Need suggestions on date range query

Question

I have a requirement to show the amount of product remaining. There is a table that holds updated "inventory" amounts with a date and tonnage, and a series of transactional tables that detail the individual disbursements from the stockpile. The trick is that the dates for the inventory adjustments may not all be the same, meaning that I need to individually resolve the stockpiles.
This query will give me the inventory disbursements:
select FN_STN_KEY(j.FACTORY_ID, j.STATION_ID) as STATION,
  count(j.LOAD_JOB_ID) as LOADS,
           CASE SUM(w.SPOT_WEIGHT)
             WHEN 0 THEN SUM(NVL(j.MAN_SPOT_WT,0))
             ELSE SUM(w.SPOT_WEIGHT)
           END TONS
       from TC c, TC_LOAD_JOBS j, SPOT_WEIGHTS w
      where c.TC_ID = j.TC_ID
        and c.DATE_INDEX = w.DATE_INDEX and j.LOAD_RATE_ID = w.LOAD_RATE_ID
        and c.DATE_INDEX BETWEEN to_date('09/01/2009','MM/DD/YYYY') and sysdate
        and FN_STN_KEY(j.FACTORY_ID, j.STATION_ID) in (810,410)
      group by FN_STN_KEY(j.FACTORY_ID, j.STATION_ID);Note that the date and the list of stations in the where clause are dynamic and selected by user. If this was only one station at a time it wouldn't be this complicated.
This query will give me the last known inventory amount:
select to_char(MAX(AS_OF_DT),'Mon DD, YYYY'), TONS
  from STATION_LOG
      where AS_OF_DT < sysdate and STN_KEY in (810,410) group by TONS;Again, the date and list of stations are selected by user. They should be identical to those selected for the other query.
Does anyone have any good ideas on how to combine these two statements into a single report?
Note: FN_STN_KEY acts as a join function. You don't really want me to get into why there isn't a single unique key to reference.

Answer 1

Hi,
I'm trying to follow your descrioption, but lots of things don't make sense to me.
blarman74 wrote:
Yeah. I put in some data so I could get the message back to you, then filled in the rest.
So the user is going to pass in two parameters: The date of the report and the list of stations they want to get an inventory count on. What were the parameters that produced the output you posted before:
STATION     INITIAL_TONS     USED_TONS     AS_OF_DATE
810               835500        465100      09/01/2010
410               495800        366900      09/02/2010
550               568900        122600      08/31/2010
What I need the report to do is
1) take a station from the list
2) find out the inventory tally from STATION_LOG where the date is the largest date less than the supplied date. This should give me AS_OF_DATE and my initial quantity.
3) query the data table for all tons hauled from the AS_OF_DATE for that station.
4) repeat for the next station.So this is what your existing PL/SQL code does. A non procedural language, like SQL, won't follow the same steps, of course.
The sample data for station_log is:
INSERT INTO STATION_LOG (1, to_date('08/31/2010','MM/DD/YYYY'), 810, 562500);
INSERT INTO STATION_LOG (2, to_date('09/02/2010','MM/DD/YYYY'), 410, 495500);
INSERT INTO STATION_LOG (3, to_date('09/01/2010','MM/DD/YYYY'), 910, 832600);
INSERT INTO STATION_LOG (4, to_date('12/31/2010','MM/DD/YYYY'), 810, 239800);How do you get the initial_tons in the output above from the data above? Did you mean to post some new sample data for station_log?
I still get ORA-00928 errors from all the INPUT statements.
As I said, I can do it inside a loop in PL/SQL, but I got completely stumped on how I could accomplish this in SQL. The trick is that if I can do it in SQL, I can allow the user to export the data to csv using built-in functionality. If I have to do it in PL/SQL, I can't provide the export as easily.
One more thing I just thought about, I am going to need to use a BETWEEN on the dates of the data I need to grab. I obviously don't want to grab data past another inventory tally record from STATION_LOG for the same station, and I can use an NVL so it cuts off at SYSDATE. I obviously haven't hauled anything in the future ;)I doubt if I'll get enough information to do this for you before I leave on vacation.
Here's an example of what you need to do using the scott.emp table instead of your station_log table:
SELECT       job
,       hiredate
,       sal
FROM       scott.emp
ORDER BY  job
,            hiredate
JOB       HIREDATE           SAL
ANALYST   03-Dec-1981       3000
ANALYST   19-Apr-1987       3000
CLERK     17-Dec-1980        800
CLERK     03-Dec-1981        950
CLERK     23-Jan-1982       1300
CLERK     23-May-1987       1100
MANAGER   02-Apr-1981       2975
MANAGER   01-May-1981       2850
MANAGER   09-Jun-1981       2450
PRESIDENT 17-Nov-1981       5000
SALESMAN  20-Feb-1981       1600
SALESMAN  22-Feb-1981       1250
SALESMAN  08-Sep-1981       1500
SALESMAN  28-Sep-1981       1250Say we want to find, for each job in a given list, the sal that corresponds to the last hiredate that is no later than the given report_date. (This seems pretty close to what you heed: fior each stn_key in a given list, the quantity that corresponds to the last row that is no later than the given report_date.)
That is, if we're only interested in the jobs CLERK, MANAGER and PRESIDENT, and only in hiredates on or before December 31, 1981, the output would be:
JOB       LAST_HIREDA   LAST_SAL
CLERK     03-Dec-1981        950
MANAGER   09-Jun-1981       2450
PRESIDENT 17-Nov-1981       5000That is, we want to ignore all jobs that are not in the given list, and all rows whose hiredate is after the given report_date. Among the rows that remain, we're interested only in the last one for each job.
Note that the last_sal for CLERK is not 1300 or 1100: those values were after the given report_date. also, the last_sal for CLERK is not 800; that's not the last one of the remaining rows.
Here's one way to get those results in pure SQL:
DEFINE       jobs_wanted     = "CLERK,MANAGER,PRESIDENT"
DEFINE       report_date     = "DATE '1981-12-31'"
SELECT       job
,       MAX (hiredate)                         AS last_hiredate
,       MAX (sal) KEEP (DENSE_RANK LAST ORDER BY hiredate)     AS last_sal
FROM       scott.emp
WHERE       hiredate     <= &report_date
AND        ',' || '&jobs_wanted' || ','     LIKE
       '%,' || job                  || ',%'
GROUP BY  job
ORDER BY  job
;I used substitution variables for the parameters. You could use bind_variables, or hard-code the values instead.
The WHERE clause is applied before aggreate functions are computed, so rows after &report_date don't matter.
"MAX (sal) KEEP (DENSE_RANK LAST ORDER BY hiredate)" means the sal that is associated with the last row, in order by hiredate. If there happens to be a tie (that is, two or more rows have exactly the same hiredate, and no row is later) then the highest sal from thsoe rows is returned; that's what MAX means here. Ties may be impossible in your data.
You need to write a similar query using your station_log table, and join the results of that to your load_data table, including only the rows that have dates between the date in the sub-query (last_hiredate in my example) and the parameter report_date. That can be part of the join condition.