Today I watched a Java server crash because a library decided it needed more than 3GB of heap space to read a 10MB file. That was after manually removed background colors from around 100,000 cells, which apparently caused the parser to create even more objects in its internal representation of the sheet.

[–] folekaule@lemmy.world 2 points 1 month ago

Yeah, I get it. I've had many libraries fail me in as many ways, which is why I consider it lucky to not have to implement my own. I work in .net these days, but there have been times where I had to just dig into the xml inside the xlsx and use xml tools. Those were mostly one-offs, thankfully.

Back when I did Java I had a frustrating experience with IBM's libxml causing our app to crash after several days due to a memory leak. I didn't have access to the production environment so it took me probably 3 weeks to find the cause and only after digging through a crash dump provided by the sysadmin. Not related, but you triggered my traumatic memory :)