When is better to use the NodeHashJoin operator?


When hashJoin operators are the best to use?

I think join hints are the only direct way to you can tell the planner to use a NodeHashJoin.

Often I find that hash joins in the plan can be problematic as far as query time and GCs, but they can be useful when you explicitly have a small set of start and end nodes, and when you have a known supernode between them, where the cost to expand out from the supernode is expensive, compared to the cost to only traverse to the supernode.

I have a knowledge base article which describes this scenario:

1 Like