Skip to main content

Module cluster_repartition

Module cluster_repartition 

Source
Expand description

Cross-instance hash repartition for distributed GROUP BY. Gated on cluster because it pulls in the shuffle transport. Cross-instance hash repartition. Replaces DataFusion’s in-process RepartitionExec::Hash between AggregateExec::Partial and AggregateExec::FinalPartitioned. One output partition per owned vnode; remote rows ship via ShuffleSender.

Structs§

ClusterRepartitionExec
Cross-instance hash repartition.
DistributedJoinRule
Physical optimizer rule that wraps HashJoinExec inputs in ClusterRepartitionExec.

Functions§

set_cluster_context
Set the global cluster context for the distributed physical optimizer rules.