Sql to fetch values for missing records.

Question

Hi All,
I'm working on oracle 10gR2.I'm having trouble writing a query for the below and I know you genious guys can help me.
I have the below table data.
The data is read as....for Name R-VA013SC between from_m 0 and to_m 0.5 the average daily value is 30 and volume is 34 for the year 1929.
similarly next record for Name R-VA013SC between from_m 0 and to_m 0.5 the average daily value is 40 and volume is 43 for the year 2001.
"NAME" | "FROM_M"|"TO_M"|"DATA_DATE"|"ADT""VOLUME"
"R-VA013SC"     | 0     |0.05     | *01-JAN-29......|*     30 |     34     
"R-VA013SC"     | 0     |0.05     | *12-SEP-01..*....| 40 |     43     
"R-VA013SC"     | 0     |0.05     | *14-JAN-08.......|*     20 |     15     
if you notice above, there are data missing for years from 1930 to 2000, the daily average and volume is assumed to be the previously available year, for example for the year 2000, we need to have a record with the data for 1929 with the current date like
"R-VA013SC"     | 0     |0.05     | *01-JAN-00.....|*     30 |     34     
for the year 2005 , the data has to copied from 2001 like .
"R-VA013SC"     | 0     |0.05     | *12-SEP-05*......| 40 |     43
for the year 2009, the value has to be fetched from the previously available year ie 2008 like
"R-VA013SC"     | 0     |0.05     | *14-JAN-09*.......|     20 |     15.
I aslo need to write a query to find the average values between 2 years,
like if someone asks for the average volume adt ADT between 2000 and 2005, I need to find the average of the same
like ... add the averages for 2000-2005/no of years requested(5).
Appreciate all your help!
Thanks in advance
H

