[
https://issues.apache.org/jira/browse/TIKA-4444?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17987230#comment-17987230
]
Peter Hoogendijk edited comment on TIKA-4444 at 7/1/25 2:10 PM:
----------------------------------------------------------------
Reading TIKA-1974 and checking the standard, it is clear that the
"dc:keywords" tag is deprecated, but something strange is still happening.
PyPDF and exiftool show separate entries for these tags (with unique values in
the attached file):
* xmp-dc:subject (from the XMP-dc metadata)
* pdf:docinfo:subject (from the PDF metadata, shown by PyPDF as /Subject)
* xmp-pdf:keywords (from the XMP-pdf metadata, not the XMP-dc metadata)
* pdf:docinfo:keywords (from the PDF metadata, shown by PyPDF as /Keywords)
Tika does _not_ report the "xmp-pdf:keywords" (from the XMP-pdf metadata) but
_does_ include the value of this tag in "dc:subject". So in my opinion Tika
should:
* either report the "xmp-pdf:keywords" (from the XMP-pdf metadata) and include
the value of this tag in "dc:subject"
* or NOT report the "xmp-pdf:keywords" (from the XMP-pdf metadata) and NOT
include the value of this tag in "dc:subject"
So it seems this is not about the xmp-dc parser but about the xmp-pdf parser:
the [XMP-pdf namespace|http://ns.adobe.com/pdf/1.3/] lists "pdf:keywords".
What is your opinion?
was (Author: JIRAUSER310088):
Reading TIKA-1974 and checking the standard, it is clear that the
"dc:keywords" tag is deprecated, but something strange is still happening.
PyPDF and exiftool show separate entries for these tags (with unique values in
the attached file):
* xmp-dc:subject (from the XMP-dc metadata)
* pdf:docinfo:subject (from the PDF metadata, shown by PyPDF as /Subject)
* xmp-pdf:keywords (from the XMP-pdf metadata, not the XMP-dc metadata)
* pdf:docinfo:keywords (from the PDF metadata, shown by PyPDF as /Keywords)
Tika does _not_ report the "xmp-pdf:keywords" (from the XMP-pdf metadata) but
_does_ include the value of this tag in "dc:subject". So in my opinion Tika
should:
* either report the "xmp-pdf:keywords" (from the XMP-pdf metadata) and include
the value of this tag in "dc:subject"
* or NOT report the "xmp-pdf:keywords" (from the XMP-pdf metadata) and NOT
include the value of this tag in "dc:subject"
So it seems this is not about the xmp-dc parser but about the xmp-pdf parser:
the [XMP-pdf namespace|[http://example.com|http://ns.adobe.com/pdf/1.3/]] lists
"pdf:keywords".
What is your opinion?
> PDFParser shows wrong data in xmp "dc:subject" tag
> --------------------------------------------------
>
> Key: TIKA-4444
> URL: https://issues.apache.org/jira/browse/TIKA-4444
> Project: Tika
> Issue Type: Bug
> Components: parser
> Affects Versions: 3.2.0
> Environment: * Docker container based on python:3-slim
> * Debian 12.11
> * Python 3.13.5
> * openjdk 17.0.15 2025-04-15
> * tika-server-standard-3.2.0.jar
> * tika-server-standard-3.2.2-20250624.143628-8.jar
> * pdfbox-app-3.0.5.jar
> * PyPDF 5.6.1
> Reporter: Peter Hoogendijk
> Assignee: Tilman Hausherr
> Priority: Major
> Labels: xmp
> Fix For: 4.0.0, 3.2.1
>
> Attachments: lorem-ipsum.pdf, lorem-ipsum.xml
>
>
> The xmp metadata "dc:subject" tag contains the wrong data: it shows a list
> with the data from the following tags:
> * pdf:docinfo:subject (from the pdf metadata)
> * pdf:docinfo:keywords (from the pdf metadata)
> * pdf:keywords (from the xmp metadata)
> And it is missing the data from the following tags:
> * dc:subject (from the xmp metadata)
> When looking at the XML for my testfile (see attachments) the xmp metadata
> contains the correct "dc:subject" and "pdf:keywords" but:
> * Tika shows the wrong data in "dc:subject" (from the xmp metadata)
> * Tika does not show "pdf:keywords" (from the xmp metadata)
> * Tika does not show the actual "dc:subject" (from the xmp metadata)
> This has been tested with tika-server-standard-3.2.0.jar and
> tika-server-standard-3.2.2-20250624.143628-8.jar
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
