Claus Ibsen created CAMEL-27201:
-----------------------------------
Summary: camel-tika - html and xml output fail on a NUL character
in document metadata
Key: CAMEL-27201
URL: https://issues.apache.org/jira/browse/CAMEL-27201
Project: Camel
Issue Type: Bug
Components: camel-tika
Reporter: Claus Ibsen
With {{tikaParseOutputFormat=html}} or {{xml}}, parsing a document whose
metadata (here the PDF title) contains a character that XML does not allow,
such as NUL (U+0000), fails. With {{text}} the same document is read fine. This
happened with 3 of 4 ordinary PDFs tested (two made by macOS {{cupsfilter}},
and {{multi_page.pdf}} from the docling test data).
h3. How to reproduce
{code}
from("file:in")
.to("tika:parse?tikaParseOutputFormat=html")
.to("log:out");
{code}
{noformat}
org.apache.tika.exception.TikaException: Unable to extract PDF content
...
Caused by: org.xml.sax.SAXException: An invalid XML character (Unicode: 0x0)
was found in the node's character data content.
at
java.xml/com.sun.org.apache.xml.internal.serializer.ToStream.accumDefaultEscape(ToStream.java:1749)
...
at
java.xml/com.sun.org.apache.xml.internal.serializer.ToHTMLStream.endElement(ToHTMLStream.java:913)
at
java.xml/com.sun.org.apache.xalan.internal.xsltc.trax.TransformerHandlerImpl.endElement(TransformerHandlerImpl.java:282)
at
org.apache.tika.sax.ContentHandlerDecorator.endElement(ContentHandlerDecorator.java:134)
at
org.apache.tika.sax.ExpandedTitleContentHandler.endElement(ExpandedTitleContentHandler.java:69)
...
at
org.apache.tika.sax.XHTMLContentHandler.lazyEndHead(XHTMLContentHandler.java:226)
{noformat}
(Camel 4.24.0-SNAPSHOT, Tika 4.1.0.)
h3. Where
{{components/camel-tika/src/main/java/org/apache/camel/component/tika/TikaProducer.java}}:
{{getContentHandler}} / {{getTransformerHandler}}. The JDK
{{TransformerHandler}} gets the SAX events as they are, and its serializer
rejects characters that are not allowed in XML.
h3. Suggested fix
Filter the characters before they reach the {{TransformerHandler}}, for example
by wrapping it in Tika's {{org.apache.tika.sax.SafeContentHandler}} (which
replaces invalid XML characters with U+FFFD), or a small
{{ContentHandlerDecorator}} that does the same for {{characters}} and attribute
values. Add a test with a document whose title holds a NUL.
Found while adding an {{outputFormat}} option to the
{{tika-extract-text-action}} Kamelet; its description now warns that html/xml
fail on such documents.
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)