Our application implements a hybrid approach to XSSF parsing, where we exclude certain large sheets from DOM parsing and run them through a streaming parse later. I don't think our local changes make sense to merge into POI itself, but if XSSFSheet were refactored slightly, it would allow us to do our changes by subclassing, avoiding the need to modify POI itself and run with a patched version.
Made my proposed refactoring in r1694531. I'll leave this open for a few days in case any other devs want to object/discuss.
Any chance you could put a stripped-down version of your subclass into examples, to show users how they might take that approach?
Added example in r1694536
Further tidy up in r1694537
Thanks for the review, Nick. Calling this fixed now I've added the example.
Any chance you could put a mention about this somewhere in the spreadsheet documentation on the site?
Sorry, yes, I was leaving out of the docs until we'd discussed. Re-opening to track documentation.
... and documented on the website, svn r1695862.