Where (Col1, Col2) in (select...

I wrote a SQL that would work with Oracle, but it is a speciality because in SQL Server there is a syntax error.
I need to retrieve all pairs of test_number, id_numeric where some of them are meeting a certain condition.

If I write:

select test_number, id_numeric from measures where ...condition...

I don't have every row having the same test_number, id_numeric pair. In Oracle I would write:

select * from measures
where  (test_number, id_numeric) in
(select test_number, id_numeric from measures where ...condition...

How would you write this SQL?
Thanks.
LVL 9
ornicarAsked:
Who is Participating?
 
Guy Hengel [angelIII / a3]Connect With a Mentor Billing EngineerCommented:
select measures.*
from measures
JOIN (select test_number ,  id_numeric from measures where ...condition... ) as CL
ON CL.test_number = measures.test_number
AND CL.id_numeric = measures.id_numberic

CHeers

0
 
ornicarAuthor Commented:
This one would work, obviously,

select * from measures
where  test_number + ' ' + id_numeric in
(select test_number + ' ' + id_numeric from measures where ...condition...

but as there are more than 6 moi records in this table, I am afraid that it would take ages as there is no index used.

0
 
BillAn1Connect With a Mentor Commented:
try this :

select * from measures m1
where  exists
(select 1 from measures m2 where m1.test_number = m2.test_number and m1.id_numeric = m2.id_numeric and ...condition...)

0
Get expert help—faster!

Need expert help—fast? Use the Help Bell for personalized assistance getting answers to your important questions.

 
BillAn1Commented:
The only problem with your answer angelIII is that if there is more tha 1 record in the table with the same values for test_number and measures_id you will return n*n records
e.g. if there are2 you will return 4, if 3 then 9 etc
0
 
Jan FranekCommented:
Slightly modified angelIII's :

select M1.*
from measures M1
JOIN measures M2
ON M1.test_number = M2.test_number
AND M2.id_numeric = M2.id_numberic
where ...condition...

in ...condition... use M2. prefix
0
 
ornicarAuthor Commented:
Thanks. I tried both of your suggestions and effectively, AngelIII's returns more rows. But the advantage is that it is the fastest. I hesitate between the two. I need this SQL to update some flags in the rows and perhaps I'll choose the fastest.
0
 
Guy Hengel [angelIII / a3]Billing EngineerCommented:
BillAn1, you are basically correct, but the same "problem" is in oracle, and subject to another question/thread :-)
Anyway, the field names suggest that the values will be unique...
CHeers

0
 
Guy Hengel [angelIII / a3]Billing EngineerCommented:
If you use it for updating, it shall be no problem, because if you get "duplicate" rows, in fact they are the same underlying rows. You might try to add DISTINCT to you query, but when doing update it doesn't matter
Cheers
0
 
ornicarAuthor Commented:
I choose angelIII's as the answer because it is the fastest and for my update I don't need the exact amount of rows. BillAn1's can be considered as more correct as it returns the rows only once.
Thanks for your help.
0
 
BillAn1Commented:
No,the oracle version ornicar provided, or the version I provided, won't give any duplicates, because it is not a join of the table back on itself, it is a "select * from mytable WHERE .....", so you won't get any duplicate rows. If you join the table to itself on a non-unique col, you get duplicates. Even if it is unique, it is less efficient to do the join.
0
 
Guy Hengel [angelIII / a3]Billing EngineerCommented:
BillAn1, I see the difference in the results, indeed the join can result in duplicated rows.
However, your suggestion also performs a join, which basically could be resolved by the optimizer to give the same/similar execution plan. IMHO
CHeers
0
 
BillAn1Commented:
In practical terms, yes the optimiser will often end up choosing a similar execution plan - particularly, if the data is unique, there will probably be no performance gain, True :-)
0
Question has a verified solution.

Are you are experiencing a similar issue? Get a personalized answer when you ask a related question.

Have a better answer? Share it in a comment.

All Courses

From novice to tech pro — start learning today.