTimestamp(), 'next_check' => $next, 'status' => isset($status) ? $status ? 1 : 0 : NULL, 'message' => isset($message) ? $message : NULL, 'details' => isset($fetch_result) ? $fetch_result : NULL);
return $result;
}
/**
* Fetch the URL and associated assets and pass it on to the designated Storage service
* @param $url
* @return
*/
public function fetch($url)
{
if (!$url) {
throw new RuntimeException("Empty URL");
}
// Check the robots.txt
if (!AmberRobots::robots_allowed($url)) {
throw new RuntimeException("Blocked by robots.txt");
}
// Send a GET request
$root_item = AmberNetworkUtils::open_url($url);
// Decide whether the item should be cached
if (!$this->cacheable_item($root_item, $reason)) {
throw new RuntimeException($reason);
}
$size = $root_item['info']['size_download'];
if ($size == 0) {
throw new RuntimeException("Empty document");
}
// Get other assets
if (isset($root_item['headers']['Content-Type']) && ($content_type = $root_item['headers']['Content-Type']) && AmberNetworkUtils::is_html_mime_type($content_type)) {
$body = $root_item['body'];
$asset_paths = $this->assetHelper->extract_assets($body);
/* Use the url of the document we end up downloading as a reference point for
relative asset references, since we may have been redirected from the one
we originally requested. */
$assets = $this->assetHelper->expand_asset_references($root_item['info']['url'], $asset_paths, $this->assetHelper->extract_base_tag($body));
$assets = $this->download_assets($assets, $root_item['info']['url']);
$assets = $this->download_css_assets_recursive($assets, $root_item['info']['url'], $size);
$body = $this->assetHelper->rewrite_links($body, $assets);
$body = $this->assetHelper->insert_banner($body, $this->headerText, array("url" => $url, "date" => date('Y/m/d')));
$root_item['body'] = $body;
/* Check total size of the file combined with its assets */
if ($size > $this->maxFileSize * 1024) {
throw new RuntimeException("File size of document + assets too large");
}
}
if ($this->storage && $root_item) {
$result = $this->storage->save($url, $root_item['body'], $root_item['headers'], isset($assets) ? $assets : array());
if (!$result) {
throw new RuntimeException("Could not save cache");
}
$storage_metadata = $this->storage->get_metadata($url);
if (!$storage_metadata || empty($storage_metadata)) {
throw new RuntimeException("Could not retrieve metadata");
}
//TODO: If cannot retrieve storage metadata, or id/url/cache not populated (perhaps due to permissions errors
// in saving the cache), fail more gracefully instead of with errors because the keys are not set
return array('id' => $storage_metadata['id'], 'url' => $storage_metadata['url'], 'type' => isset($storage_metadata['type']) ? $storage_metadata['type'] : 'application/octet-stream', 'date' => strtotime($storage_metadata['cache']['amber']['date']), 'location' => $storage_metadata['cache']['amber']['location'], 'size' => $size, 'provider' => $this->storage->provider_id(), 'provider_id' => $storage_metadata['id']);
} else {
throw new RuntimeException("Content empty or could not save to disk");
}
}
