3 namespace Wallabag\CoreBundle\Helper
;
6 use Psr\Log\LoggerInterface
;
7 use Symfony\Component\HttpFoundation\File\MimeType\MimeTypeExtensionGuesser
;
8 use Symfony\Component\Validator\Constraints\Locale
as LocaleConstraint
;
9 use Symfony\Component\Validator\Constraints\Url
as UrlConstraint
;
10 use Symfony\Component\Validator\Validator\ValidatorInterface
;
11 use Wallabag\CoreBundle\Entity\Entry
;
12 use Wallabag\CoreBundle\Tools\Utils
;
15 * This kind of proxy class take care of getting the content from an url
16 * and update the entry with what it found.
24 protected $mimeGuesser;
25 protected $fetchingErrorMessage;
26 protected $eventDispatcher;
27 protected $storeArticleHeaders;
29 public function __construct(Graby
$graby, RuleBasedTagger
$tagger, ValidatorInterface
$validator, LoggerInterface
$logger, $fetchingErrorMessage, $storeArticleHeaders = false)
31 $this->graby
= $graby;
32 $this->tagger
= $tagger;
33 $this->validator
= $validator;
34 $this->logger
= $logger;
35 $this->mimeGuesser
= new MimeTypeExtensionGuesser();
36 $this->fetchingErrorMessage
= $fetchingErrorMessage;
37 $this->storeArticleHeaders
= $storeArticleHeaders;
41 * Update entry using either fetched or provided content.
43 * @param Entry $entry Entry to update
44 * @param string $url Url of the content
45 * @param array $content Array with content provided for import with AT LEAST keys title, html, url to skip the fetchContent from the url
46 * @param bool $disableContentUpdate Whether to skip trying to fetch content using Graby
48 public function updateEntry(Entry
$entry, $url, array $content = [], $disableContentUpdate = false)
50 if (!empty($content['html'])) {
51 $content['html'] = $this->graby
->cleanupHtml($content['html'], $url);
54 if ((empty($content) || false === $this->validateContent($content)) && false === $disableContentUpdate) {
55 $fetchedContent = $this->graby
->fetchContent($url);
56 $fetchedContent['title'] = $this->sanitizeContentTitle($fetchedContent['title'], $fetchedContent['content_type']);
58 // when content is imported, we have information in $content
59 // in case fetching content goes bad, we'll keep the imported information instead of overriding them
60 if (empty($content) || $fetchedContent['html'] !== $this->fetchingErrorMessage
) {
61 $content = $fetchedContent;
65 // be sure to keep the url in case of error
66 // so we'll be able to refetch it in the future
67 $content['url'] = !empty($content['url']) ? $content['url'] : $url;
69 $this->stockEntry($entry, $content);
73 * Use a Symfony validator to ensure the language is well formatted.
76 * @param string $value Language to validate and save
78 public function updateLanguage(Entry
$entry, $value)
80 // some lang are defined as fr-FR, es-ES.
81 // replacing - by _ might increase language support
82 $value = str_replace('-', '_', $value);
84 $errors = $this->validator
->validate(
86 (new LocaleConstraint())
89 if (0 === \
count($errors)) {
90 $entry->setLanguage($value);
95 $this->logger
->warning('Language validation failed. ' . (string) $errors);
99 * Use a Symfony validator to ensure the preview picture is a real url.
101 * @param Entry $entry
102 * @param string $value URL to validate and save
104 public function updatePreviewPicture(Entry
$entry, $value)
106 $errors = $this->validator
->validate(
108 (new UrlConstraint())
111 if (0 === \
count($errors)) {
112 $entry->setPreviewPicture($value);
117 $this->logger
->warning('PreviewPicture validation failed. ' . (string) $errors);
123 * @param Entry $entry
124 * @param string $value Date to validate and save
126 public function updatePublishedAt(Entry
$entry, $value)
130 // is it a timestamp?
131 if (false !== filter_var($date, FILTER_VALIDATE_INT
)) {
136 // is it already a DateTime?
137 // (it's inside the try/catch in case of fail to be parse time string)
138 if (!$date instanceof \DateTime
) {
139 $date = new \
DateTime($date);
142 $entry->setPublishedAt($date);
143 } catch (\Exception
$e) {
144 $this->logger
->warning('Error while defining date', ['e' => $e, 'url' => $entry->getUrl(), 'date' => $value]);
149 * Helper to extract and save host from entry url.
