MCP Report performance slow about acs-aem-commons

[kwin]

PRs are always appreciated in open source!

[ghenzler]

changing from true to false is easy - the question is more if everybody is happy with generating the xls file directly at the end so it does not have to be regenerated on every xlsx-download? (maybe there was reasons it was done the way it is today, before I create a PR for that I want to hear that the "MCP maintainers" are fine with it...)

[YegorKozlov]

MCP uses Apache POI to build xlsx which is a memory hog. Generating a report with 80K rows can take quite a bit of time and memory, and even potentially result in OOM). Maybe we should make the output format configurable and allow user to choose between xlsx and csv. So there are two potential improvements:

true => false will reduce the execution time
xlsx => csv will reduce the download time

[ghenzler]

I think it's ok to use xlsx, it's just more convenient to open & distribute (as the reports are often sent to business people from my experience). I think there is little reason to not change from true to false to speed up the creation of the JCR nodes significantly.

My question is mostly if somebody objects to create the xlsx report immediately along with the JCR nodes, so on download the existing binary can just be quickly spooled out.

[YegorKozlov]

My question is mostly if somebody objects to create the xlsx report immediately along with the JCR nodes, so on download the existing binary can just be quickly spooled out.

it should be perfectly fine to create the report as a step in the MCP job. Ideally it would be backwards compatible, i.e. if the report is pre-built (new mode) then download it, otherwise create it on the fly (current behavior).