You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I initially thought no body will ever read a 20GB parquet file over network, but it turns out that many people do -- it's not entirely unreasonable, as we only need to download the portion of the file that is relevant to the query, usually only a few MBs.
Unfortunately, wasm32 is a 32 bit platform, in which the usize is 32 bit, meaning that larger parquet ranges are chopped to 4GB, which leads to incorrect decoding.
We need to fix the upstream object_store to actually fix this. apache/arrow-rs#5351
The text was updated successfully, but these errors were encountered:
I initially thought no body will ever read a 20GB parquet file over network, but it turns out that many people do -- it's not entirely unreasonable, as we only need to download the portion of the file that is relevant to the query, usually only a few MBs.
Unfortunately, wasm32 is a 32 bit platform, in which the
usize
is 32 bit, meaning that larger parquet ranges are chopped to 4GB, which leads to incorrect decoding.We need to fix the upstream
object_store
to actually fix this. apache/arrow-rs#5351The text was updated successfully, but these errors were encountered: