Take previous results and add to row

Question

I have the below table results.

Criteria:

Partition by CUSTID and ITEMID Where end_dt is not null If the RESP column is null then take the latest value that has an entry.

CUSTID    ITEMID   START_DT    END_DT      RESP  START_CYCLE   END_CYCLE
  1        101     1/1/2019    4/1/2019    400   1/1/2019      1/12/2019
  1        101     1/1/2019    4/1/2019          1/13/2019     1/18/2019
  1        101     1/1/2019    4/1/2019    750   1/19/2019     2/15/2019
  1        101     1/1/2019    4/1/2019          2/16/2019     4/1/2019
  2        909     3/1/2019                444   3/1/2019      3/2/2019
  2        909     3/1/2019                      3/3/2019      3/10/2019
  2        909     3/1/2019                767   3/11/2019     3/28/2019
  2        909     3/1/2019                      3/29/2019     12/31/3000

Expected Results:

CUSTID    ITEMID   START_DT    END_DT      RESP  START_CYCLE   END_CYCLE
  1        101     1/1/2019    4/1/2019    400   1/1/2019      1/12/2019
  1        101     1/1/2019    4/1/2019          1/13/2019     1/18/2019
  1        101     1/1/2019    4/1/2019    750   1/19/2019     2/15/2019
  1        101     1/1/2019    4/1/2019    750   2/16/2019     4/1/2019
  2        909     3/1/2019                444   3/1/2019      3/2/2019
  2        909     3/1/2019                      3/3/2019      3/10/2019
  2        909     3/1/2019                767   3/11/2019     3/28/2019
  2        909     3/1/2019                      3/29/2019     12/31/3000

The only row that is changing is

     1        101     1/1/2019    4/1/2019    750   2/16/2019     4/1/2019

This row should not change, this is correct:

 1        101     1/1/2019    4/1/2019          1/13/2019     1/18/2019

score 1 · Accepted Answer · answered Apr 17 '19 at 13:24

You have to check three things: resp is null, end_dt is not null and if this is the last row for this custid, itemid. Only in this case use last_value, like here, column resp2:

select custid, itemid, start_dt, end_dt, resp, start_cycle, end_cycle, 
       case when resp is null 
                 and end_dt is not null 
                 and lead(itemid) over (partition by custid, itemid order by start_cycle) is null 
            then last_value(resp) ignore nulls 
                 over (partition by custid, itemid order by start_cycle) 
            else resp 
       end resp2
  from t

Take previous results and add to row

1 Answers1