[
https://issues.apache.org/jira/browse/TIKA-1365?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14063467#comment-14063467
]
Ken Krugler commented on TIKA-1365:
-----------------------------------
Turning the XML parser into a fuzzy parser (like what TagSoup or NekoHTML do)
is a big change, and you'd be replicating what those components do in any case.
What I think might be in scope here is to improve HTML detection, as the
http://lucene.apache.org/core/discussion.html does return a Content-Type of
"text/html; charset=utf-8". The main problem is that the HTML is badly broken,
as it starts with:
<!-- core-simple.html -->
<head>
Which is really an HTML fragment.
> Incorrectly MimeType detection for Apache Lucene web site
> ---------------------------------------------------------
>
> Key: TIKA-1365
> URL: https://issues.apache.org/jira/browse/TIKA-1365
> Project: Tika
> Issue Type: Bug
> Components: detector
> Affects Versions: 1.5
> Reporter: Tien Nguyen Manh
> Attachments: discussion.html
>
>
> Tika 1.5 detect many page from apache lucene web site as xml, for example
> this page
> http://lucene.apache.org/core/discussion.html
> Here are error log:, it failed to parse becuase it use xml parser
> Apache Tika was unable to parse the document
> at http://lucene.apache.org/core/discussion.html.
> The full exception stack trace is included below:
> org.apache.tika.exception.TikaException: XML parse error
> at org.apache.tika.parser.xml.XMLParser.parse(XMLParser.java:78)
> at
> org.apache.tika.parser.CompositeParser.parse(CompositeParser.java:242)
> at
> org.apache.tika.parser.CompositeParser.parse(CompositeParser.java:242)
> at
> org.apache.tika.parser.AutoDetectParser.parse(AutoDetectParser.java:120)
> at org.apache.tika.gui.TikaGUI.handleStream(TikaGUI.java:320)
> at org.apache.tika.gui.TikaGUI.openURL(TikaGUI.java:293)
> at org.apache.tika.gui.TikaGUI.actionPerformed(TikaGUI.java:247)
> at
> javax.swing.AbstractButton.fireActionPerformed(AbstractButton.java:2018)
--
This message was sent by Atlassian JIRA
(v6.2#6252)
