Le jeudi 1 juillet 2021, 09:09:38 CEST Ronan Dunklau a écrit : > > Yes, thanks! I was making a big mistake here thinking the executor can > > stop after the first match. That's not true. We need to use each outer > > tuple to find all the matches and mark the corresponding hashtable > > entries. I have updated the patch with the fix. > > It looks OK to me.
I forgot to mention: you also have failing tests due to the plans changing to use the new join type. This might not be the case anymore once you update the cost model, but if that's the case the tests should be updated.
Thanks! Test cases are updated in v3 patch. Also merge join can do the 'right anti join' too in the same patch.
Thanks again for reviewing this patch.
Thanks
Richard
Hi,
Minor comment:
+ * In a right-antijoin, we never return a matched tuple.