X-Git-Url: https://git.immae.eu/?a=blobdiff_plain;f=src%2FWallabag%2FCoreBundle%2FHelper%2FContentProxy.php;h=d4ea608f3c724d6ae9f896199fb23ede3fc96435;hb=83f1c3274f02ece12bd06c0b8df61d5e4b3236e7;hp=1ac7ad83c8ab753231ea72ec5c05db1205ae9aaf;hpb=e56f82142ac890c916b74700087424017bb2acf9;p=github%2Fwallabag%2Fwallabag.git

diff --git a/src/Wallabag/CoreBundle/Helper/ContentProxy.php b/src/Wallabag/CoreBundle/Helper/ContentProxy.php
index 1ac7ad83..d4ea608f 100644
--- a/src/Wallabag/CoreBundle/Helper/ContentProxy.php
+++ b/src/Wallabag/CoreBundle/Helper/ContentProxy.php
@@ -24,8 +24,9 @@ class ContentProxy
     protected $mimeGuesser;
     protected $fetchingErrorMessage;
     protected $eventDispatcher;
+    protected $storeArticleHeaders;
 
-    public function __construct(Graby $graby, RuleBasedTagger $tagger, ValidatorInterface $validator, LoggerInterface $logger, $fetchingErrorMessage)
+    public function __construct(Graby $graby, RuleBasedTagger $tagger, ValidatorInterface $validator, LoggerInterface $logger, $fetchingErrorMessage, $storeArticleHeaders = false)
     {
         $this->graby = $graby;
         $this->tagger = $tagger;
@@ -33,6 +34,7 @@ class ContentProxy
         $this->logger = $logger;
         $this->mimeGuesser = new MimeTypeExtensionGuesser();
         $this->fetchingErrorMessage = $fetchingErrorMessage;
+        $this->storeArticleHeaders = $storeArticleHeaders;
     }
 
     /**
@@ -51,6 +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
@@ -83,7 +86,7 @@ class ContentProxy
             (new LocaleConstraint())
         );
 
-        if (0 === count($errors)) {
+        if (0 === \count($errors)) {
             $entry->setLanguage($value);
 
             return;
@@ -105,7 +108,7 @@ class ContentProxy
             (new UrlConstraint())
         );
 
-        if (0 === count($errors)) {
+        if (0 === \count($errors)) {
             $entry->setPreviewPicture($value);
 
             return;
@@ -125,7 +128,7 @@ class ContentProxy
         $date = $value;
 
         // is it a timestamp?
-        if (filter_var($date, FILTER_VALIDATE_INT) !== false) {
+        if (false !== filter_var($date, FILTER_VALIDATE_INT)) {
             $date = '@' . $date;
         }
 
@@ -142,6 +145,91 @@ class ContentProxy
         }
     }
 
+    /**
+     * Helper to extract and save host from entry url.
+     *
+     * @param Entry $entry
+     */
+    public function setEntryDomainName(Entry $entry)
+    {
+        $domainName = parse_url($entry->getUrl(), PHP_URL_HOST);
+        if (false !== $domainName) {
+            $entry->setDomainName($domainName);
+        }
+    }
+
+    /**
+     * Helper to set a default title using:
+     * - url basename, if applicable
+     * - hostname.
+     *
+     * @param Entry $entry
+     */
+    public function setDefaultEntryTitle(Entry $entry)
+    {
+        $url = parse_url($entry->getUrl());
+        $path = pathinfo($url['path'], PATHINFO_BASENAME);
+
+        if (empty($path)) {
+            $path = $url['host'];
+        }
+
+        $entry->setTitle($path);
+    }
+
+    /**
+     * 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.
@@ -153,10 +241,7 @@ class ContentProxy
     {
         $entry->setUrl($content['url']);
 
-        $domainName = parse_url($entry->getUrl(), PHP_URL_HOST);
-        if (false !== $domainName) {
-            $entry->setDomainName($domainName);
-        }
+        $this->setEntryDomainName($entry);
 
         if (!empty($content['title'])) {
             $entry->setTitle($content['title']);
@@ -181,11 +266,11 @@ class ContentProxy
             $entry->setHttpStatus($content['status']);
         }
 
-        if (!empty($content['authors']) && is_array($content['authors'])) {
+        if (!empty($content['authors']) && \is_array($content['authors'])) {
             $entry->setPublishedBy($content['authors']);
         }
 
-        if (!empty($content['all_headers'])) {
+        if (!empty($content['all_headers']) && $this->storeArticleHeaders) {
             $entry->setHeaders($content['all_headers']);
         }
 
@@ -202,7 +287,7 @@ class ContentProxy
         }
 
         // if content is an image, define it as a preview too
-        if (!empty($content['content_type']) && in_array($this->mimeGuesser->guess($content['content_type']), ['jpeg', 'jpg', 'gif', 'png'], true)) {
+        if (!empty($content['content_type']) && \in_array($this->mimeGuesser->guess($content['content_type']), ['jpeg', 'jpg', 'gif', 'png'], true)) {
             $this->updatePreviewPicture($entry, $content['url']);
         }