[
https://issues.apache.org/jira/browse/TIKA-2648?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16486904#comment-16486904
]
Sebastian Nagel commented on TIKA-2648:
---------------------------------------
Hi Gerard,
thanks for bringing this up. It actually affects some more MIME types related
to server-side scripting/webapp/etc. formats, see [discussion
user@tika|https://lists.apache.org/thread.html/1e4f4b6c249618a446f2e92f56ef90e6bfa0dfe51ce10197461df3d9@%3Cuser.tika.apache.org%3E]
(sorry, I've never found the time to dig into it).
The file extension {{.php}} is an indicator for the MIME type {{text/x-php}}
only for a (local) file system. It does practically give no glue for a URL -
it's likely HTML but the server may use PHP also to render PDF or to deliver
images. I don't know whether
* there is a way to model this distinction between URL and file glob patterns
in the tika-mimetypes.xml ?
* could the [probabilistic
detector|https://wiki.apache.org/tika/BaysianMimeTypeSelector] help? We could
give more weight to the Content-Type sent by the server. Of course, this may
affect the detection of {{.pdf}}, {{.docx}}, etc.
* only as last solution: modify tika-mimetypes.xml so that it matches the needs
of a crawler and maintain it separately (as part of Nutch).
Regarding your example: if the content is passed to Tika the MIME type is
correctly recognized. Of course, that's no excuse and often fails because
content is only a HTML fragment, not well-formed, or just not easy to detect.
> mime detection based on resource name detects resources as "text/x-php"
> instead of "text/html"
> -----------------------------------------------------------------------------------------------
>
> Key: TIKA-2648
> URL: https://issues.apache.org/jira/browse/TIKA-2648
> Project: Tika
> Issue Type: Bug
> Reporter: Gerard Bouchar
> Priority: Major
>
> When using tika to detect a mime type given only an URL containing ".php" and
> a content-type hint of "text/html", it guesses "text/x-php", whereas one
> could expect "text/html".
> {code}
> TikaConfig tika = new TikaConfig();
> Metadata metadata = new Metadata();
> String url = "https://www.facebook.com/home.php";;
> metadata.set(Metadata.RESOURCE_NAME_KEY, url);
> metadata.set(Metadata.CONTENT_TYPE, "text/html");
> MediaType type = tika.getDetector().detect(null, metadata);
> System.out.println(url + " is of type " + type.toString());
> // Prints https://www.facebook.com/home.php is of type text/x-php
> {code}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
