My table is like
RUNDATE TRANSACTION_ID CUSTOMER_ID TRANSACTION_TYPE 18/10/2020 2464 4007715 T9
I am trying to get customer_ids where they didn’t do any transactions previous month and made transactions with at least 2 transaction_type at this_month.While doing this I am converting rundate to MM-YYYY format. What I’ve done for getting previous month and this month on MM-YYY format is :
SELECT EXTRACT( MONTH FROM RUNDATE )|| '-' || EXTRACT(YEAR FROM RUNDATE ) AS MONTH_YEAR, TRANSACTION_TYPE, CUSTOMER_ID, TRANSACTION_ID, EXTRACT( MONTH FROM PREV_MONTH)|| '-' || EXTRACT(YEAR FROM PREV_MONTH) AS PREV_MONTH_YEAR FROM ( SELECT TRANSACTION_ID,RUNDATE,CUSTOMER_ID,TRANSACTION_TYPE, ADD_MONTHS(rundate ,-1) as PREV_MONTH FROM TRANSACTIONS )
and this is what i got:
PREV_MONTH_YEAR MONTH_YEAR TRANSACTION_TYPE CUSTOMER_ID 9-2020 10-2020 T9 4007715
But the problem is I couldn’t figure how to check my conditions for both PREV_MONT_YEAR AND MONTH_YEAR at the same time
Advertisement
Answer
You can use window functions: You can aggregate by month and use window functions:
SELECT t.* FROM (SELECT t.CUSTOMER_ID, TRUNC(RUNDATE, 'MON') as RUN_MONTH, COUNT(*) as cnt_this_month, LAG(TRUNC(RUNDATE, 'MON')) OVER (PARTITION BY CUSTOMER_ID) as PREV_RUNMONTH FROM TRANSACTIONS t GROUP BY CUSTOMER_ID, TRUNC(RUNDATE, 'MON') ) t WHERE cnt_this_month >= 2 and (prev_runmonth is null or prev_runmonth < run_month - interval '1' month);
Note: This truncates the date to the first day of the month. I prefer working with dates rather than strings. Of course, you can use to_char(runmonth, 'YYYY-MM')
if you prefer strings.