Sum Over Time

Question

Hi,
I'm trying to do something which I would guess is quite a common query, but after scratching my head and perusing the web I am still no closer to a solution.
I am running:
Oracle Database 10g Enterprise Edition Release 10.2.0.4.0 - 64bi
I'm looking to sum up a set of values, taking into account both a parent grouping and start and end dates.
For the parent grouping I am using:
+SUM([value]) over (Partition by [Parent] order by [Parent],[Child])+
And I was hoping to be able to extend this SUM to also handle the start and end dates, so the final output would contain a sum of the values for each different time period.
As an example, using the data below I'm trying to sum up the price of the components of a car over time:
row, product, component, rate, start date, end date
1, car, chassis, 180, 01/01/2000, 31/12/2009
2, car, chassis, 200, 01/01/2010, 01/01/2050
3, car, engine, 100, 01/01/2000, 01/01/2050
Notice there is a change of price for Component 'chassis', so the output I'm looking for is:
row, product, component, rate, start date, end date, sum
1, car, chassis, 180, 01/01/2000, 31/12/2009, 280
2, car, engine, 100, 01/01/2000, 31/12/2009, 280
3, car, chassis, 200, 01/01/2010, 01/01/2050, 300
4, car, engine, 100, 01/01/2010, 01/01/2050, 300
But in reality all I need is:
row, product, start date, end date, sum
1, car, 01/01/2000, 31/12/2009, 280
2, car, 01/01/2010, 01/01/2050, 300
Preferably the query would be in a view rather than a stored procedure, and it needs to be able to handle many 'products', 'components' and start/end dates.
All help most appreciated, and if any more info is required, please let me know.
Thanks,
Julian

