implement hosts crawl queue, move robots, sitemaps, manifests to this task

config/app.php.example

@@ -270,7 +270,7 @@ define('CRAWL_HOST_DEFAULT_NSFW', false);
 /*
  * Collect sitemap index when available
  *
- * At this moment, works with CRAWL_ROBOTS_SECONDS_OFFSET/CRAWL_ROBOTS_LIMIT options enabled only
+ * At this moment, works with CRAWL_HOST_SECONDS_OFFSET/CRAWL_HOST_LIMIT options enabled only
  *
  * When sitemap path not provided in robots.txt, crawler scans default /sitemap.xml
  *
@@ -290,18 +290,23 @@ define('CRAWL_SITEMAPS', true);
 define('CRAWL_PAGE_RANK_UPDATE', true);
 
 /*
- * Renew robots.txt index by timing offset provided
+ * Renew hosts index by timing offset provided
  *
  */
-define('CRAWL_ROBOTS_SECONDS_OFFSET', 60*60*24*7);
+define('CRAWL_HOST_SECONDS_OFFSET', 60*60*24*7);
 
 /*
- * Hosts Robots.txt processing limit in the crawler.php queue
+ * Hosts hosts processing limit in the crawler.php queue
  *
  * Set 0 to disable
  *
  */
-define('CRAWL_ROBOTS_LIMIT', 1);
+define('CRAWL_HOST_LIMIT', 1);
+
+/*
+ * Crawl robots.txt
+ */
+define('CRAWL_ROBOTS', true); // true|false
 
 /*
  * Default robots.txt rules on remote file not exists

crontab/crawler.php

@@ -44,11 +44,12 @@ $httpRequestsSizeTotal = 0;
 $httpDownloadSizeTotal = 0;
 $httpRequestsTimeTotal = 0;
 
+$hostsProcessed        = 0;
 $hostsAdded            = 0;
-$hostPagesBanned       = 0;
-$hostPagesSnapAdded    = 0;
 
 $hostPagesProcessed    = 0;
+$hostPagesBanned       = 0;
+$hostPagesSnapAdded    = 0;
 $hostPagesAdded        = 0;
 
 $manifestsProcessed    = 0;
@@ -67,8 +68,18 @@ try {
   exit;
 }
 
-// Process robots crawl queue
-foreach ($db->getHostRobotsCrawlQueue(CRAWL_ROBOTS_LIMIT, time() - CRAWL_ROBOTS_SECONDS_OFFSET) as $host) {
+// Process hosts crawl queue
+foreach ($db->getHostCrawlQueue(CRAWL_HOST_LIMIT, time() - CRAWL_HOST_SECONDS_OFFSET) as $host) {
+
+  $db->beginTransaction();
+
+  try {
+
+    // Update host crawl queue
+    $hostsProcessed += $db->updateHostCrawlQueue($host->hostId);
+
+    // Crawl robots.txt
+    if (CRAWL_ROBOTS) {
 
       // Update robots
       $curl = new Curl($host->url . '/robots.txt', CRAWL_CURLOPT_USERAGENT);
@@ -91,6 +102,7 @@ foreach ($db->getHostRobotsCrawlQueue(CRAWL_ROBOTS_LIMIT, time() - CRAWL_ROBOTS_
 
       // Update host index
       $db->updateHostRobots($host->hostId, $hostRobots, time());
+    }
 
     // Process sitemaps when enabled
     if (CRAWL_SITEMAPS) {
@@ -138,7 +150,8 @@ foreach ($db->getHostRobotsCrawlQueue(CRAWL_ROBOTS_LIMIT, time() - CRAWL_ROBOTS_
       }
     }
 
-  // Update manifest if available for this host
+    // Update manifests
+    if (CRAWL_MANIFEST) {
       if ($manifestURL = $db->getHostSetting($host->hostId, 'MANIFEST_URL')) {
 
