From: "Lin Kun Hsin" <s9154083@cyut.edu.tw>
To: "pgsql-sql" <pgsql-sql@postgresql.org>
Subject: Re: A SQL Question About distinct, limit, group by, having, aggregate
Date: Thu, 31 Mar 2005 20:24:03 +0800
Message-Id: <20050331121514.M9051@cyut.edu.tw>
MIME-Version: 1.0
Content-Type: text/plain;
	charset=big5

Actually, i have a method to solve this problem.
But i really want to know, we have to write more statement to do one thing?


First step, we have to create 2 sequence. Let call them "foo" and "foo1".

create sequence foo;
create sequence foo1;

then, you can run below statement, and you will see the result that is we want.


select setval('foo',1);
select setval('foo1',1);
select id, score, class from (
		
		select id, score, ph1.class , sequence1, sequence2, CASE WHEN ph1.cc > 5
THEN ph3.sequence + 4 
		ELSE ph3.sequence + ph1.cc - 1 END as tail from 
			
			(select class, count(*) as cc from allscore group by class) 
		
		as ph1 join 	
			
			(select id, class, score, nextval('foo') as sequence1, currval('foo') as
sequence2 from (select * from allscore order by class, score desc) as t2)
		
		as ph2 on (ph1.class = ph2.class) join 
		
			(select distinct on (class) class, nextval('foo1'), currval('foo1') as
sequence from (
			select  id , score , class from allscore order by class, score desc) as t6)
			
		as ph3 on (ph2.class = ph3.class) order by ph1.class , score desc 
	
) as con where sequence2 <= tail;

--
http://alumni.cyut.edu.tw
Open WebMail Project (http://openwebmail.org)