SQL Query - Max Date - Group By

Question

Please see SQL below - this code is counting customers and grouping by their tenure since activation
However the table FACT_TRANS I am linking to contains multiple transactions for different dates and I only want to select the latest date.
How can I return only the latest transaction from the FACT_TRANS table as the code below (In Bold)is not returning the correct number of records
SELECT TO_NUMBER(MONTHS_BETWEEN(To_DATE(a.DATE_KEY, 'yyyymmdd' ),
TO_DATE(ADD_MONTHS((LAST_DAY(c.DATE_JOIN)+1),-1),'dd-mm-yyyy')) -24000, '999') AS TENURE,
SUM(DECODE(A.RESULT_MNTHM12_VALUE,0,0,1)) AS M12,
SUM(DECODE(A.RESULT_MNTHM11_VALUE,0,0,1)) AS M11,
SUM(DECODE(A.RESULT_MNTHM10_VALUE,0,0,1)) AS M10,
SUM(DECODE(A.RESULT_MNTHM9_VALUE,0,0,1)) AS M9,
SUM(DECODE(A.RESULT_MNTHM8_VALUE,0,0,1)) AS M8,
SUM(DECODE(A.RESULT_MNTHM7_VALUE,0,0,1)) AS M7,
SUM(DECODE(A.RESULT_MNTHM6_VALUE,0,0,1)) AS M6,
SUM(DECODE(A.RESULT_MNTHM5_VALUE,0,0,1)) AS M5,
SUM(DECODE(A.RESULT_MNTHM4_VALUE,0,0,1)) AS M4,
SUM(DECODE(A.RESULT_MNTHM3_VALUE,0,0,1)) AS M3,
SUM(DECODE(A.RESULT_MNTHM2_VALUE,0,0,1)) AS M2,
SUM(DECODE(A.RESULT_MNTHM1_VALUE,0,0,1)) AS M1,
SUM(DECODE(A.CURRENT_RESULT_VALUE,0,0,1)) AS M0
FROM FACT_RESULT_VW a,
PLAN_VW b,
SUB_VW c ,
CUST_VW d,
FACT_TRANS e,
WHERE e.DATE_TRANS IN (SELECT MAX (FACT_TRANS.DATE_TRANS) FROM FACT_TRANS )+
AND a.PP_KEY = b.PP_key
AND a.ID = c.ID
AND c.CUST_ID = d.CUST_ID
AND c.ID = E.ID
AND c.DATE_JOIN >= To_DATE('01-09-2002 00:00:00', 'dd-mm-yyyy hh24:mi:ss' )
AND c.DATE_JOIN < To_DATE('01-07-2010 00:00:00', 'dd-mm-yyyy hh24:mi:ss' )
AND a.DATE_KEY = 20100601
AND A.PEF_ID = 1
GROUP BY TO_NUMBER(MONTHS_BETWEEN(To_DATE(a.DATE_KEY, 'yyyymmdd' ),
     TO_DATE(ADD_MONTHS((LAST_DAY(c.DATE_JOIN)+1),-1),'dd-mm-yyyy')) -24000, '999')

