Paul Tillotson <spam1011@xxxxxxxxxxxx> writes: > Tom Lane wrote: >> Hardly likely, considering it's estimating only 296 rows in the subquery >> output. My bet is that you've chosen a datatype whose comparisons are >> not hashable (like char(n)). What is the datatype of parentid in these >> tables, anyway? >> > I don't have access to the machine now, but my memory is that > parent.parentid is numeric(10,2) and child.parentid is int. Offhand I don't believe there are any hashable crosstype comparisons. In this case the int is probably getting promoted to numeric, but I think numeric comparison isn't hashable either (because for example '0.0' = '0.000' but the internal representations are different). regards, tom lane ---------------------------(end of broadcast)--------------------------- TIP 6: explain analyze is your friend
