[
https://issues.apache.org/jira/browse/PDFBOX-6268?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18117871#comment-18117871
]
Tim Allison commented on PDFBOX-6268:
-------------------------------------
Thank you [~lehmi]! I ran a before and after on the 250k PDFs in our standard
regression corpus on Tika. This dramatically improved 4 pdfs, and somewhat
improved a bunch others. I didn't see any new surprises on the "extract text,
annotations and attachments" code paths.
> Avoid rebuilding key cache when possible
> ----------------------------------------
>
> Key: PDFBOX-6268
> URL: https://issues.apache.org/jira/browse/PDFBOX-6268
> Project: PDFBox
> Issue Type: Task
> Components: Parsing
> Reporter: Tim Allison
> Assignee: Andreas Lehmkühler
> Priority: Minor
> Attachments: G7BO7PNCCREVF2BCY5YSYOPYDLMBYASY.pdf,
> RFFQRH5P3P4YLUL5MWWZIPDQRKAPES2H.pdf
>
>
> Over on Tika, we found that our {{COSDocument.getObjectsByType(FILESPEC)}}
> was really expensive on a couple of PDFs.
> The issue was that every object stream parse rebuilds the cache from scratch:
> xref entries × object streams map insertions.
> With incoming PR, we could get >600s -> ~4 seconds or less.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