Answer 1

Hi,
Whenever you have a question, post a little sample data (CREATE TABLE and INSERT statements) and the results you want from that data.
Simplify the problem as much as possible.
Would you have the same problem if only two tables, fact_result_vw and fact_trans, were involved? Then post a question involving only those two tables. Once you have a solution for the simplified problem, adding the other tables will be easy.
Would the problem be essentially the same if you were pivoting only 2 or 3 columns, instead of 13? Then post a question where you're only pivoting 2 or 3 columns. Once you have a solution, adding the others will be easy.
787771 wrote:
Please see SQL below - this code is counting customers and grouping by their tenure since activation
However the table FACT_TRANS I am linking to contains multiple transactions for different dates and I only want to select the latest date.
How can I return only the latest transaction from the FACT_TRANS table as the code below (In Bold)is not returning the correct number of recordsDo you want only the latest transaction from the fact_trans table, or the latest transaction for each customer ?
What if there's a tie, and the same customer has two (or more) transactions at the same date, and none later?
Since you're not using any of the data from the fact_trans table, the effect is just to eliminate customers who are not in fact_trans. You can do that more simply with an IN or EXISTS sub-query.
If you really do need something from the fact_trans table in the main table, then one way you can get the results you want is to use the analytic RANK (or ROW_NUMBER) function to number the rows in fact_trans by trans_date (latest first), with a separate set of numbers (1, 2, 3, ...) for each customer, like this:
WITH     fact_trans_1     AS
     SELECT     id     -- all columns that you need
     ,     RANK () OVER ( PARTITION BY  id
                      ORDER BY          date_trans     DESC
                    )     AS r_num
     FROM     fact_trans
SELECT        TO_NUMBER(MONTHS_BETWEEN(To_DATE(a.DATE_KEY, 'yyyymmdd' ),
       TO_DATE(ADD_MONTHS((LAST_DAY(c.DATE_JOIN)+1),-1),'dd-mm-yyyy')) -24000, '999') AS TENURE,
       SUM(DECODE(A.RESULT_MNTHM12_VALUE,0,0,1)) AS M12,
       SUM(DECODE(A.RESULT_MNTHM11_VALUE,0,0,1)) AS M11,
       SUM(DECODE(A.RESULT_MNTHM10_VALUE,0,0,1)) AS M10,
       SUM(DECODE(A.RESULT_MNTHM9_VALUE,0,0,1)) AS M9,
       SUM(DECODE(A.RESULT_MNTHM8_VALUE,0,0,1)) AS M8,
       SUM(DECODE(A.RESULT_MNTHM7_VALUE,0,0,1)) AS M7,
       SUM(DECODE(A.RESULT_MNTHM6_VALUE,0,0,1)) AS M6,
       SUM(DECODE(A.RESULT_MNTHM5_VALUE,0,0,1)) AS M5,
       SUM(DECODE(A.RESULT_MNTHM4_VALUE,0,0,1)) AS M4,
       SUM(DECODE(A.RESULT_MNTHM3_VALUE,0,0,1)) AS M3,
       SUM(DECODE(A.RESULT_MNTHM2_VALUE,0,0,1)) AS M2,
       SUM(DECODE(A.RESULT_MNTHM1_VALUE,0,0,1)) AS M1,
       SUM(DECODE(A.CURRENT_RESULT_VALUE,0,0,1)) AS M0
FROM        FACT_RESULT_VW      a,
       PLAN_VW           b,
       SUB_VW           c,
       CUST_VW           d,
       FACT_TRANS_1          e,
WHERE        e.DATE_TRANS IN (SELECT MAX (FACT_TRANS.DATE_TRANS) FROM FACT_TRANS )
AND       e.r_num     = 1
AND        a.PP_KEY      = b.PP_key
AND        a.ID           = c.ID
AND       c.CUST_ID      = d.CUST_ID
AND       c.ID           = E.ID
AND       c.DATE_JOIN       >= To_DATE('01-09-2002 00:00:00', 'dd-mm-yyyy hh24:mi:ss' )
AND       c.DATE_JOIN      <  To_DATE('01-07-2010 00:00:00', 'dd-mm-yyyy hh24:mi:ss' )
AND       a.DATE_KEY = 20100601
AND       A.PEF_ID = 1
GROUP BY  TO_NUMBER(MONTHS_BETWEEN(To_DATE(a.DATE_KEY, 'yyyymmdd' ),
       TO_DATE(ADD_MONTHS((LAST_DAY(c.DATE_JOIN)+1),-1),'dd-mm-yyyy')) -24000, '999')I hope this answers your question.
If not, post some sample data (CREATE TABLE and INSERT statements) and the results you want from that data.