"D'Arcy" "J.M." Cain <darcy@druid.net> writes:
> After recent changes I find an error with SUM when summing more than
> one column. ...
> See how the individual sums are correct but I can no longer get both
> sums in one select.
Good eye!
Actually, it looks like *any* two aggregates conflict --- we're
reporting the result of the rightmost aggregate for all aggregate
functions in a SELECT. Using D'Arcy's test table, I also tried
treetest=> SELECT AVG(a), SUM(a) FROM x;
avg|sum
---+--- 3| 3
(1 row)
treetest=> SELECT AVG(a), SUM(b) FROM x;
avg|sum
---+---12| 12
(1 row)
treetest=> SELECT AVG(a), COUNT(b) FROM x;
avg|count
---+----- 2| 2
(1 row)
Oops.
This bug appears to explain some of the regression-test failures I'm
seeing --- numerology and select_having both contain multiple-aggregate
commands that are failing.
In the select_having test, it looks like multiple aggregates used in
the HAVING clause of a SELECT are suffering the same sort of fate
as those in the target list.
regards, tom lane