Question about GROUP BY and HAVING

Question

Good afternoon,
I have the following query which returns the desired result (set of students who take CS112 or CS114 but not both). I wanted to "condense" it into a single SELECT statement (if that is at all possible - DDL to execute the statement is provided at the end of this post):
-- is this select distinct * and its associated where clause absolutely
-- necessary to obtain the result ?
select distinct *
  from (
        select s.sno,
               s.sname,
               s.age,
               sum(case when t.cno in ('CS112', 'CS114')
                        then 1
                        else 0
                   end)
                 over (partition by s.sno) as takes_either_or_both
          from student s join take t
                           on (s.sno = t.sno)
  where takes_either_or_both = 1
;The following looked reasonable but, unfortunately unsuccessful:
  Window functions not allowed here (in Having Clause)
select max(s.sno),
       max(s.sname),
       max(s.age),
       sum(case when t.cno in ('CS112', 'CS114')
                then 1
                else 0
           end)
         over (partition by s.sno) as takes_either_or_both
  from student s join take t
                   on (s.sno = t.sno)
group by s.sno
having sum(case when t.cno in ('CS112', 'CS114')
                then 1
                else 0
           end)
         over (partition by s.sno) = 1
Invalid identifier in Having clause
select s.sno,
       s.sname,
       s.age,
       sum(case when t.cno in ('CS112', 'CS114')
                then 1
                else 0
           end)
         over (partition by s.sno) as takes_either_or_both
  from student s join take t
                   on (s.sno = t.sno)
group by s.sno, s.sname, s.age
having takes_either_or_both = 1
;I have searched for a document that completely defines the sequence in which the clauses are executed. I have found tidbits here and there but not something complete. I realize that my running into problems like this one is due to my lack of understanding of the sequence and the scope of the clauses that make up a statement. Because of this, I cannot even tell if it is possible to write the above query using a single select statement. Pardon my bit of frustration...
Thank you for your help,
John.
DDL follows.
        /* drop any preexisting tables */
        drop table student;
        drop table courses;
        drop table take;
        /* table of students */
        create table student
        ( sno integer,
          sname varchar(10),
          age integer
        /* table of courses */
        create table courses
        ( cno varchar(5),
          title varchar(10),
          credits integer
        /* table of students and the courses they take */
        create table take
        ( sno integer,
          cno varchar(5)
        insert into student values (1,'AARON',20);
        insert into student values (2,'CHUCK',21);
        insert into student values (3,'DOUG',20);
        insert into student values (4,'MAGGIE',19);
        insert into student values (5,'STEVE',22);
        insert into student values (6,'JING',18);
        insert into student values (7,'BRIAN',21);
        insert into student values (8,'KAY',20);
        insert into student values (9,'GILLIAN',20);
        insert into student values (10,'CHAD',21);
        insert into courses values ('CS112','PHYSICS',4);
        insert into courses values ('CS113','CALCULUS',4);
        insert into courses values ('CS114','HISTORY',4);
        insert into take values (1,'CS112');
        insert into take values (1,'CS113');
        insert into take values (1,'CS114');
        insert into take values (2,'CS112');
        insert into take values (3,'CS112');
        insert into take values (3,'CS114');
        insert into take values (4,'CS112');
        insert into take values (4,'CS113');
        insert into take values (5,'CS113');
        insert into take values (6,'CS113');
        insert into take values (6,'CS114');

Answer 1

Hi, John,
Just use the aggregate SUM function.
        select s.sno,
               s.sname,
               s.age,
               sum(case when t.cno in ('CS112', 'CS114')
                        then 1
                        else 0
                   end) as takes_either_or_both
          from student s join take t
                           on (s.sno = t.sno)
       GROUP BY  s.sno,
                s.sname,
              s.age
       HAVING  sum(case when t.cno in ('CS112', 'CS114')
                        then 1
                        else 0
                   end)  = 1;Analytic functions are computed after the WHERE- and HAVING clause have been applied. To use the results of an analytic fucntion in a WHERE- or HAVING clause, you have to compute it in a sub-query, and then you can use it in a WHERE- or HAVING clause of a super-query.