}
}
-$es = new Elasticsearch($GLOBALS['phinde']['elasticsearch']);
-
$url = $argv[1];
-$existingDoc = $es->get($url);
-if ($existingDoc && $existingDoc->status == 'indexed') {
- echo "URL already indexed: $url\n";
- exit(0);
-}
-//FIXME: size limit
-//FIXME: sourcetitle, sourcelink
$req = new \HTTP_Request2($url);
$req->setConfig('follow_redirects', true);
$req->setConfig('connect_timeout', 5);
$req->setConfig('timeout', 10);
$req->setConfig('ssl_verify_peer', false);
+//FIXME: size limit
+
+$es = new Elasticsearch($GLOBALS['phinde']['elasticsearch']);
+$existingDoc = $es->get($url);
+if ($existingDoc && $existingDoc->status == 'indexed') {
+ $nMoDate = strtotime($existingDoc->modate);
+ $refreshtime = $GLOBALS['phinde']['refreshtime'];
+ if (time() - $nMoDate < $refreshtime) {
+ echo "URL already indexed less than $refreshtime seconds ago: $url\n";
+ exit(0);
+ }
+
+ $req->setHeader('If-Modified-Since: ' . date('r', $nMoDate));
+}
+//FIXME: sourcetitle, sourcelink
+
$res = $req->send();
//FIXME: try-catch
-//FIXME: delete if 401 gone or 404 when updating
-if ($res->getStatus() !== 200) {
+if ($res->getStatus() === 304) {
+ //not modified since last time
+ //FIXME: store "last try" time
+ exit(0);
+} else if ($res->getStatus() !== 200) {
+ //FIXME: delete if 401 gone or 404 when updating
echo "Response code is not 200 but " . $res->getStatus() . ", stopping\n";
//FIXME: update status
exit(3);
public function crawl($url)
{
- $res = $this->fetch($url);
+ $res = $this->fetch($url);
+ if ($res === false) {
+ return;
+ }
+
$linkInfos = $this->extractLinks($res);
if ($this->showLinksOnly) {
$this->showLinks($linkInfos);
protected function fetch($url)
{
+ $existingDoc = $this->es->get($url);
+
$req = new HttpRequest($url);
$req->setHeader(
'accept',
implode(',', array_keys(static::$supportedIndexTypes))
);
+ if ($existingDoc) {
+ $nMoDate = strtotime($existingDoc->modate);
+ $req->setHeader('If-Modified-Since: ' . date('r', $nMoDate));
+ }
+
$res = $req->send();
- if ($res->getStatus() !== 200) {
+ if ($res->getStatus() === 304) {
+ //not modified since last time, so don't crawl again
+ return false;
+ } else if ($res->getStatus() !== 200) {
throw new \Exception(
"Response code is not 200 but "
. $res->getStatus() . ", stopping"