Question regarding XSLX Performance issues

[Johan Warnar Lando Boekkooi]

Hey all,

First of thank you every much for looking at this issue and maintaining the Okapi Framework!

Now I was debugging a issue with a 15MB xslx file (sorry can’t share it) and noticed during profiling that this.stringItems.get in SharedStrings.java was taking forever to get a single item.

I did a quick MonkeyPath to replace the new LinkedList<>() on line 85 with a new java.util.ArrayList() this really increased the speed from over 15 minutes (after which I aborted the run) to around 20 seconds for my simple timeline.

Now I was wondering if there was a specific reason for using a LinkedList or if this could be changed to some other list to improve performance?

Looking forward to your answer and have a great weekend!

[jhargrave-straker]

That is very interesting! We should do a profile to see if there are any other bottlenecks.

[Johan Warnar Lando Boekkooi]

Hey Jim,

So https://www2.gov.bc.ca/assets/gov/bc-procurement-resources/bc-bid-guides/unspsc_v250901_updated-excel-version_27-feb-2023.xlsx is an example where you can clearly see that the this.stringItems.get is an issue.

Cheers,
Warnar

[Johan Warnar Lando Boekkooi]

So I opened https://bitbucket.org/okapiframework/okapi/pull-requests/769 which contains the patches that I have applied to Okapi to improve the performance a bit.
Sadly I didn’t find a benchmark method in Okapi Framework so proving the improvements was not really possible,

[jhargrave-straker]

Partially resolved with https://bitbucket.org/okapiframework/okapi/pull-requests/769