         $curl = new Curl($manifestURL, CRAWL_CURLOPT_USERAGENT);
@@ -325,6 +338,21 @@ foreach ($db->getHostRobotsCrawlQueue(CRAWL_ROBOTS_LIMIT, time() - CRAWL_ROBOTS_
       }
     }
 
+    $db->commit();
+
+  // Process update errors
+  } catch (Exception $e) {
+
+    // Debug std
+    var_dump($e);
+
+    // Skip item
+    $db->rollBack();
+
+    continue;
+  }
+}
+
 // Process pages crawl queue
 foreach ($db->getHostPageCrawlQueue(CRAWL_PAGE_LIMIT, time() - CRAWL_PAGE_SECONDS_OFFSET, time() - CRAWL_PAGE_HOME_SECONDS_OFFSET) as $queueHostPage) {
 
@@ -1207,20 +1235,21 @@ $executionTimeTotal    = microtime(true) - $timeStart;
 $httpRequestsTimeTotal = $httpRequestsTimeTotal / 1000000;
 
 // Debug output
-echo 'Hosts added: ' . $hostsAdded . PHP_EOL;
+echo 'Hosts processed: ' . $hostsProcessed . PHP_EOL;
+echo 'Hosts added: ' . $hostsAdded . PHP_EOL . PHP_EOL;
 
 echo 'Pages processed: ' . $hostPagesProcessed . PHP_EOL;
 echo 'Pages added: ' . $hostPagesAdded . PHP_EOL;
 echo 'Pages snaps added: ' . $hostPagesSnapAdded . PHP_EOL;
-echo 'Pages banned: ' . $hostPagesBanned . PHP_EOL;
+echo 'Pages banned: ' . $hostPagesBanned . PHP_EOL . PHP_EOL;
 
-echo 'Sitemaps processed: ' . $sitemapsProcessed . PHP_EOL;
+echo 'Sitemaps processed: ' . $sitemapsProcessed . PHP_EOL . PHP_EOL;
 
-echo 'Manifests processed: ' . $manifestsProcessed . PHP_EOL;
+echo 'Manifests processed: ' . $manifestsProcessed . PHP_EOL . PHP_EOL;
 
 echo 'HTTP Requests total: ' . $httpRequestsTotal . PHP_EOL;
 echo 'HTTP Requests total size: ' . $httpRequestsSizeTotal . PHP_EOL;
 echo 'HTTP Download total size: ' . $httpDownloadSizeTotal . PHP_EOL;
-echo 'HTTP Requests total time: ' . $httpRequestsTimeTotal . PHP_EOL;
+echo 'HTTP Requests total time: ' . $httpRequestsTimeTotal . PHP_EOL . PHP_EOL;
 
 echo 'Total time: ' . $executionTimeTotal . PHP_EOL . PHP_EOL;

library/mysql.php

@@ -667,7 +667,7 @@ class MySQL {
     return $query->rowCount();
   }
 
-  public function getHostRobotsCrawlQueue(int $limit, int $timeFrom) {
+  public function getHostCrawlQueue(int $limit, int $timeFrom) {
 
     $result = [];
 
@@ -693,9 +693,19 @@ class MySQL {
     return (object) $result;
   }
 
+  public function updateHostCrawlQueue(int $hostId, int $timeUpdated) {
+
+    $query = $this->_db->prepare('UPDATE `host` SET `timeUpdated` = ? WHERE `hostId` = ? LIMIT 1');
+
+    $query->execute([$timeUpdated, $hostId]);
+
+    return $query->rowCount();
+  }
+
   public function optimize() {
 
     $this->_db->query('OPTIMIZE TABLE `host`');
+    $this->_db->query('OPTIMIZE TABLE `hostSetting`');
     $this->_db->query('OPTIMIZE TABLE `hostPage`');
     $this->_db->query('OPTIMIZE TABLE `hostPageDescription`');
     $this->_db->query('OPTIMIZE TABLE `hostPageDom`');