Question

Two tables:
a) svn3 with columns name revno compname
b) archdetails with columns name version compname inFlat

Shouldn’t these 2 queries output the same result. The output of query 2 is correct. What is wrong with query1?

query1:
select sum(CASE WHEN inFlat=1 THEN 1 ELSE 0 END) from archdetails inner join svn3 on archdetails.compname=svn3.compname where archdetails.name='ant' AND version='1.4' AND (revno='r274642' OR revno='r274578' OR revno='r274533')

query2:
select sum(CASE WHEN inFlat=1 THEN 1 ELSE 0 END) from archdetails where name='ant' AND version='1.4' AND compname IN (select compname from svn3 where revno ='r274642' OR revno='r274578' OR revno='r274533')

Answer 1

If svn3 has more than one row with the same compname that also have one of the requested revno values, then query 1 will count each svn3 row once. Query 2 will only count each compname once. The IN clause is eliminating any duplicate compname entries in the second query.

If you don’t want to count duplicate compname entries, then you need to group them in a subquery. One way to do that is using the IN clause like you did in query 2. Another option would be to use GROUP BY on a subquery, and name that subquery. Then join the named subquery to the archdetails table.

If you’re trying to solve a performance problem, then I would suggest adding primary keys and foreign keys to the tables. I think that would speed things up.

I’ve seen some comments that EXISTS is sometimes faster than IN, but I haven’t seen any evidence to back that up. You might want to experiment with that.