"1. Is it so those projects could use the arrow compute kernels instead of
their own?"
This is the case we met. We offloaded Spark JVM SQL engine to native library
(gazelle-plugin) which is based on arrow computer kernel, but customer asked if
it's possible to use Velox as native backend becaus
This is excellent proposal. We worked on apache arrow based native Spark SQL
engine offloading for 1.5 years. Currently we extended Gandiva's node and
expression definition to pass query plan from Spark to native code. We will try
start to try substrait. These features can definitely help:
(O)
