How to Delete the records based upon Prev and Next rows and assign the date based upon certain conditions

Question

This is my insert Statements for the Source data.

REM INSERTING into EXPORT_TABLE  
SET DEFINE OFF;  
Insert into EXPORT_TABLE   values (\'4VKMH\',\'GUIDFREE         


        

Answer 1

The reason why you got an error on your query is that before the sub-query is defined, you must indicate what you want to select from it. So if you had prefixed that with select * from it would have been a valid query.

Note that you don't have to do those or operations, as you can do that shorter with an in operator.

You should also negate some comparisons (as you already have NOT) and truncate dates with TRUNC.

Here is the query I would suggest:

SELECT      TEMP.REG_ID, 
            TEMP.EVENT_TYPE,
            TEMP.EVENT_DATE,
            TEMP.PRODUCT_CD,
            TEMP.TERM_START_DATE,
            CASE WHEN TEMP.EVENT_TYPE IN ('NEW SUBSCRIPTION', 'RENEWAL', 'UPSELL') 
                  AND TEMP.NEXT_EVENT_TYPE = 'CANCELLATION' THEN
                        LEAST(TEMP.TERM_END_DATE, TEMP.NEXT_TERM_END_DATE)
                 ELSE TEMP.TERM_END_DATE
            END AS TERM_END_DATE,
            TEMP.DAYS,
            TEMP.AMT
FROM    (SELECT     REG_ID, 
                    EVENT_TYPE,
                    EVENT_DATE,
                    PRODUCT_CD,
                    TERM_START_DATE,
                    TERM_END_DATE,
                    DAYS,
                    AMT,
                    LAG(EVENT_TYPE, 1, '-') over (
                        PARTITION BY REG_ID, PRODUCT_CD
                        ORDER BY EVENT_DATE, TERM_START_DATE) as PREV_EVENT_TYPE,
                    LAG(EVENT_DATE, 1) over (
                        PARTITION BY REG_ID, PRODUCT_CD
                        ORDER BY EVENT_DATE, TERM_START_DATE) as PREV_EVENT_DATE,
                    LEAD(EVENT_TYPE, 1, '-') over (
                        PARTITION BY REG_ID, PRODUCT_CD
                        ORDER BY EVENT_DATE, TERM_START_DATE) as NEXT_EVENT_TYPE,
                    LEAD(EVENT_DATE, 1) over (
                        PARTITION BY REG_ID, PRODUCT_CD
                        ORDER BY EVENT_DATE, TERM_START_DATE) as NEXT_EVENT_DATE,  
                    LEAD(TERM_END_DATE, 1) over (
                        PARTITION BY REG_ID, PRODUCT_CD
                        ORDER BY EVENT_DATE, TERM_START_DATE) as NEXT_TERM_END_DATE
            FROM    export_table) TEMP
WHERE   NOT (TEMP.EVENT_TYPE = 'CANCELLATION' 
             AND TEMP.PREV_EVENT_TYPE IN ('NEW SUBSCRIPTION', 'RENEWAL', 'UPSELL') 
             AND TRUNC(TEMP.EVENT_DATE) = TRUNC(TEMP.PREV_EVENT_DATE))
AND     NOT (TEMP.NEXT_EVENT_TYPE = 'CANCELLATION'
             AND TEMP.EVENT_TYPE IN ('NEW SUBSCRIPTION', 'RENEWAL', 'UPSELL') 
             AND TRUNC(TEMP.NEXT_EVENT_DATE) = TRUNC(TEMP.EVENT_DATE))

Note that the term_end_date of record 6 is also modified, as rule 2 is applicable to it.

Answer 2

Let me state up front that I didn't test this in Oracle as I don't have an Oracle database handy.

I pared it down to a single join, it may be useful to compare performance with the accepted answer.

select  
    e1.reg_id,
    e1.product_cd,
    e1.event_type, 
    e1.event_date, 
    e1.term_start_date,
    case e1.event_type when 'CANCELLATION' then e1.term_end_date else coalesce(e2.term_end_date, e1.term_end_date) end as term_end_date,
    e1.days, 
    e1.amt
from event e1 
    left outer join event e2 on 
        e1.reg_id = e2.reg_id and 
        e1.product_cd = e2.product_cd and 
        e1.term_start_date = e2.term_start_date and 
        (e1.event_type = 'CANCELLATION' or e2.event_type = 'CANCELLATION') and 
        e1.event_date <> e2.event_date  
where trunc(e1.event_date) <> trunc(e2.event_date) or e2.reg_id is null