Hacker News new | ask | show | jobs
by tpetry 1346 days ago
Most parallel operations in PG are implemented by simple merge the dataset, work independently and merge the results. I expect the new distinct to behave the same and not work on a shared data structure.
1 comments

Thanks! May be helpful to include this in the documentation, since I guess it will then often depend on the numDistinctRows estimate [1] if the parallel plan is used.

[1] https://git.postgresql.org/gitweb/?p=postgresql.git;a=blob;f...