Rob Tulloh created TIKA-1584:
--------------------------------
Summary: Tika 1.7 possible regression (nested attachment files not
getting parsed)
Key: TIKA-1584
URL: https://issues.apache.org/jira/browse/TIKA-1584
Project: Tika
Issue Type: Bug
Components: server
Affects Versions: 1.7
Reporter: Rob Tulloh
I tried to send this to the tika user list, but got a qmail failure so I am
opening a jira to see if I can get help with this.
There appears to be a change in the behavior of tika since 1.5 (the last
version we have used). In 1.5, if we pass a file with content type of rfc822
which contains a zip that contains a docx file, the entire content would get
recursed and the text returned. In 1.7, tika only unwinds as far as the zip
file and ignores the content of the contained docx file. This is causing a
regression failure in our search tests because the contents of the docx file
are not found when searched for.
We are testing with tika-server if this helps. If we ask the meta service to
just characterize the test data, it correctly determines the input is of type
rfc822. However, on extract, the contents of the attachment are not extracted
as expected.
curl -X PUT -T test.eml -q -H Content-Type:application/octet-stream
http://localhost:9998/meta 2>/dev/null | grep Content-Type
"Content-Type","message/rfc822"
curl -X PUT -T test.eml -q -H Content-Type:application/octet-stream
http://localhost:9998/tika 2>/dev/null | grep docx
sign.docx <<<<--- this is not expected, need contents of this extracted
We can easily reproduce this problem with a simple eml file with an attachment.
Can someone please comment if this seems like a problem or perhaps we need to
change something in our call to get the old behavior?
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
