SQL query join help

Question

I have two tables DEVICE and CONTACT. The tables are joined DEVICE_ID to CONTACT_ID. I want to print device_ID and the corresponding contact info with the lowest CONTACT.PRIORITY.
SQL> describe DEVICE;
DEVICE_ID              NOT NULL VARCHAR2(50)
SQL> describe CONTACT;
CONTACT_ID               NOT NULL VARCHAR2(50)
CONTACT_TYPE             NOT NULL VARCHAR2(4)
PRIORITY                 NOT NULL DOUBLE PRECISION
LASTNAME                 NOT NULL VARCHAR2(30)
FIRSTNAME                NOT NULL VARCHAR2(80)
SQL> SELECT a.DEVICE_ID, b.LASTNAME,b.PRIORITY from DEVICE a, CONTACT b where a.DEVICE_ID = 'DEVICEA' and a.DEVICE_ID=b.CONTACT_ID(+);
DEVICE_ID              LASTNAME    FIRSTNAME PRIORITY CONTACT_TYPE
DEVICEA                CONTACT1     GN               1 ROUT
DEVICEA                CONTACT2     Hans             2 ROUTHow do I print just one record with the lowest CONTACT.PRIORITY?
Thanks
Ravi

Answer 1

rmalghan wrote:
Thanks Frank, Sundar and Sean. I got the below 3 to work. Two questions. My DEVICE table has 31936 records.
1. The first one returned 30207, 2nd 31936 records and 3rd one 30348. Since the 2nd one result looks like what I expected, the 1st and 3rd one seem to be missing some records. For troubleshooting, I tried to limit the query for a few devices, but can't seem to figure why the difference. Any suggestions? The 2nd query is not checking for CONTACT_TYPE = 'ROUT', so it's not surprising that it returns more rows.
The 3rd query is effectively doing an inner join. The condition "WHERE c.rnum = 1" is being applied after the join, so it rejects all the rows that were added because of the outer join (that is, where all the columns from c are NULL). You probably meant:
LEFT OUTER JOIN got_rnum c ON  d.DEVICE_ID     = c.CONTACT_ID
                           AND c.rnum          = 1; without a WHERE clause.
2. Is there benefit in terms of efficiency. The 2nd one does seem to run faster so I am guessing that is the best. Any comments?
SELECT a.DEVICE_ID, b.LASTNAME,b.FIRSTNAME,b.PRIORITY
FROM DEVICE a, CONTACT b
where
a.DEVICE_ID=b.CONTACT_ID(+) and
b.CONTACT_TYPE = 'ROUT' and
b.PRIORITY = (select min(b1.PRIORITY ) from CONTACT b1 where b1.CONTACT_ID = b.CONTACT_ID);
30207 rows selected (10.12 seconds)
SELECT DEVICE_ID, LASTNAME, PRIORITY
FROM (
SELECT a.DEVICE_ID, b.LASTNAME, b.PRIORITY, ROW_NUMBER() OVER (PARTITION BY a.DEVICE_ID ORDER BY b.PRIORITY) row_num
from DEVICE a, CONTACT b
where a.DEVICE_ID=b.CONTACT_ID(+))
WHERE row_num = 1;
31936 rows selected (6.68 seconds)
WITH got_rnum AS (
SELECT LASTNAME, FIRSTNAME, PRIORITY, CONTACT_TYPE, CONTACT_ID,
ROW_NUMBER () OVER ( PARTITION BY  CONTACT_ID ORDER BY priority) AS rnum
FROM contact
where CONTACT_TYPE = 'ROUT'
SELECT d.DEVICE_ID, c.LASTNAME, c.FIRSTNAME, c.PRIORITY
FROM device d
JOIN got_rnum c ON d.DEVICE_ID     = c.CONTACT_ID(+)
WHERE c.rnum = 1;
30348 rows selected (9.07 seconds)
Timing that way isn't very accurate. Whichever one you run first is apt to be slower, because the data is less likely to be cached in memory.
The 2nd query has only 3 columns in the result set. That probably doesn't change the speed much, but you should fix it anyway.