support base href

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

diff --git a/src/phinde/LinkExtractor/Html.php b/src/phinde/LinkExtractor/Html.php
index 4acd19f513017b1eff58c1e49b61763c62fdf184..7b987e3280fa0b298c7c0f9b2ef774eb15cc5a87 100644 (file)
--- a/src/phinde/LinkExtractor/Html.php
+++ b/src/phinde/LinkExtractor/Html.php
@@ -2,6 +2,7 @@
 namespace phinde\LinkExtractor;
 
 use phinde\LinkInfo;
+use phinde\Helper;
 
 class Html
 {
@@ -21,6 +22,13 @@ class Html
 
         $dx = new \DOMXPath($doc);
 
+        $xbase = $dx->evaluate('/html/head/base[@href]')->item(0);
+        if ($xbase) {
+            $base = $base->resolve(
+                $xbase->attributes->getNamedItem('href')->textContent
+            );
+        }
+
         $meta = $dx->evaluate('/html/head/meta[@name="robots" and @content]')
             ->item(0);
         if ($meta) {
@@ -36,10 +44,10 @@ class Html
         $links = $dx->evaluate('//a');
         //FIXME: link rel, img, video
 
-        $alreadySeen = array();
+        $alreadySeen = array($url => true);
 
         foreach ($links as $link) {
-            $linkTitle = $link->textContent;
+            $linkTitle = Helper::sanitizeTitle($link->textContent);
             $href = '';
             foreach ($link->attributes as $attribute) {
                 if ($attribute->name == 'href') {