Group by and record count simplification, try #2

Question

I know what I’m trying to do, not sure how to get there. I am providing the make table and inserts along with the query that doesn’t work.
/*table*/
CREATE TABLE claims_step_table ( 
table_id       INTEGER PRIMARY KEY,
record_id      NUMBER, 
step_code      VARCHAR2(20 Byte),
date_completed DATE,
step_date      DATE
/*inserts*/
INSERT INTO claims_step_table VALUES (1,  123456,'P96','10-JAN-09','05-JAN-09');
INSERT INTO claims_step_table VALUES (2,  123456,'L61',null,'05-JAN-09');
INSERT INTO claims_step_table VALUES (3,  67,'P96','10-MAR-08','05-MAR-08');
INSERT INTO claims_step_table VALUES (4,  67,'L61',null,'05-MAR-08');
INSERT INTO claims_step_table VALUES (5,  67,'U18',null,'05-MAR-08');
INSERT INTO claims_step_table VALUES (6,  67,'429',null,'05-MAR-08');
INSERT INTO claims_step_table VALUES (7,  9876,'L61',null,'05-FEB-08');
INSERT INTO claims_step_table VALUES (8,  9876,'429',null,'05-FEB-08');
INSERT INTO claims_step_table VALUES (9,  123456,'P96','10-AUG-07','05-JUN-07');
INSERT INTO claims_step_table VALUES (10, 123456,'L61',null,'05-JUN-07');
INSERT INTO claims_step_table VALUES (11, 123456,'U18',null,'05-JUN-07');
INSERT INTO claims_step_table VALUES (12, 123456,'429',null,'05-JUN-07');
INSERT INTO claims_step_table VALUES (13, 67555,'L61',null,'07-NOV-08');
INSERT INTO claims_step_table VALUES (14, 67555,'P96','08-NOV-08','07-NOV-08');
INSERT INTO claims_step_table VALUES (15, 67555,'429','15-NOV-08','07-NOV-08');
COMMIT;
/*query*/
SELECT
    RECORD_ID,
    max(STEP_DATE)
    from(
            select
            RECORD_ID,
            STEP_DATE,
            case
                when STEP_CODE = 'L61' then DATE_COMPLETED
            end as STEP_L61,
            case
                when STEP_CODE = 'P96' then DATE_COMPLETED
            end as STEP_P96,
            case
                when STEP_CODE = 'U18' then DATE_COMPLETED
            end as STEP_U18,
            case
                when STEP_CODE = '429' then DATE_COMPLETED
            end as STEP_429
        from
        CLAIMS_STEP_TABLE
        where
            STEP_CODE in('L61','P96','U18','429')
where
    STEP_L61 is null
and
    STEP_P96 is not null
and
    STEP_U18 is null
and
    STEP_429 is null
group by RECORD_ID, STEP_DATE
order by RECORD_ID, STEP_DATE DESCWhat I need is all the RECORD_ID’s and STEP_DATE’s where
•     STEP_CODE’s L61, U18 and 429 are null AND STEP_CODE P96 is completed
•     Only the newest records based on the STEP_DATE(so the max step_date), if multiple RECORD_ID's.
The query should return *2* tuples
RECORD_ID       MAX(STEP_DATE)
67             03/05/2008
123456             01/05/2009I currently get the following
RECORD_ID       MAX(STEP_DATE)
67             03/05/2008
67555             11/07/2008 <= step 429 not null
123456             01/05/2009
123456              06/05/2007 <= old record need newestThis is a much more simplified query than the original as there are many table joins. There are 100k records + in the real CLAIMS_STEP_TABLE with over 1k different STEP_CODE’s. (this is dummy table for this example). BTW on the real DB i only have read access.
Thanks for your patience and understanding with a noob, hope this attempt is clearer.

Answer 1

The reason it's failing is because your pivot (inner query) is returning sparse results:
RECORD_ID STEP_DATE STEP_L61  STEP_P96  STEP_U18  STEP_429
    123456 05-JAN-09           10-JAN-09
    123456 05-JAN-09
        67 05-MAR-08           10-MAR-08
        67 05-MAR-08
        67 05-MAR-08
        67 05-MAR-08
      9876 05-FEB-08
      9876 05-FEB-08
    123456 05-JUN-07           10-AUG-07
    123456 05-JUN-07
    123456 05-JUN-07
    123456 05-JUN-07
     67555 07-NOV-08
     67555 07-NOV-08           08-NOV-08
     67555 07-NOV-08                               15-NOV-08So in the case of 67555 there is a situation where all the conditions of the outer query are met. What you need to do is remove the extra NULLs like so:
SELECT  RECORD_ID
,       STEP_DATE
,       MAX(CASE
                WHEN STEP_CODE = 'L61' THEN DATE_COMPLETED
        END) AS STEP_L61,
        MAX(CASE
                WHEN STEP_CODE = 'P96' THEN DATE_COMPLETED
        END) AS STEP_P96,
        MAX(CASE
                WHEN STEP_CODE = 'U18' THEN DATE_COMPLETED
        END) AS STEP_U18,
        MAX(CASE
                WHEN STEP_CODE = '429' THEN DATE_COMPLETED
        END) AS STEP_429
FROM    CLAIMS_STEP_TABLE
WHERE   STEP_CODE IN ('L61','P96','U18','429')
GROUP BY RECORD_ID
,       STEP_DATETo get this result:
RECORD_ID STEP_DATE STEP_L61  STEP_P96  STEP_U18  STEP_429
     67555 07-NOV-08           08-NOV-08           15-NOV-08
      9876 05-FEB-08
    123456 05-JUN-07           10-AUG-07
        67 05-MAR-08           10-MAR-08
    123456 05-JAN-09           10-JAN-09Then your whole query returns the following:
SQL> SELECT  RECORD_ID
  2  ,       MAX(STEP_DATE)
  3  FROM
  4  (
  5          SELECT  RECORD_ID
  6          ,       STEP_DATE
  7          ,       MAX(CASE
  8                          WHEN STEP_CODE = 'L61' THEN DATE_COMPLETED
  9                  END) AS STEP_L61,
10                  MAX(CASE
11                          WHEN STEP_CODE = 'P96' THEN DATE_COMPLETED
12                  END) AS STEP_P96,
13                  MAX(CASE
14                          WHEN STEP_CODE = 'U18' THEN DATE_COMPLETED
15                  END) AS STEP_U18,
16                  MAX(CASE
17                          WHEN STEP_CODE = '429' THEN DATE_COMPLETED
18                  END) AS STEP_429
19          FROM    CLAIMS_STEP_TABLE
20          WHERE   STEP_CODE IN ('L61','P96','U18','429')
21          GROUP BY RECORD_ID
22          ,       STEP_DATE
23  )
24  WHERE   STEP_L61 IS NULL
25  AND     STEP_P96 IS NOT NULL
26  AND     STEP_U18 IS NULL
27  AND     STEP_429 IS NULL
28  GROUP BY RECORD_ID
29  /
RECORD_ID MAX(STEP_
    123456 05-JAN-09
        67 05-MAR-08
SQL>HTH!