Answer 1

Hi,
user32322435 wrote:
Thanks again..
Here is the table data
ID     NAME     FROM_M     TO_M     ADT     VOLUME     DATA_DATE
2055801     SC-723N      0     0.05     30     34     01/01/1929
2055801     SC-723N      0     0.05     40     43     09/12/2001
2055801     SC-723N      0     0.05     20     15     01/14/2008I'm considering the following
start date ='1999-07-01'
end date ='2003-07-01'
end_date-start_date = 1461
So,
days considered in 1999 is 184 *adt for 1999 is 30 = 5520
days considered in 2000 is *366(leap year)* *adt for 2000 is 30 = 10980
days considered in 2001 is 365 *adt for 2001 is 40 = 14600
days considered in 2002 is 365 *adt for 2001is 40 = 14600
days considered in 2003 is 184 *adt for 2003 is 40= 7360January 1 through July 1, 2010 is only 182 days, isn't it? That would make the total for 2003 = 7280 (net loss of 80).
>
5520+10980+14600+14600+7360 = 53060/no of days taken into account (1461)July 1, 1999 through July 1, 2003 inclusive is 1462 days, right? 4 * 365 = 1460, plus one February 29th, plus an extra July 1st.
Expected ADT 53060/1461 =*36.31* With the corrections mentioned above, that's 52980 / 1462 = 36.23 (rounded).
......very sorry about the 38.01 . I missed something earlier.
(select to_date('2003-07-01','YYYY-MM-DD') - to_date('1999-07-01','YYYY-MM-DD') from dual;) = 1461That's the number of days between the two. You have to add 1 if both endpoints are to be included.
1999-07-01 through 1999-07-01 is 1 day, but the difference is 0.
select to_date('1999-12-31','YYYY-MM-DD') - to_date('1999-07-01','YYYY-MM-DD') +1 from dual; = 184Now I understand!
Until this message I had no idea that you wanted data_date rounded down to the beginning of the year, no matter how late in the year it actually is.
In that case, replace data_date in the sub-query with TRUNC (data_date, 'YEAR'), and the previous query will do what you requested:
WITH   params                      AS
     SELECT     DATE '1999-07-01'     AS begin_date
     ,     DATE '2003-07-01' + 1     AS end_date
     FROM dual
,     got_next_data_date     AS
     SELECT  id
--     ,     data_date          -- REPLACED WITH FOLLOWING LINE
     ,     TRUNC (data_date, 'YEAR')     AS data_date
     ,     adt
     ,     LEAD (
              -- data_date     -- REPLACED WITH FOLLOWING LINE
                 TRUNC (data_date, 'YEAR')
               , 1
               , DATE '9999-12-31'
               ) OVER ( PARTITION BY  id
                              ORDER BY      data_date
                       ) AS next_data_date
     FROM    table_test
SELECT       n.id
,       SUM ( n.adt
           * ( LEAST ( next_data_date
               , p.end_date
          - GREATEST ( n.data_date
                    , p.begin_date
           ) / (p.end_date - p.begin_date)      AS avg_adt
FROM       got_next_data_date  n
JOIN       params           p       ON     n.data_date     <= p.end_date
                        AND     p.begin_date     <= n.next_data_date
GROUP BY  id
;Of course, you don't need the lines marked "REPLACED WITH FOLLOWING LINE". I left them there in case you were wondering how the query had changed.
I also used id, instead of name, for GROUPing and PARTIONing, since that seems to reflect your posting. If one is functionally dependent on the other, then it doesn't matter which one you use.
... our expectation for select * from new view should be some thing like ...for those missing year the dates can be defaulted to 01_JAN. the available years are spaced just to differentiate
ID     NAME     FROM_M     TO_M     ADT     VOLUME     DATA_DATE
2055801     SC-723N      0     0.05     40     43     01/01/1999
2055801     SC-723N      0     0.05     40     43     01/01/2000
2055801     SC-723N      0     0.05     40     43     09/12/2001
2055801     SC-723N      0     0.05     20     15     01/01/2002
2055801     SC-723N      0     0.05     20     15     01/01/2003
2055801     SC-723N      0     0.05     20     15     01/01/2004
2055801     SC-723N      0     0.05     20     15     01/01/2005
2055801     SC-723N      0     0.05     20     15     01/01/2004
2055801     SC-723N      0     0.05     20     15     01/01/2005
2055801     SC-723N      0     0.05     20     15     01/01/2006
2055801     SC-723N      0     0.05     20     15     01/01/2007
2055801     SC-723N      0     0.05     20     15     01/14/2008
2055801     SC-723N      0     0.05     20     15     01/01/2009
This does what you requested.
It includes all years from the earliest data_date through either (a) the last data_date, or (b) the current year, whichever is later.
CREATE OR REPLACE VIEW     yearly_view
AS
WITH     got_years_in_effect     AS
     SELECT     id, name, from_m, to_m, adt, volume, data_date
     ,     MONTHS_BETWEEN ( LEAD ( TRUNC (data_date, 'YEAR')
                                 , 1
                          , ADD_MONTHS ( TRUNC (SYSDATE, 'YEAR')
                                       , 12
                          ) OVER ( PARTITION BY  id
                                      ORDER BY          data_date
                    , TRUNC (data_date, 'YEAR')
                    ) / 12          AS years_in_effect
     FROM     table_test
,     cntr     AS
     SELECT     LEVEL     AS n
     FROM     (
               SELECT  MAX (years_in_effect)     AS max_years
               FROM     got_years_in_effect
     CONNECT BY     LEVEL <= max_years
SELECT     y.id, y.name, y.from_m, y.to_m, y.adt, y.volume
,     CASE
          WHEN  c.n = 1
          THEN  y.data_date
          ELSE  ADD_MONTHS ( TRUNC (y.data_date, 'YEAR')
                           , 12 * (c.n - 1)
     END     AS data_date
FROM     got_years_in_effect y
JOIN     cntr              c     ON     c.n     <= y.years_in_effect
;