Having count- count(count(1))

Question

I have two queries below- first of which outputs one record, the Second query doesn't.
Can you explain why First query returns always records, at least one record? In explanation, can you say which process is runned after which (something like where is runned firstly, then group by, then...)?
--First:
select count(count(1)) from dual
where 1=2
group by dummy
having count(dummy) > 0;--1 row, value: 0
--Second:
select count(1) from dual
where 1=2
group by dummy
having count(dummy) > 0;--0 rows

Answer 1

Hi,
Aggregate queries with nested aggregate functions (such as "COUNT (COUNT (1))") always return exactly 1 row, even if there are no rows in the table (or, as in your first example, no rows survive the WHERE clause).
The outer layer of nested aggreate functions is evaluated after the GROUP BY and HAVING clauses have been applied.
If you have a GROUP BY clause, but no nested aggregate fucntions, then the output contains one row for every distinct group, subject to the WHERE- and HAVING clauses. In your second example, no rows are left after the WHERE clause, so there will be no rows in the output.
You can think of things as happening in this order
(1) WHERE
(2) GROUP BY
(3) (inner) aggregate fucntions
(4) HAVING
(5) Nested aggregate functions
The optimizer may not actually follow the steps above, but you won't notice that.
Edited by: Frank Kulash on Mar 25, 2010 11:51 AM
The following thread may help you:
Re: group function