Get rows where the last row finish off

Question

Hi, i have two tables AND would LIKE TO get data BY combining both.
here IS my data
WITH hist AS
  SELECT To_Date('4/23/2010','mm/dd/yyyy') dt, 999 alias, 'PROC' dom FROM dual UNION ALL
  SELECT To_Date('4/27/2010','mm/dd/yyyy') dt, 999 alias, 'LON' dom FROM dual UNION all
  SELECT To_Date('4/1/2010','mm/dd/yyyy') dt, 111 alias, 'SOC' dom FROM dual UNION all
  SELECT To_Date('4/10/2010','mm/dd/yyyy') dt, 111 alias, 'NAO' dom FROM dual UNION ALL
  SELECT To_Date('3/23/2010','mm/dd/yyyy') dt, 222 alias, 'PSE' dom FROM dual
final AS
  SELECT To_Date('2/26/2010','mm/dd/yyyy') dt, 999 alias FROM dual UNION ALL
  SELECT To_Date('4/22/2010','mm/dd/yyyy') dt, 999 alias FROM dual UNION all
  SELECT To_Date('4/26/2010','mm/dd/yyyy') dt, 999 alias FROM dual UNION ALL
  SELECT To_Date('4/30/2010','mm/dd/yyyy') dt, 999 alias FROM dual UNION ALL
  SELECT To_Date('2/25/2010','mm/dd/yyyy') dt, 111 alias FROM dual UNION ALL
  SELECT To_Date('2/26/2010','mm/dd/yyyy') dt, 222 alias FROM dual UNION ALL
  SELECT To_Date('4/22/2010','mm/dd/yyyy') dt, 222 alias FROM dual UNION all
  SELECT To_Date('4/26/2010','mm/dd/yyyy') dt, 222 alias FROM dual
the output should be as follow(without the extra blank line of course)
DT           ALIAS   DOM
2/26/2010     999     PROC
4/22/2010     999     PROC
4/26/2010     999     LON
4/30/2010     999     LON
4/27/2010     999     LON
4/23/2010     999     PROC
2/25/2010     111     SOC
4/1/2010     222     SOC
4/10/2010     222     NAO
2/26/2010     222     PSE
4/22/2010     222     PSE
4/26/2010     222     PSEso what i am doing here is as follow, take one row in hist table (4/23) and join with final table and give me all rows in final table
where dt <= to the row in hist table and include the row from hist table.
this logic will give me rows 2/26/2010,4/22/10 4/23/2010
then the second row in hist table (4/27/2010) wiill get all rows
in final table that is <= to the current row and pick up the rows starting from the row > than the last row where the 4/23/2010 finished off
in this case the output will be 4/26/10, 4/27/2010(we need to include row from hist)
since there is no row in hist that is greater than 4/30/2010, this date will still be display and dom column value should be taking from the max date in hist
which is 4/27/2010. see output above
this sound a little confusing to explain but look at output of what to expect as output. the other ids should follow the same logic
can someone help write a query for this? thanks

Answer 1

Hi,
Devx wrote:
Frank, thanks again, i ran the query in oracle 11g and oracle 9i. 11g runs ok but 9i doesnt. it looks like the ignore null option is not supported in 9i. That's right: IGNORE NULLS was new in Oracle 10. You should always mention your Oracle version whenever you ask a quiestion, especially if it's as old as Oracle 9.
i will be running this query in 9i. is there any alternative to re-write this query without using last value since ignore null is not supported and the output is not as i expected when i take that keyword out.
i really appreciate your help. please let me know how would i re-write the query. thanksOne work-around is to use LEAD or LAG instead of LAST_VALUE. This means you have to know exactly where (how many rows away) the most recent non-NULL value is, which in turn requires other analytuic funtions, such as ROW_NUMBER, and more sub-queries:
WITH     combined_tables        AS
     SELECT     dt, alias, NVL (dom, '_?_') AS dom     FROM hist
     UNION
     SELECT     dt, alias, NULL          AS dom      FROM final
,     got_r_num     AS
     SELECT     dt, alias, dom
     ,     ROW_NUMBER () OVER ( PARTITION BY  alias
                               ORDER BY          dt
                       )               AS r_num
     ,     COUNT (*)     OVER ( PARTITION BY  alias
                       )               AS alias_cnt
     FROM    combined_tables
,     got_skip_cnts     AS
     SELECT     dt, alias, dom, r_num
     ,     r_num - MAX (CASE WHEN dom IS NOT NULL THEN r_num END)
                             OVER ( PARTITION BY  alias
                           ORDER BY          r_num
                         )                    AS skip_before
     ,     MIN (CASE WHEN dom IS NOT NULL THEN r_num END)
                             OVER ( PARTITION BY  alias
                           ORDER BY          r_num     DESC
                         ) - r_num               AS skip_after
     FROM    got_r_num
,     got_next_dom     AS
     SELECT     dt, alias, dom, r_num, skip_before
     ,     LEAD (dom, skip_after) OVER ( PARTITION BY  alias
                                            ORDER BY      r_num
                             ) AS next_dom
     FROM    got_skip_cnts
SELECT       dt
,       alias
,       NULLIF ( COALESCE ( next_dom
                        , LAG (dom, skip_before) OVER ( PARTITION BY  alias
                                                     ORDER BY         r_num
           )     AS dom
FROM       got_next_dom
ORDER BY  alias
,            dt
;You should be able to calculate bot LEAD and LAG in the same query, but there seems to be a bug that only calculates one of them correctly in this case. The sub-query got_next_dom gets around that, by doing the LEAD in a separate sub-query.