However, as per the previous bug that I was investigating, I'm running into an offsets error. We're pushing i32 offsets into some arrow array, but this is causing a panic when the offset inevitably overflows for numbers larger than 2^31 - 1. I'll need to find the place where we're instantiating this i32 based indexing and probably just switch that over to using i64s instead. Or the original data could be deserialized into a LargeString instead (maybe by changing the original parquet schema type to "LargeString" or something?).