/**
* Send a request for scraping a single url (product) to keyword servers
* @param object $product product
* @param string $server keyword server
* @author Stom
*/
protected function create_product_crawl_request($product, $server)
{
$result = false;
$product_url_host = preg_replace("/^www\\./", '', parse_url(strtolower($product->url), PHP_URL_HOST));
if (empty($product->id) || empty($product_url_host)) {
return $result;
}
$crawler_name = Ranking_model::getCrawlerName($product_url_host);
$server_url = "http://{$server}:6543/ranking_data/";
$post_array = array('site' => $crawler_name, 'product_url' => $product->url);
$ch = curl_init();
curl_setopt($ch, CURLOPT_URL, $server_url);
curl_setopt($ch, CURLOPT_RETURNTRANSFER, true);
curl_setopt($ch, CURLOPT_POST, true);
curl_setopt($ch, CURLOPT_POSTFIELDS, $post_array);
curl_setopt($ch, CURLOPT_FOLLOWLOCATION, true);
curl_setopt($ch, CURLOPT_MAXREDIRS, 1);
$curl_result = curl_exec($ch);
$curl_log_data = array('type' => 'keyword_cron_job', 'method' => 'POST', 'url' => $server_url, 'params' => json_encode($post_array), 'http_code' => curl_getinfo($ch, CURLINFO_HTTP_CODE), 'curl_info' => json_encode(curl_getinfo($ch)), 'result' => $curl_result);
$curl_log_id = $this->ranking_model->create_('curl_logs', $curl_log_data);
$http_code = curl_getinfo($ch, CURLINFO_HTTP_CODE);
if (preg_match('/[23]\\d\\d/', $http_code)) {
// 2xx 3xx codes (usually 202 or 302)
// create job
$eff_url = curl_getinfo($ch, CURLINFO_EFFECTIVE_URL);
$eff_url_parts = explode('/', $eff_url);
$request_id = $eff_url_parts[sizeof($eff_url_parts) - 2];
$job_url = curl_getinfo($ch, CURLINFO_REDIRECT_URL) ?: $eff_url;
// create cron job
$keyword_cron_job_data = array('url' => $job_url, 'request_id' => $request_id, 'status' => $this->ranking_model->CRON_STATUS_WAITING, 'product_list_item_id' => $product->id);
$keyword_cron_job_id = $this->ranking_model->create_('keyword_cron_jobs', $keyword_cron_job_data);
$keyword_cron_jobs_log_data = array('job_id' => $keyword_cron_job_id, 'message' => json_encode(array('id' => $request_id, 'crawler_name' => $crawler_name, 'product_url' => $product->url, 'status' => "STATUS REQUESTED ({$http_code})")), 'child_job_id' => $request_id, 'job_phase' => $this->ranking_model->JOB_PHASE_REQUEST, 'job_phase_status' => 'true', 'curl_log_id' => $curl_log_id);
$this->ranking_model->create_('keyword_cron_jobs_log', $keyword_cron_jobs_log_data);
$result = true;
}
curl_close($ch);
return $result;
}
public function create_jobs()
{
if (!$this->_check_setting()) {
echo 'ranking data jobs disabled';
return;
}
$this->load->model('ranking_model');
$this->load->library('logger');
$start_time = microtime(true);
$log_message = array('number_of_messages_queued' => 0, 'time' => 0);
$log_id = $this->logger->info('ranking_data_create_jobs', json_encode($log_message));
// create 'search term' jobs
$search_terms_to_crawl = $this->ranking_data_model->getSearchTermsToBeCrawled();
// create jobs
$ranking_data_jobs_batch = array();
foreach ($search_terms_to_crawl as $search_term_to_crawl) {
if (!empty($search_term_to_crawl->search_terms_group_id) && !empty($search_term_to_crawl->search_term) && !empty($search_term_to_crawl->site_id)) {
$ranking_data_jobs_batch[] = array('search_term_group_id' => $search_term_to_crawl->search_terms_group_id, 'search_term' => $search_term_to_crawl->search_term, 'site_id' => $search_term_to_crawl->site_id);
}
}
if (!empty($ranking_data_jobs_batch)) {
$this->ranking_data_model->create_batch_('ranking_data_jobs', $ranking_data_jobs_batch);
}
// create 'product' jobs
$products_to_crawl = $this->ranking_data_model->getProductsToBeCrawled();
$ranking_data_jobs_batch = array();
foreach ($products_to_crawl as $product) {
$ranking_data_jobs_batch[] = array('product_list_item_id' => $product->id);
}
if (!empty($ranking_data_jobs_batch)) {
$create_batch_result = $this->ranking_data_model->create_batch_('ranking_data_jobs', $ranking_data_jobs_batch);
}
// re-create failed jobs
$failed_jobs_res = $this->ranking_data_model->getFailedJobs();
$failed_jobs = array();
foreach ($failed_jobs_res as $failed_job_res) {
$failed_job_key = implode("|", array_intersect_key((array) $failed_job_res, array_flip(array('search_term_group_id', 'search_term', 'site_id', 'product_list_item_id'))));
$failed_jobs[$failed_job_key] = $failed_job_res;
// grouped by "search_term_group_id|search_term|site_id|product_list_item_id"
}
if (!empty($failed_jobs)) {
$ranking_data_jobs_batch = array();
foreach ($failed_jobs as $failed_job) {
$failed_jobs_same_params = array('search_term_group_id' => $failed_job->search_term_group_id, 'search_term' => $failed_job->search_term, 'site_id' => $failed_job->site_id, 'product_list_item_id' => $failed_job->product_list_item_id);
$failed_jobs_same = $this->ranking_data_model->getFailedJobSame($failed_jobs_same_params);
