Nicolas Seinlet <nicolas@xxxxxxxxxxx> writes: > I'm facing a strange behaviour and I don't understand why. From a wider query plan, https://explain.dalibo.com/plan/7yh ,a table is scanned and filter based on its primary key, returning 98 records. Then, those records are sorted on another field, and the output is 758,247,643 records. You haven't showed us the whole plan; tsk tsk. However, I'm going to bet that this sort is the inner input of a merge join, and that the apparent "extra" tuples are caused by the merge backing up and rescanning the sort result to deal with duplicate join keys in its outer input. There must be a heck of a lot of duplicates. The planner would typically avoid using merge join if it knew that, so I'm wondering if you have up-to-date statistics for the outer side of the join. regards, tom lane
