rework crawler; add atom link extraction

[phinde.git] / src / phinde / LinkExtractor / Atom.php

diff --git a/src/phinde/LinkExtractor/Atom.php b/src/phinde/LinkExtractor/Atom.php
new file mode 100644 (file)
index 0000000..bb4d90b
--- /dev/null
+++ b/src/phinde/LinkExtractor/Atom.php
@@ -0,0 +1,35 @@
+<?php
+namespace phinde\LinkExtractor;
+
+use phinde\LinkInfo;
+
+class Atom
+{
+    public function extract(\HTTP_Request2_Response $res)
+    {
+        $url  = $res->getEffectiveUrl();
+        $base = new \Net_URL2($url);
+
+        $sx = simplexml_load_string($res->getBody());
+        $linkInfos   = array();
+        $alreadySeen = array();
+
+        foreach ($sx->entry as $entry) {
+            $linkTitle = (string) $entry->title;
+            foreach ($entry->link as $xlink) {
+                $linkUrl = (string) $base->resolve((string) $xlink['href']);
+                if (isset($alreadySeen[$linkUrl])) {
+                    continue;
+                }
+
+                if ($xlink['rel'] == 'alternate') {
+                    $linkInfos[] = new LinkInfo($linkUrl, $linkTitle, $url);
+                }
+                $alreadySeen[$linkUrl] = true;
+            }
+        }
+
+        return $linkInfos;
+    }
+}
+?>