Using Dense Rank

Question

Hello,
I am using ORACLE 10g and having an error when updating a table and generating ORA-30926 'Unable to get a stable set of rows in the source tables'.
I create a temporary working table with records pulled from 2 different tables. The query to make the temp table uses
the DENSE RANK FUNCTION to eliminate any duplicates. But, I am still getting duplicate records.
When it is time to update my primary table with the records from the temp table, it sees the duplicates and
generates the error. I have all the code to create the problem. I think I almost have it but maybe need different query structure.Any help is appreciated.
-- Create table
create table CUSTOMER
EMP_ID NUMBER(4) ,
CUST_ID NUMBER(9) not null,
CONTACT_CD          VARCHAR2(1),
PURCHASE_DATE DATE,
NEW_CUST_CD          VARCHAR2(1),
CUST_NAME          VARCHAR2(25)
-- Create table
create table DEFERRED_MAILING
CUST_ID NUMBER(9) not null,
EMP_ID NUMBER(4) not null,
PURCHASE_DATE          DATE
insert into customer ( emp_id, cust_id, contact_cd,PURCHASE_DATE, NEW_CUST_CD,cust_name)
values (10,444,'Y',to_date('11-01-2011 16:11:05', 'dd-mm-yyyy hh24:mi:ss'),'Y','J.J.BROWN');
insert into customer ( emp_id, cust_id,contact_cd, PURCHASE_DATE, NEW_CUST_CD,cust_name)
values (10,444,'Y',to_date('11-01-2011 16:11:05', 'dd-mm-yyyy hh24:mi:ss'),'Y','J.BROWN');
insert into deferred_mailing (cust_id, emp_id, purchase_date)
values (444, 10, to_date('11-01-2011 16:11:05', 'dd-mm-yyyy hh24:mi:ss'));
create table UPDTE_DEFERRED_MAILING_RECORDS nologging as
SELECT a.CUST_ID,
a.EMP_ID,
a.PURCHASE_DATE,
a.DM_ROW_ID,
a.drank,
c.CONTACT_CD,
c.NEW_CUST_CD/*,
TRANS_ID_SEQ.NEXTVAL as TRANS_ID*/
FROM (SELECT a.ROWID AS DM_ROW_ID,
a.CUST_ID,
a.EMP_ID,
a.PURCHASE_DATE,
dense_rank() over(PARTITION BY a.CUST_ID, a.EMP_ID ORDER
BY a.PURCHASE_DATE DESC, a.ROWID) DRANK
FROM deferred_mailing a) a,
customer c
WHERE a.CUST_ID = c.CUST_ID
AND a.EMP_ID = c.EMP_ID
AND (a.PURCHASE_DATE <= c.PURCHASE_DATE OR
c.PURCHASE_DATE IS NULL);
SELECT * FROM UPDTE_DEFERRED_MAILING_RECORDS;
}

Answer 1

Please try this
create table UPDTE_DEFERRED_MAILING_RECORDS nologging as
SELECT distinct a.CUST_ID,
a.EMP_ID,
a.PURCHASE_DATE,
a.drank,
c.CONTACT_CD,
c.NEW_CUST_CD,
a.DM_ROW_ID
FROM (SELECT a.ROWID AS DM_ROW_ID,
a.CUST_ID,
a.EMP_ID,
a.PURCHASE_DATE,
dense_rank() over(PARTITION BY a.CUST_ID, a.EMP_ID ORDER
BY a.PURCHASE_DATE DESC, a.ROWID) DRANK
FROM deferred_mailing a) a,
customer c
WHERE a.CUST_ID = c.CUST_ID
AND a.EMP_ID = c.EMP_ID
AND (a.PURCHASE_DATE <= c.PURCHASE_DATE OR
c.PURCHASE_DATE IS NULL)
and a.drank=1;
The query you've posted is behaving according to the expectations. The inner select is returning one row and the outer is returning two as the
WHERE a.CUST_ID = c.CUST_ID
AND a.EMP_ID = c.EMP_ID
AND (a.PURCHASE_DATE <= c.PURCHASE_DATE OR
c.PURCHASE_DATE IS NULL)
conditions are seeing two rows in the table customer.
I've added the a.drank=1 clause to skip the duplicates from the inner table and distinct in the final result to remove duplicates from the overall query result.
For eg, if you have one more row in the deferred_mailing like this
SQL> select * from DEFERRED_MAILING;
CUST_ID EMP_ID PURCHASE_
444 10 11-JAN-11
444 10 11-JAN-11
then the query without "a.drank=1" will return 4 rows like this by the outer query.
CUST_ID EMP_ID PURCHASE_ DM_ROW_ID DRANK C N
444 10 11-JAN-11 AAATi2AAGAAAACcAAB 2 Y Y
444 10 11-JAN-11 AAATi2AAGAAAACcAAA 1 Y Y
444 10 11-JAN-11 AAATi2AAGAAAACcAAB 2 Y Y
444 10 11-JAN-11 AAATi2AAGAAAACcAAA 1 Y Y
It'll return the below even if we use distinct on the same query(i.e. without a.drank=1)
CUST_ID EMP_ID PURCHASE_ DM_ROW_ID DRANK C N
444 10 11-JAN-11 AAATi2AAGAAAACcAAB 2 Y Y
444 10 11-JAN-11 AAATi2AAGAAAACcAAA 1 Y Y
which contains duplicates again.
So, we need a combination of distinct and dense here.
btw, Please mark the thread as 'answered', if you feel you got your question answered. This will save the time of others who search for open questions to answer.
Regards,
CSM