Question

Given the following table and data.

 create table prices
(productKey int
,PriceType char(10)
,BeginDate date
,EndDate date
,price decimal(18,2))

insert into prices(productKey, PriceType,BeginDate,EndDate, price)
values
(1,'LIST','1-1-2010','1-15-2010',10),
(1,'LIST','1-16-2010','10-15-2010',20),
(1,'DISCOUNT','1-10-2010','1-15-2010',-5),
(2,'LIST','2-1-2010','10-15-2010',30),
(2,'LIST','10-16-2010','1-1-9999',35),
(2,'DISCOUNT','2-10-2010','10-25-2010',-10),
(2,'LIST','1-1-2010','1-15-2010',10),
(3,'DISCOUNT','1-12-2010','1-1-9999',-5),
(3,'LIST','1-16-2010','1-1-9999',10)

I need to insert records into that same table that calculates the actual price (list-discount) for each time period.

e.g. for product 1, I should have the following “ACTUAL” records

Begin     End      Price
1-1-2010  1-9-2010 10
1-10-2010  1-15-2010 5
1-16-2010  10-15-2010 20

I kind of have it figured out for anything where a discount starts within a list price span, but I’m at a loss for anything else.

Thanks for the help

EDIT

There can be multiple discounts per ProductKey, but the discount periods won’t overlap. So you could have one for 2010, and another one for 2012, but not 2 for 2010.

Also, if someone can come up with a better title, please do so. My poor brain is completely challenged at this point.

EDIT2

It’s SQL server 2008R2. I’d love a beautiful set based answer (or someone that gives me a start in that direction), but will be just as happy with a cursor solution that works.

Answer 1

Clever puzzle.

You need to reconstruct all the time spans. To do this, I take out all the dates from the price ranges and reconstruct the possible date ranges.

with alldates as (select d.*, ROW_NUMBER() over (partition by productkey order by thedate) as seqnum
                  from ((select productkey, BeginDate as thedate from prices)
                        union all
                        (select productkey, enddate as thedate from prices)
                       ) d
                 ),
     datepair as (select d1.productkey, d1.thedate as BeginDate, d2.thedate as EndDate
                  from alldates d1 left outer join 
                       alldates d2
                       on d1.seqnum = d2.seqnum - 1 and d1.productKey = d2.productKey
                 )
select dp.productkey, dp.BeginDate, dp.EndDate, SUM(p.price)
from datepair dp join
     prices p
     on dp.productkey = p.productkey and
         dp.BeginDate >= p.BeginDate and
        dp.EndDate <= p.EndDate
group by dp.productkey, dp.BeginDate, dp.EndDate
order by 1, 2, 3

I’ve thought about this some more. The basic idea above is correct. The basic idea is to break up the time dimension into intervals where the list and discount is constant over the entire interval. The question is how to create these intervals, which are in the datepairs alias.

These intervals have just a few rules:

• A datepair interval can begin when any time period begins.
• A datepair interval can begin one day after any time period ends.
• A datepair interval can end when any time period ends
• A datepair interval can end one day before any time period begins

Once we have the intervals, it is a simple matter to join in the appropriate list price and discounts for that period. The following query uses this logic:

with begindates as (select distinct productKey, thedate
                    from ((select productkey, BeginDate as thedate from prices)
                          union all
                          (select productkey, dateadd(d, 1, enddate) as thedate from prices)
                         ) d
                     ),
     enddates as (select distinct productKey, thedate
                  from ((select productkey, DATEADD(d, -1, begindate) as thedate from prices)
                        union all
                        (select productkey, enddate as thedate from prices)
                       ) d
                 ),
     datepair as (select *
                  from (select d1.productkey, d1.thedate as BeginDate,
                               MIN(d2.thedate) as EndDate
                        from begindates d1 left outer join 
                             enddates d2
                             on d1.productKey = d2.productKey and d1.thedate < d2.thedate
                        group by d1.productkey, d1.thedate
                       ) t
                  where BeginDate <> EndDate
                 )
select dp.productkey, dp.BeginDate, dp.EndDate, SUM(p.price)
from datepair dp join
     prices p
     on dp.productkey = p.productkey and
         dp.BeginDate >= p.BeginDate and
        dp.EndDate <= p.EndDate
group by dp.productkey, dp.BeginDate, dp.EndDate
order by 1, 2, 3