SQL Query Execution using parameter - sql

Question

I am having problems figuring out why a query is taking drastically longer to run based on me swapping out a parameter against a real value.
DECLARE #quarter int
DECLARE #year int
DECLARE #countOfUnitsBought int
set #year = 2009
set #quarter = 1
set #countOfUnitsBought = 4;
with res
as
(
select
o.account_id
--,orderyear
--,orderquarter
from
fmtables.[dbo].[orders] o
--cross apply(values(year(o.[ship_date]))) as a1(orderyear)
--cross apply(values(DatePart(quarter,(o.[ship_date])))) as a2(orderquarter)
where
ship_date = (select min(ship_date) from fmtables.[dbo].[orders] mo where [account_id] = o.account_id) and
total_value > 0 AND
order_status NOT LIKE 'return%' AND
order_status NOT LIKE 'cancel%' AND
order_status NOT LIKE 'freeze%' and
CAST(DatePart(quarter,(o.[ship_date])) as int) = #quarter and
year(o.[ship_date]) = #year and
(select sum(quantity) from fmtables..[orders] ox inner join fmtables..[orderlines] olx on ox.order_id = olx.order_id
where olx.order_id = o.order_id and [product_code] in(select [product_code] from fmtables..[products] where [category_code] in('1','2','3','4'))) >= #countOfUnitsBought
)
select * from res;
This query takes 43 seconds to run.
Now if I simply replace the #quarter and change to a literal
CAST(DatePart(quarter,(o.[ship_date])) as int) = 1 and
it now takes 1 second.
Can anyone please give me a clue as to why and if I need to change some casting to help.
Thanks
Scott
EDIT:
So I manage to get it whizzing through with the help of everyone's comments.
I used a mix of passing the parameters from the inputs and then through to a 'local' variable inside the procedure.
alter procedure [dbo].[Lifetime_HeadsetUnits]
#inquarter int , #inyear int, #incountOfUnitsBought int
as
DECLARE #quarter int
DECLARE #year int
declare #countOfUnitsBought int
select #quarter = #inquarter
select #year = #inyear
select #countOfUnitsBought = #incountOfUnitsBought
and also the
OPTION (OPTIMIZE FOR(#quarter = 1))
as part of the final output query.

Answer 1

Try this instead. I rewrote the datepart so indexes can be, used and the database is not making the long calculation on all rows. In other words I made your datecalculation sargable:
DECLARE #quarter int
DECLARE #year int
DECLARE #countOfUnitsBought int
set #year = 2009
set #quarter = 1
declare #from datetime = dateadd(quarter, #quarter - 1, cast(#year as char(4)))
set #countOfUnitsBought = 4;
with res
as
(
select
o.account_id
from
fmtables.[dbo].[orders] o
where
ship_date =
(select min(ship_date)
from fmtables.[dbo].[orders] mo
where [account_id] = o.account_id) and
total_value > 0 AND
order_status NOT LIKE 'return%' AND
order_status NOT LIKE 'cancel%' AND
order_status NOT LIKE 'freeze%' and
o.[ship_date] >= #quarter and
o.[ship_date] < DATEADD(QUARTER, 1, #from) and
(select sum(quantity) from fmtables..[orders] ox
inner join fmtables..[orderlines] olx on ox.order_id = olx.order_id
where [product_code] in(select [product_code] from fmtables..[products]
where [category_code] in('1','2','3','4'))) >= #countOfUnitsBought
)
select * from res;
Are you running an sql-server 2008 ? There is a bug that could also explain your performance issues.