Optimizations of queries containing non-deterministic functions - difference between Calcite and Hive

[X L <xi...@gmail.com>]

Hi all,

When I read source code from Calcite and Hive, I notice that Hive invests
more logic than Calcite in checking whether a function call will result in
non-deterministic output. On the other hand, Calcite seems by default to
assume that a function is deterministic.
Can someone provide any insights on such a difference? I am wondering if
this derives from their different context of usages.

Thanks,
Xinyu