151 * @param Entry $entry
153 public function setEntryDomainName(Entry
$entry)
155 $domainName = parse_url($entry->getUrl(), PHP_URL_HOST
);
156 if (false !== $domainName) {
157 $entry->setDomainName($domainName);
162 * Helper to set a default title using:
163 * - url basename, if applicable
166 * @param Entry $entry
168 public function setDefaultEntryTitle(Entry
$entry)
170 $url = parse_url($entry->getUrl());
171 $path = pathinfo($url['path'], PATHINFO_BASENAME
);
174 $path = $url['host'];
177 $entry->setTitle($path);
181 * Try to sanitize the title of the fetched content from wrong character encodings and invalid UTF-8 character.
184 * @param $contentType
188 private function sanitizeContentTitle($title, $contentType)
190 if ('application/pdf' === $contentType) {
191 $title = $this->convertPdfEncodingToUTF8($title);
194 return $this->sanitizeUTF8Text($title);
198 * If the title from the fetched content comes from a PDF, then its very possible that the character encoding is not
199 * UTF-8. This methods tries to identify the character encoding and translate the title to UTF-8.
203 * @return string (maybe contains invalid UTF-8 character)
205 private function convertPdfEncodingToUTF8($title)
207 // first try UTF-8 because its easier to detect its present/absence
208 foreach (['UTF-8', 'UTF-16BE', 'WINDOWS-1252'] as $encoding) {
209 if (mb_check_encoding($title, $encoding)) {
210 return mb_convert_encoding($title, 'UTF-8', $encoding);
218 * Remove invalid UTF-8 characters from the given string.
220 * @param string $rawText
224 private function sanitizeUTF8Text($rawText)
226 if (mb_check_encoding($rawText, 'UTF-8')) {
230 return iconv('UTF-8', 'UTF-8//IGNORE', $rawText);
234 * Stock entry with fetched or imported content.
235 * Will fall back to OpenGraph data if available.
237 * @param Entry $entry Entry to stock
238 * @param array $content Array with at least title, url & html
240 private function stockEntry(Entry
$entry, array $content)
242 $entry->setUrl($content['url']);
244 $this->setEntryDomainName($entry);
246 if (!empty($content['title'])) {
247 $entry->setTitle($content['title']);
248 } elseif (!empty($content['open_graph']['og_title'])) {
249 $entry->setTitle($content['open_graph']['og_title']);
252 $html = $content['html'];
253 if (false === $html) {
254 $html = $this->fetchingErrorMessage
;
256 if (!empty($content['open_graph']['og_description'])) {
257 $html .= '<p><i>But we found a short description: </i></p>';
258 $html .= $content['open_graph']['og_description'];
262 $entry->setContent($html);
263 $entry->setReadingTime(Utils
::getReadingTime($html));
265 if (!empty($content['status'])) {
266 $entry->setHttpStatus($content['status']);
269 if (!empty($content['authors']) && \
is_array($content['authors'])) {
270 $entry->setPublishedBy($content['authors']);
273 if (!empty($content['all_headers']) && $this->storeArticleHeaders
) {
274 $entry->setHeaders($content['all_headers']);
277 if (!empty($content['date'])) {
278 $this->updatePublishedAt($entry, $content['date']);
281 if (!empty($content['language'])) {
282 $this->updateLanguage($entry, $content['language']);
285 if (!empty($content['open_graph']['og_image'])) {
286 $this->updatePreviewPicture($entry, $content['open_graph']['og_image']);
289 // if content is an image, define it as a preview too
290 if (!empty($content['content_type']) && \
in_array($this->mimeGuesser
->guess($content['content_type']), ['jpeg', 'jpg', 'gif', 'png'], true)) {
291 $this->updatePreviewPicture($entry, $content['url']);
294 if (!empty($content['content_type'])) {
295 $entry->setMimetype($content['content_type']);
299 $this->tagger
->tag($entry);
300 } catch (\Exception
$e) {
301 $this->logger
->error('Error while trying to automatically tag an entry.', [
302 'entry_url' => $content['url'],
303 'error_msg' => $e->getMessage(),
309 * Validate that the given content has at least a title, an html and a url.
311 * @param array $content
313 * @return bool true if valid otherwise false
315 private function validateContent(array $content)
317 return !empty($content['title']) && !empty($content['html']) && !empty($content['url']);