// check same jobs count and status
$max_status = Ranking_data_model::STATUS_CREATED;
$min_status = Ranking_data_model::STATUS_RESULTS_IMPORTED;
$age_1_hour = null;
foreach ($failed_jobs_same as $failed_job_same) {
$max_status = max($max_status, $failed_job_same->status);
$min_status = min($min_status, $failed_job_same->status);
$age_1_hour = (is_null($age_1_hour) ? true : $age_1_hour) && ($failed_job_same->age_1_hour === 't' ? true : false);
}
if ($min_status > Ranking_data_model::STATUS_CREATED && $max_status < Ranking_data_model::STATUS_DONE && count($failed_jobs_same) < self::FAILED_JOB_ATTEMPTS && $age_1_hour) {
$ranking_data_jobs_batch[] = $failed_jobs_same_params;
}
}
if (!empty($ranking_data_jobs_batch)) {
$this->ranking_data_model->create_batch_('ranking_data_jobs', $ranking_data_jobs_batch);
}
}
// queue jobs
$jobs_to_queue = $this->ranking_data_model->getJobsToQueue(self::JOBS_TO_QUEUE);
// get jobs' ids
$jobs_ids = array_values(array_unique(array_filter(array_map(function ($job) {
return $job->id;
}, $jobs_to_queue)), SORT_NUMERIC));
// get corresponding sites
$sites_ids = array_values(array_unique(array_filter(array_map(function ($job) {
return $job->site_id;
}, $jobs_to_queue)), SORT_NUMERIC));
$sites = array();
if (!empty($sites_ids)) {
foreach ($this->ranking_model->get_('sites', array('id' => $sites_ids)) as $site) {
$sites[$site->id] = $site;
}
}
// get corresponding product list items
$product_list_item_ids = array_values(array_unique(array_filter(array_map(function ($job) {
return $job->product_list_item_id;
}, $jobs_to_queue)), SORT_NUMERIC));
$product_list_items = array();
if (!empty($product_list_item_ids)) {
foreach ($this->ranking_data_model->get_('product_list_items', array('id' => $product_list_item_ids)) as $product_list_item) {
$product_list_items[$product_list_item->id] = $product_list_item;
}
}
$entries = array();
foreach ($jobs_to_queue as $job_to_queue) {
$message = array('server_name' => $this->_get_server_name(), 'task_id' => $job_to_queue->id);
if (!empty($job_to_queue->search_term) && !empty($job_to_queue->site_id) && !empty($sites[$job_to_queue->site_id])) {
$job_site = $sites[$job_to_queue->site_id];
$message['searchterms_str'] = $job_to_queue->search_term;
$message['site'] = Ranking_model::getCrawlerName($job_site->name);
$message['cmd_args'] = array('quantity' => 1000);
} elseif (!empty($job_to_queue->product_list_item_id) && !empty($product_list_items[$job_to_queue->product_list_item_id])) {
$job_product = $product_list_items[$job_to_queue->product_list_item_id];
$product_url_host = preg_replace("/^www\\./", '', parse_url(strtolower($job_product->url), PHP_URL_HOST));
$message['site'] = Ranking_model::getCrawlerName($product_url_host);
$message['url'] = $job_product->url;
}
if (!(empty($message['searchterms_str']) && empty($message['url']))) {
$entries[] = array('Id' => $job_to_queue->id, 'MessageBody' => json_encode($message));
}
}
if (!empty($entries)) {
foreach (array_chunk($entries, self::SQS_MAX_MESSAGES) as $entries_sub) {
$sqs_call_args = array('QueueUrl' => $this->_get_queue_url(), 'Entries' => $entries_sub);
$aws_log_id = null;
$sqs_result = $this->aws->sqs->call('sendMessageBatch', $sqs_call_args, $this->_get_log_callback(), $aws_log_id);
if (!empty($sqs_result['Successful'])) {
foreach ($sqs_result['Successful'] as $job_queue_result) {
if (!empty($job_queue_result['Id']) && in_array($job_queue_result['Id'], $jobs_ids)) {
// change ranking_data_jobs status
$update_data = array('status' => Ranking_data_model::STATUS_QUEUED, 'updated_at' => 'now()');
$where_data = array('id' => $job_queue_result['Id']);
$this->ranking_data_model->update_('ranking_data_jobs', $update_data, $where_data);
// create ranking_data_job_logs
$ranking_data_job_logs_data = array_filter(array('ranking_data_job_id' => $job_queue_result['Id'], 'phase' => Ranking_data_model::STATUS_QUEUED, 'aws_log_id' => $aws_log_id));
$this->ranking_data_model->create_('ranking_data_job_logs', $ranking_data_job_logs_data);
$log_message['number_of_messages_queued']++;
}
}
} else {
foreach ($jobs_ids as $jobs_id) {
// change ranking_data_jobs status
$update_data = array('updated_at' => 'now()');
$where_data = array('id' => $jobs_id);
$this->ranking_data_model->update_('ranking_data_jobs', $update_data, $where_data);
// create ranking_data_job_logs
$ranking_data_job_logs_data = array_filter(array('ranking_data_job_id' => $jobs_id, 'phase' => Ranking_data_model::STATUS_QUEUED, 'success' => 'false', 'aws_log_id' => $aws_log_id));
$this->ranking_data_model->create_('ranking_data_job_logs', $ranking_data_job_logs_data);
}
}
}
}
// update log
$log_message['time'] = number_format(microtime(true) - $start_time, 2);
$this->logger->update($log_id, array('message' => json_encode($log_message)));
echo 'ok';
}
