X-Git-Url: https://git.immae.eu/?a=blobdiff_plain;f=src%2FWallabag%2FCoreBundle%2FHelper%2FContentProxy.php;h=1a2a330ff085f5e922da295fbd5c908611936545;hb=fc040c749dec0275e562182562c1c1cb89e6cfa1;hp=1ac7ad83c8ab753231ea72ec5c05db1205ae9aaf;hpb=d69a66a58e958a5aa426085083bb93637ca3f61f;p=github%2Fwallabag%2Fwallabag.git diff --git a/src/Wallabag/CoreBundle/Helper/ContentProxy.php b/src/Wallabag/CoreBundle/Helper/ContentProxy.php index 1ac7ad83..1a2a330f 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 @@ -63,6 +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, $content); } @@ -83,7 +93,7 @@ class ContentProxy (new LocaleConstraint()) ); - if (0 === count($errors)) { + if (0 === \count($errors)) { $entry->setLanguage($value); return; @@ -105,7 +115,7 @@ class ContentProxy (new UrlConstraint()) ); - if (0 === count($errors)) { + if (0 === \count($errors)) { $entry->setPreviewPicture($value); return; @@ -125,7 +135,7 @@ 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 +152,91 @@ 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. @@ -151,12 +246,9 @@ class ContentProxy */ private function stockEntry(Entry $entry, array $content) { - $entry->setUrl($content['url']); + $this->updateOriginUrl($entry, $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 +273,11 @@ 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 +294,7 @@ 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']); } @@ -220,6 +312,51 @@ 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) { + $parsed_entry_url = parse_url($entry->getUrl()); + $parsed_content_url = parse_url($url); + + $diff_ec = array_diff_assoc($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($diff_keys); + + switch ($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']: + case ['query']: + case ['fragment', 'query']: + // noop + break; + default: + if (empty($entry->getOriginUrl())) { + $entry->setOriginUrl($entry->getUrl()); + } + $entry->setUrl($url); + break; + } + } + } + /** * Validate that the given content has at least a title, an html and a url. *