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'd prefer supporting Apache Arrow due to it's use in Nvidia RAPIDS and the base format of their GPU accelerated DataFrame library CuDF: https://github.com/rapidsai/cudf.
I have no time to tackle that for the foreseeable future though.
Just for your consideration.
Feather file format seems to have excellent performance while Parquet seems to be more oriented for long term storage as explained here.
It looks like feather development is now maintained under Apache's Arrow
Some results, benchmarking: csv, pickle, messagepack, HDF5, feather and parquet
It looks like Feather requires the use of Flatbuffers. There seems to be a pure Nim library: skflatbuffers.
Another serialization format to explore: fst. Feather, Parquet and FST are explained here.
The text was updated successfully, but these errors were encountered: