cloud-fan commented on issue #25347: [SPARK-28610][SQL] Allow having a decimal
buffer for long sum
URL: https://github.com/apache/spark/pull/25347#issuecomment-525763443
Now I'm wondering if it's worthwhile to sacrifice 10x perf to handle
intermedia overflow. It's not about correctness but
cloud-fan commented on issue #25347: [SPARK-28610][SQL] Allow having a decimal
buffer for long sum
URL: https://github.com/apache/spark/pull/25347#issuecomment-524876439
@mgaido91 can you do a simple microbenchmark? If the performance overhead is
not significant, we can use decimal as sum
cloud-fan commented on issue #25347: [SPARK-28610][SQL] Allow having a decimal
buffer for long sum
URL: https://github.com/apache/spark/pull/25347#issuecomment-523465552
We need to be careful when adding configs to control query semantic. We have
a lot of such configs now, but most of them
cloud-fan commented on issue #25347: [SPARK-28610][SQL] Allow having a decimal
buffer for long sum
URL: https://github.com/apache/spark/pull/25347#issuecomment-523426015
If different databases have different behaviors, then I don't see a strong
reason to change the return type of sum in Sp
cloud-fan commented on issue #25347: [SPARK-28610][SQL] Allow having a decimal
buffer for long sum
URL: https://github.com/apache/spark/pull/25347#issuecomment-522569449
We can create the coarser flag later.
For this PR, in general LGTM, but let's make the story about sum completed.
cloud-fan commented on issue #25347: [SPARK-28610][SQL] Allow having a decimal
buffer for long sum
URL: https://github.com/apache/spark/pull/25347#issuecomment-522481955
@maropu good point. I think it's more than overflow now. Maybe we should
create a config to specify the dialect, and pro
