X-Git-Url: https://git.immae.eu/?a=blobdiff_plain;f=src%2FWallabag%2FCoreBundle%2FHelper%2FContentProxy.php;h=f0d8c1b420c9ba5fa192eb43b4949d84d1f79b5a;hb=4a81360efcdfe4bab8d75f7227c9cf5bfd514189;hp=0c971863bcd670957a7f4c0f6d172d98bebdf4f0;hpb=a8993999893fb665b5e96597f9b649a2a8b940a2;p=github%2Fwallabag%2Fwallabag.git

diff --git a/src/Wallabag/CoreBundle/Helper/ContentProxy.php b/src/Wallabag/CoreBundle/Helper/ContentProxy.php
index 0c971863..f0d8c1b4 100644
--- a/src/Wallabag/CoreBundle/Helper/ContentProxy.php
+++ b/src/Wallabag/CoreBundle/Helper/ContentProxy.php
@@ -4,12 +4,12 @@ namespace Wallabag\CoreBundle\Helper;
 
 use Graby\Graby;
 use Psr\Log\LoggerInterface;
-use Wallabag\CoreBundle\Entity\Entry;
-use Wallabag\CoreBundle\Tools\Utils;
 use Symfony\Component\HttpFoundation\File\MimeType\MimeTypeExtensionGuesser;
 use Symfony\Component\Validator\Constraints\Locale as LocaleConstraint;
 use Symfony\Component\Validator\Constraints\Url as UrlConstraint;
 use Symfony\Component\Validator\Validator\ValidatorInterface;
+use Wallabag\CoreBundle\Entity\Entry;
+use Wallabag\CoreBundle\Tools\Utils;
 
 /**
  * This kind of proxy class take care of getting the content from an url
@@ -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
@@ -63,9 +66,178 @@ 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);
     }
 
+    /**
+     * Use a Symfony validator to ensure the language is well formatted.
+     *
+     * @param Entry  $entry
+     * @param string $value Language to validate and save
+     */
+    public function updateLanguage(Entry $entry, $value)
+    {
+        // some lang are defined as fr-FR, es-ES.
+        // replacing - by _ might increase language support
+        $value = str_replace('-', '_', $value);
+
+        $errors = $this->validator->validate(
+            $value,
+            (new LocaleConstraint())
+        );
+
+        if (0 === \count($errors)) {
+            $entry->setLanguage($value);
+
+            return;
+        }
+
+        $this->logger->warning('Language validation failed. ' . (string) $errors);
+    }
+
+    /**
+     * Use a Symfony validator to ensure the preview picture is a real url.
+     *
+     * @param Entry  $entry
+     * @param string $value URL to validate and save
+     */
+    public function updatePreviewPicture(Entry $entry, $value)
+    {
+        $errors = $this->validator->validate(
+            $value,
+            (new UrlConstraint())
+        );
+
+        if (0 === \count($errors)) {
+            $entry->setPreviewPicture($value);
+
+            return;
+        }
+
+        $this->logger->warning('PreviewPicture validation failed. ' . (string) $errors);
+    }
+
+    /**
+     * Update date.
+     *
+     * @param Entry  $entry
+     * @param string $value Date to validate and save
+     */
+    public function updatePublishedAt(Entry $entry, $value)
+    {
+        $date = $value;
+
+        // is it a timestamp?
+        if (false !== filter_var($date, FILTER_VALIDATE_INT)) {
+            $date = '@' . $date;
+        }
+
+        try {
+            // is it already a DateTime?
+            // (it's inside the try/catch in case of fail to be parse time string)
+            if (!$date instanceof \DateTime) {
+                $date = new \DateTime($date);
+            }
+
+            $entry->setPublishedAt($date);
+        } catch (\Exception $e) {
+            $this->logger->warning('Error while defining date', ['e' => $e, 'url' => $entry->getUrl(), 'date' => $value]);
+        }
+    }
+
+    /**
+     * 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.
@@ -75,9 +247,14 @@ class ContentProxy
      */
     private function stockEntry(Entry $entry, array $content)
     {
-        $title = $content['title'];
-        if (!$title && !empty($content['open_graph']['og_title'])) {
-            $title = $content['open_graph']['og_title'];
+        $entry->setUrl($content['url']);
+
+        $this->setEntryDomainName($entry);
+
+        if (!empty($content['title'])) {
+            $entry->setTitle($content['title']);
+        } elseif (!empty($content['open_graph']['og_title'])) {
+            $entry->setTitle($content['open_graph']['og_title']);
         }
 
