2 ms·
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 sa
by tpetry 4y 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.
- jeltz 4y agoYou are correct: https://git.postgresql.org/gitweb/?p=postgresql.git;a=commitdiff;h=22c4e88ebff408acd52e212543a77158bde59e69 https://git.postgresql.org/gitweb/?p=postgresql.git;a=commit...
- chrstr 4y agoThanks! 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=src/backend/optimizer/plan/planner.c;h=468105d91ea71dc195aae1ed12ab3a2044128ca2;hb=refs/heads/REL_15_STABLE#l4420 https://git.postgresql.org/gitweb/?p=postgresql.git;a=blob;f...