Hash the urls to check if they exist

[github/wallabag/wallabag.git] / src / Wallabag / CoreBundle / Helper / ContentProxy.php

diff --git a/src/Wallabag/CoreBundle/Helper/ContentProxy.php b/src/Wallabag/CoreBundle/Helper/ContentProxy.php
index f0d8c1b420c9ba5fa192eb43b4949d84d1f79b5a..0534d27b22ca58fe48c88a3d34a8835bac864bfe 100644 (file)
--- a/src/Wallabag/CoreBundle/Helper/ContentProxy.php
+++ b/src/Wallabag/CoreBundle/Helper/ContentProxy.php
@@ -68,9 +68,8 @@ class ContentProxy
 
         // In one case (at least in tests), url is empty here
         // so we set it using $url provided in the updateEntry call.
-        // Not sure what are the other possible cases where this property is empty 
-        if (empty($entry->getUrl()) && !empty($url))
-        {
+        // Not sure what are the other possible cases where this property is empty
+        if (empty($entry->getUrl()) && !empty($url)) {
             $entry->setUrl($url);
         }
 
@@ -247,7 +246,9 @@ class ContentProxy
      */
     private function stockEntry(Entry $entry, array $content)
     {
-        $entry->setUrl($content['url']);
+        $this->updateOriginUrl($entry, $content['url']);
+
+        $entry->setHashedUrl(hash('md5', $entry->getUrl()));
 
         $this->setEntryDomainName($entry);
 
@@ -257,18 +258,17 @@ class ContentProxy
             $entry->setTitle($content['open_graph']['og_title']);
         }
 
-        $html = $content['html'];
-        if (false === $html) {
-            $html = $this->fetchingErrorMessage;
+        if (empty($content['html'])) {
+            $content['html'] = $this->fetchingErrorMessage;
 
             if (!empty($content['open_graph']['og_description'])) {
-                $html .= '<p><i>But we found a short description: </i></p>';
-                $html .= $content['open_graph']['og_description'];
+                $content['html'] .= '<p><i>But we found a short description: </i></p>';
+                $content['html'] .= $content['open_graph']['og_description'];
             }
         }
 
-        $entry->setContent($html);
-        $entry->setReadingTime(Utils::getReadingTime($html));
+        $entry->setContent($content['html']);
+        $entry->setReadingTime(Utils::getReadingTime($content['html']));
 
         if (!empty($content['status'])) {
             $entry->setHttpStatus($content['status']);
@@ -313,6 +313,115 @@ class ContentProxy
         }
     }
 
+    /**
+     * Update the origin_url field when a redirection occurs
+     * This field is set if it is empty and new url does not match ignore list.
+     *
+     * @param Entry  $entry
+     * @param string $url
+     */
+    private function updateOriginUrl(Entry $entry, $url)
+    {
+        if (empty($url) || $entry->getUrl() === $url) {
+            return false;
+        }
+
+        $parsed_entry_url = parse_url($entry->getUrl());
+        $parsed_content_url = parse_url($url);
+
+        /**
+         * The following part computes the list of part changes between two
+         * parse_url arrays.
+         *
+         * As array_diff_assoc only computes changes to go from the left array
+         * to the right one, we make two differents arrays to have both
+         * directions. We merge these two arrays and sort keys before passing
+         * the result to the switch.
+         *
+         * The resulting array gives us all changing parts between the two
+         * urls: scheme, host, path, query and/or fragment.
+         */
+        $diff_ec = array_diff_assoc($parsed_entry_url, $parsed_content_url);
+        $diff_ce = array_diff_assoc($parsed_content_url, $parsed_entry_url);
+
+        $diff = array_merge($diff_ec, $diff_ce);
+        $diff_keys = array_keys($diff);
+        sort($diff_keys);
+
+        if ($this->ignoreUrl($entry->getUrl())) {
+            $entry->setUrl($url);
+
+            return false;
+        }
+
+        /**
+         * This switch case lets us apply different behaviors according to
+         * changing parts of urls.
+         *
+         * As $diff_keys is an array, we provide arrays as cases. ['path'] means
+         * 'only the path is different between the two urls' whereas
+         * ['fragment', 'query'] means 'only fragment and query string parts are
+         * different between the two urls'.
+         *
+         * Note that values in $diff_keys are sorted.
+         */
+        switch ($diff_keys) {
+            case ['path']:
+                if (($parsed_entry_url['path'] . '/' === $parsed_content_url['path']) // diff is trailing slash, we only replace the url of the entry
+                    || ($url === urldecode($entry->getUrl()))) { // we update entry url if new url is a decoded version of it, see EntryRepository#findByUrlAndUserId
+                    $entry->setUrl($url);
+                }
+                break;
+            case ['scheme']:
+                $entry->setUrl($url);
+                break;
+            case ['fragment']:
+                // noop
+                break;
+            default:
+                if (empty($entry->getOriginUrl())) {
+                    $entry->setOriginUrl($entry->getUrl());
+                }
+                $entry->setUrl($url);
+                break;
+        }
+    }
+
+    /**
+     * Check entry url against an ignore list to replace with content url.
+     *
+     * XXX: move the ignore list in the database to let users handle it
+     *
+     * @param string $url url to test
+     *
+     * @return bool true if url matches ignore list otherwise false
+     */
+    private function ignoreUrl($url)
+    {
+        $ignored_hosts = ['feedproxy.google.com', 'feeds.reuters.com'];
+        $ignored_patterns = ['https?://www\.lemonde\.fr/tiny.*'];
+
+        $parsed_url = parse_url($url);
+
+        $filtered = array_filter($ignored_hosts, function ($var) use ($parsed_url) {
+            return $var === $parsed_url['host'];
+        });
+
+        if ([] !== $filtered) {
+            return true;
+        }
+
+        $filtered = array_filter($ignored_patterns, function ($var) use ($url) {
+            return preg_match("`$var`i", $url);
+        });
+
+        if ([] !== $filtered) {
+            return true;
+        }
+
+        return false;
+    }
+
     /**
      * Validate that the given content has at least a title, an html and a url.
      *