Answer 1

Hi Frank,
Thanks for picking up this query, I'll try to explain my points in more detail:
+SUM([value]) over (Partition by [Parent] order by [Parent],[Child])+I don't see columns called value, parent or child in the sample data below.
Is value the same as rate? What are parent and child? In the example:
Product is the parent
Component is the child
Rate is the value
Whenever you have a problem, post CREATE TABLE and INSERT statments for your sample data.CREATE TABLE "REPOSITORY"."PRODUCT_RATES"
(     "PRODUCT" VARCHAR2(255 BYTE),
     "COMPONENT" VARCHAR2(255 BYTE),
     "RATE" NUMBER(9,2),
     "START_DATE" DATE,
     "END_DATE" DATE
) PCTFREE 10 PCTUSED 40 INITRANS 1 MAXTRANS 255 NOCOMPRESS LOGGING
STORAGE(INITIAL 65536 NEXT 1048576 MINEXTENTS 1 MAXEXTENTS 2147483645
PCTINCREASE 0 FREELISTS 1 FREELIST GROUPS 1 BUFFER_POOL DEFAULT)
TABLESPACE "SHOP_AREA" ;
insert into REPOSITORY.PRODUCT_RATES (PRODUCT, COMPONENT, RATE, START_DATE, END_DATE) values ('car', 'chassis', 180, to_date('01-01-2000','dd-mm-yyyy'), to_date('31-12-2009','dd-mm-yyyy'))
insert into REPOSITORY.PRODUCT_RATES (PRODUCT, COMPONENT, RATE, START_DATE, END_DATE) values ('car', 'chassis', 200, to_date('01-01-2010','dd-mm-yyyy'), to_date('01-01-2050','dd-mm-yyyy'))
insert into REPOSITORY.PRODUCT_RATES (PRODUCT, COMPONENT, RATE, START_DATE, END_DATE) values ('car', 'engine', 100, to_date('01-01-2000','dd-mm-yyyy'), to_date('01-01-2050','dd-mm-yyyy'))
Although the above short scenario highlights my issue, to expand on the example data set:
insert into REPOSITORY.PRODUCT_RATES (PRODUCT, COMPONENT, RATE, START_DATE, END_DATE) values ('family', 'wife', 500, to_date('01-01-2000','dd-mm-yyyy'), to_date('31-12-2001','dd-mm-yyyy'))
insert into REPOSITORY.PRODUCT_RATES (PRODUCT, COMPONENT, RATE, START_DATE, END_DATE) values ('family', 'wife', 999, to_date('01-01-2002','dd-mm-yyyy'), to_date('01-01-2050','dd-mm-yyyy'))
insert into REPOSITORY.PRODUCT_RATES (PRODUCT, COMPONENT, RATE, START_DATE, END_DATE) values ('family', 'baby', 250, to_date('01-01-2000','dd-mm-yyyy'), to_date('31-12-2004','dd-mm-yyyy'))
insert into REPOSITORY.PRODUCT_RATES (PRODUCT, COMPONENT, RATE, START_DATE, END_DATE) values ('family', 'baby', 500, to_date('01-01-2005','dd-mm-yyyy'), to_date('01-01-2050','dd-mm-yyyy'))
Notice there is a change of price for Component 'chassis', so the output I'm looking for is:
row, product, component, rate, start date, end date, sum
1, car, chassis, 180, 01/01/2000, 31/12/2009, 280
2, car, engine, 100, 01/01/2000, 31/12/2009, 280
3, car, chassis, 200, 01/01/2010, 01/01/2050, 300
4, car, engine, 100, 01/01/2010, 01/01/2050, 300Explain how you get 4 rows of output when the table contains only 3 rows. Are you saying that, because some row has end_date=31/12/2009, then any other row that includes that date has to be split into two, with one row ending on 31/12/2009 and the other one beginning on the next day?
Explain, step by step, how you get the values in the desired output, especially the last column.
But in reality all I need is:Sorry, I can;'t understand what you want.
Are you saying that the output above sould be acceptable, but the output below would be even better?
row, product, start date, end date, sum
1, car, 01/01/2000, 31/12/2009, 280
2, car, 01/01/2010, 01/01/2050, 300
Preferably the query would be in a view rather than a stored procedure, and it needs to be able to handle many 'products', 'components' and start/end dates.Include a couple of differtent products in your sample data and results.
I'm not sure what you want, but there's nothing in what you've said so far that makes me think a stored procedure would be needed.The only output I actually require is:
row, product, component, rate, start date, end date, sum
1, car, 01/01/2000, 31/12/2009, 280
2, car, 01/01/2010, 01/01/2050, 300and with the extended data set:
3, family, 750, 01/01/2000, 31/12/2001
4, family, 1249, 01/01/2002, 31/12/2004
5, family, 1499, 01/01/2005, 31/12/2050however, I was thinking that the data set would need to be somehow expanded to get to the above end result, hence why I included the 'middle step' of:
row, product, component, rate, start date, end date, sum
1, car, chassis, 180, 01/01/2000, 31/12/2009, 280
2, car, engine, 100, 01/01/2000, 31/12/2009, 280
3, car, chassis, 200, 01/01/2010, 01/01/2050, 300
4, car, engine, 100, 01/01/2010, 01/01/2050, 300however, this may be irrelevent.
By the way, there's no point in using the same expression in both the PARTITON BY and ORDER BY clauses of the same analytic function call. For example, if you "PARTITION BY parent", then, when "ORDER BY parent, child" is evaluated, rows will only be compared to other rows with the same parent, so they'll all tie for first place in "ORDER BY parent". OK, thanks.
So far I have got to:
select
sum(rate) over (partition by product) as sum,
a.*
from product_rates a
which results in:
SUM     PRODUCT     COMPONENT     RATE     START_DATE     END_DATE
480     car     engine     100     2000-01-01 00:00:00.0     2050-01-01 00:00:00.0
480     car     chassis     200     2010-01-01 00:00:00.0     2050-01-01 00:00:00.0
480     car     chassis     180     2000-01-01 00:00:00.0     2009-12-31 00:00:00.0
2249     family     baby     250     2000-01-01 00:00:00.0     2004-12-31 00:00:00.0
2249     family     wife     999     2002-01-01 00:00:00.0     2050-01-01 00:00:00.0
2249     family     baby     500     2005-01-01 00:00:00.0     2050-01-01 00:00:00.0
2249     family     wife     500     2000-01-01 00:00:00.0     2001-12-31 00:00:00.0
but this shows that all price variations for a component over time are being summed (e.g. car enging 100 + car chassis 200 + car chassis 180 = 480)
Hope that goes someway expaling my query better.
Also, quick query to improve my postings - how do i indent without making text ittallic, and how do you make code a different font?
Thanks again.
Julian