1 files changed, 171 insertions, 1 deletions

diff --git a/src/Wallabag/CoreBundle/Helper/ContentProxy.php b/src/Wallabag/CoreBundle/Helper/ContentProxy.php
index 3fe31c2c..d38811a2 100644
--- a/src/Wallabag/CoreBundle/Helper/ContentProxy.php
+++ b/src/Wallabag/CoreBundle/Helper/ContentProxy.php
@@ -53,6 +53,7 @@ class ContentProxy
 
         if ((empty($content) || false === $this->validateContent($content)) && false === $disableContentUpdate) {
             $fetchedContent = $this->graby->fetchContent($url);
+            $fetchedContent['title'] = $this->sanitizeContentTitle($fetchedContent['title'], $fetchedContent['content_type']);
 
             // when content is imported, we have information in $content
             // in case fetching content goes bad, we'll keep the imported information instead of overriding them
@@ -65,6 +66,13 @@ class ContentProxy
         // so we'll be able to refetch it in the future
         $content['url'] = !empty($content['url']) ? $content['url'] : $url;
 
+        // 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)) {
+            $entry->setUrl($url);
+        }
+
         $this->stockEntry($entry, $content);
     }
 
@@ -177,6 +185,59 @@ class ContentProxy
     }
 
     /**
+     * Try to sanitize the title of the fetched content from wrong character encodings and invalid UTF-8 character.
+     *
+     * @param $title
+     * @param $contentType
+     *
+     * @return string
+     */
+    private function sanitizeContentTitle($title, $contentType)
+    {
+        if ('application/pdf' === $contentType) {
+            $title = $this->convertPdfEncodingToUTF8($title);
+        }
+
+        return $this->sanitizeUTF8Text($title);
+    }
+
+    /**
+     * If the title from the fetched content comes from a PDF, then its very possible that the character encoding is not
+     * UTF-8. This methods tries to identify the character encoding and translate the title to UTF-8.
+     *
+     * @param $title
+     *
+     * @return string (maybe contains invalid UTF-8 character)
+     */
+    private function convertPdfEncodingToUTF8($title)
+    {
+        // first try UTF-8 because its easier to detect its present/absence
+        foreach (['UTF-8', 'UTF-16BE', 'WINDOWS-1252'] as $encoding) {
+            if (mb_check_encoding($title, $encoding)) {
+                return mb_convert_encoding($title, 'UTF-8', $encoding);
+            }
+        }
+
+        return $title;
+    }
+
+    /**
+     * Remove invalid UTF-8 characters from the given string.
+     *
+     * @param string $rawText
+     *
+     * @return string
+     */
+    private function sanitizeUTF8Text($rawText)
+    {
+        if (mb_check_encoding($rawText, 'UTF-8')) {
+            return $rawText;
+        }
+
+        return iconv('UTF-8', 'UTF-8//IGNORE', $rawText);
+    }
+
+    /**
      * Stock entry with fetched or imported content.
      * Will fall back to OpenGraph data if available.
      *
@@ -185,7 +246,7 @@ class ContentProxy
      */
     private function stockEntry(Entry $entry, array $content)
     {
-        $entry->setUrl($content['url']);
+        $this->updateOriginUrl($entry, $content['url']);
 
         $this->setEntryDomainName($entry);
 
@@ -252,6 +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.
      *
      * @param array $content