SqlQuery Help [date range]

Question

Database: Oracle 10g
Table: SOURCE_TABLE
#of records: 500k
Primary Key: ID, EFF_DATE
Data/Table:
with SOURCE_TABLE AS
     (select 'P1' AS ID, TO_DATE('03/01/2012','MM/DD/YYYY') AS EFF_DATE, 'V1' AS VALUE1, 'V2' AS VALUE2 FROM DUAL
      UNION ALL
      select 'P1' AS ID, TO_DATE('02/01/2012','MM/DD/YYYY') AS EFF_DATE, 'V3' AS VALUE1, 'V24' AS VALUE2 FROM DUAL
      UNION ALL
      select 'P2' AS ID, TO_DATE('01/01/2012','MM/DD/YYYY') AS EFF_DATE, 'V10' AS VALUE1, 'V20' AS VALUE2 FROM DUAL
      UNION ALL
      select 'P2' AS ID, TO_DATE('03/01/2012','MM/DD/YYYY') AS EFF_DATE, 'V11' AS VALUE1, 'V21' AS VALUE2 FROM DUAL)                                 
SELECT * FROM SOURCE_TABLE  Conditions:
For a given date range i.e. 01/01/2012 to 04/30/2012
what we need is: For each ID, we need part "P" value1 and value2 column values for each specific month between the default date ranges.
In this case it would be Jan-2012, FEB-2012, MAR-2012, APR-2012
If you see below output for P1, there will not be JAN-2012 record as the part P1 is effective from FEB-2012
What i have done so far:
i have the select query basically it is looping through 01/2012 to 04/2012 and for every month,
find out the max eff_date record less then the month that we are computing the value columns for
Output:
with output as
     (select 'P1' AS ID , '02/01/2012' AS AS_OF_DATE, 'V3' AS VALUE1, 'V24' AS VALUE2 FROM DUAL
      UNION ALL
      select 'P1' AS ID , '03/01/2012' AS AS_OF_DATE, 'V1' AS VALUE1, 'V2' AS VALUE2 FROM DUAL
      UNION ALL
      select 'P1' AS ID , '04/01/2012' AS AS_OF_DATE, 'V1' AS VALUE1, 'V2' AS VALUE2 FROM DUAL
      UNION ALL
      select 'P2' AS ID , '01/01/2012' AS AS_OF_DATE, 'V10' AS VALUE1, 'V20' AS VALUE2 FROM DUAL
      UNION ALL
      select 'P2' AS ID , '02/01/2012' AS AS_OF_DATE, 'V10' AS VALUE1, 'V20' AS VALUE2 FROM DUAL
      UNION ALL
      select 'P2' AS ID , '03/01/2012' AS AS_OF_DATE, 'V11' AS VALUE1, 'V21' AS VALUE2 FROM DUAL
      UNION ALL
      select 'P2' AS ID , '04/01/2012' AS AS_OF_DATE, 'V11' AS VALUE1, 'V21' AS VALUE2 FROM DUAL
SELECT * FROM OUTPUT;
        ID     AS_OF_DATE     VALUE1     VALUE2
     P1     02/01/2012     V3     V24
     P1     03/01/2012     V1     V2
     P1     04/01/2012     V1     V2
     P2     01/01/2012     V10     V20
     P2     02/01/2012     V10     V20
     P2     03/01/2012     V11     V21
     P2     04/01/2012     V11     V21Thanks in advance

Answer 1

Hi,
ora1001 wrote:
Hi frank,
I changed the query to inner join instead of outer join because if a part is not effective for a given month, then i dont need a record.
ex: if P1 EFF_DATE IS 03/01/2012, then i dont need records for JAN, FEB... You're right; inner join is all you need here. My mistake.
What is "P1 EFF_DATE"? Is that the start of the date range you're interested in (January 1, 2012 in your original example)?
Don't you still need to cosider rows before that date? For example, say we want to show the same 4 months (January to April, 2012), but instead of
select 'P2' AS ID, TO_DATE('01/01/2012','MM/DD/YYYY') AS EFF_DATE, 'V10' AS VALUE1, 'V20' AS VALUE2 FROM DUALthe data is changed to
select 'P2' AS ID, TO_DATE('11/01/2011','MM/DD/YYYY') AS EFF_DATE, 'V10' AS VALUE1, 'V20' AS VALUE2 FROM DUALWould you still want the same results? In this example, the values from November 2011 serve as the values for January and Febrruay 2012. It doesn't matter that Noverber 2011 is outside of the target range.
It works fine... its faster now.
however, i would like to know the difference in my query vs your query in terms of performance..
the subquery runs faster when compared to analytic function...
select * from part p,
source_table st
where p.id = st.id
and st.eff_date = (select max(eff_date) from source_table st1
where st1.eff_date < '01-mar-2012');
If st1.eff_date is a DATE, then don't try to compare it to a VARCHAR2, such as '01-mar-2012'. Use TO_DATE:
where st1.eff_date < TO_DATE ( '01-mar-2012'
                               , 'DD-mon-YYYY'
                    )or a DATE literal:
where st1.eff_date < DATE '2012-03-01'instead.
vs
select * from (
select p.id,
st.eff_date,
row_number() over (partition by st.id,st.eff_date order by st.id,st.eff_date desc) as consider_order from part p,
source_table st
where p.id = st.id) where consider_order = 1;
The first approach is doing much less work. It just has to remember what the latest eff_date is. Say it has already found a row with March 1, 2012, and it comes across a row with an earlier date, such as January 1, 2012. As soon as it sees that January 1, 2012 is not the new record holder, it can move on to the next row. (Using the LAST function, as I suggested, would cause it to do even less work.)
The second approach is doing much more work. It has to remember all the dates, in order. Say it has already found a row with March 1, 2012, and it comes across a row with an earlier date, such as January 1, 2012. After it sees that January 1, 2012 is not the new record holder, it still has to determine if January 1 comes 2nd, or 3rd on the list.