Ken Reese created TIKA-5017:
-------------------------------
Summary: PPTX SAX extractor iterates slides in "lexographic
relationship id" order instead "presentation" order
Key: TIKA-5017
URL: https://issues.apache.org/jira/browse/TIKA-5017
Project: Tika
Issue Type: Bug
Components: parser
Affects Versions: 4.0.0
Reporter: Ken Reese
Attachments: libreoffice-twelve-slides.pptx
This seems like a regression in behavior between the how the old PPTX DOM
extractor worked and how the newer PPTX SAX extractor works.
The current PPTX SAX extractor is iterating slides in lexographic order of
their relationship id instead of actual presentation order. This doesn't always
make sense - you can have `rId10` show up before `rId4` for example.
Tika is currently just using POI to iterate `ppt/_rels/presentation.xml.rels`
and returning the slides in whichever order POI returns in - which is just
using the relationship id as a key in a `TreeMap`. So we really shouldn't be
relying on the POI iteration order to give us the actual slide order back here.
I think we should instead be looking at `ppt/presentation.xml` to get the slide
ordering, which is how Tika used to work when it used the DOM extractor by
default. As of 4.0.0, the DOM extractor implementation has been completely
remove from Tika, so there is no longer an easy workaround.
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