         $html = $content['html'];
@@ -90,58 +267,40 @@ class ContentProxy
             }
         }
 
-        $entry->setUrl($content['url']);
-        $entry->setTitle($title);
         $entry->setContent($html);
-        $entry->setHttpStatus(isset($content['status']) ? $content['status'] : '');
-
-        if (!empty($content['date'])) {
-            $date = $content['date'];
-
-            // is it a timestamp?
-            if (filter_var($date, FILTER_VALIDATE_INT) !== false) {
-                $date = '@'.$content['date'];
-            }
+        $entry->setReadingTime(Utils::getReadingTime($html));
 
-            try {
-                $entry->setPublishedAt(new \DateTime($date));
-            } catch (\Exception $e) {
-                $this->logger->warning('Error while defining date', ['e' => $e, 'url' => $content['url'], 'date' => $content['date']]);
-            }
+        if (!empty($content['status'])) {
+            $entry->setHttpStatus($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']);
         }
 
-        $this->validateAndSetLanguage(
-            $entry,
-            isset($content['language']) ? $content['language'] : ''
-        );
+        if (!empty($content['date'])) {
+            $this->updatePublishedAt($entry, $content['date']);
+        }
 
-        $this->validateAndSetPreviewPicture(
-            $entry,
-            isset($content['open_graph']['og_image']) ? $content['open_graph']['og_image'] : ''
-        );
+        if (!empty($content['language'])) {
+            $this->updateLanguage($entry, $content['language']);
+        }
 
-        // 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)) {
-            $this->validateAndSetPreviewPicture(
-                $entry,
-                $content['url']
-            );
+        if (!empty($content['open_graph']['og_image'])) {
+            $this->updatePreviewPicture($entry, $content['open_graph']['og_image']);
         }
 
-        $entry->setMimetype(isset($content['content_type']) ? $content['content_type'] : '');
-        $entry->setReadingTime(Utils::getReadingTime($html));
+        // 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)) {
+            $this->updatePreviewPicture($entry, $content['url']);
+        }
 
-        $domainName = parse_url($entry->getUrl(), PHP_URL_HOST);
-        if (false !== $domainName) {
-            $entry->setDomainName($domainName);
+        if (!empty($content['content_type'])) {
+            $entry->setMimetype($content['content_type']);
         }
 
         try {
@@ -165,52 +324,4 @@ class ContentProxy
     {
         return !empty($content['title']) && !empty($content['html']) && !empty($content['url']);
     }
-
-    /**
-     * Use a Symfony validator to ensure the language is well formatted.
-     *
-     * @param Entry  $entry
-     * @param string $value Language to validate
-     */
-    private function validateAndSetLanguage($entry, $value)
-    {
-        // some lang are defined as fr-FR, es-ES.
-        // replacing - by _ might increase language support
-        $value = str_replace('-', '_', $value);
-
-        $errors = $this->validator->validate(
-            $value,
-            (new LocaleConstraint())
-        );
-
-        if (0 === count($errors)) {
-            $entry->setLanguage($value);
-
-            return;
-        }
-
-        $this->logger->warning('Language validation failed. '.(string) $errors);
-    }
-
-    /**
-     * Use a Symfony validator to ensure the preview picture is a real url.
-     *
-     * @param Entry  $entry
-     * @param string $value URL to validate
-     */
-    private function validateAndSetPreviewPicture($entry, $value)
-    {
-        $errors = $this->validator->validate(
-            $value,
-            (new UrlConstraint())
-        );
-
-        if (0 === count($errors)) {
-            $entry->setPreviewPicture($value);
-
-            return;
-        }
-
-        $this->logger->warning('PreviewPicture validation failed. '.(string) $errors);
-    }
 }