Declaring variables in CTEs SQL Server 2008 - sql

Question

How can I implement something like the following:
declare #myInt int
set #myInt=(select count(*) from x)
;with x as
(
select row_number() over(partition by c.patientid order by c.admissiondate) as rn
,c.patientid,c.admissiondate
,max(c.claimsfromdate) as maxHemiDate
,min(c.claimsfromdate) minHemiDate
,(
select max(c2.claimsfromdate)
from claims as c2
where c2.patientid=c.patientid
group by c2.patientid
) as maxClaimsDate
,p.drgCode
,datediff(dd,min(c.claimsfromdate),max(c.claimsfromdate)) /7 as weeksWithHemi
from claims as c inner join icdclaims as ci on ci.id=c.id
inner join tblicd as t on t.icd_id=ci.icd_id
inner join patient as p on p.patientid=c.patientid
and p.admissiondate = c.admissiondate
and p.dischargedate = c.dischargedate
where t.icdText like '%X%' and p.statecode='21'
group by c.patientid, c.admissiondate, p.drgCode
)
select p.patientid, count(*)
from patient as p
left join x on x.patientid=p.patientid
where x.patientid is null
group by p.patientid
The error thrown when this is executed is
invalid object name x
I kinda figured that this would happen since the variable declaration is outside of the CTE. If I move the declaration inside of the parentheses of WITH I get another error.
How can I assign a variable like this inside a CTE? Or can you not use a variable that draws data from the CTE at all?

Answer 1

If you are using SQL Server 2005 or later, you can put the total on each row:
select t.*, cnt, count(*) over () as NumRows
from (select p.patientid, count(*) as cnt
from patient as p left join
x
on x.patientid=p.patientid
where x.patientid is null
group by p.patientid
) t
In case you are using the total to calculate percentages, or something like that, it might be convenient to have the value on each row.

Answer 2

To get in #myInt number of rows, you can do:
declare #myInt int
;with x as
(
__your query__
)
select p.patientid, count(*)
from patient as p
left join x on x.patientid=p.patientid
where x.patientid is null
group by p.patientid
set #myInt=##ROWCOUNT
EDITED (Due Jon Egerton comment) If you need to count x rows, then a temporay table is the way:
declare #myInt int
;with x as
(
__your query__
)
select *
into #tmp_x
from x;
set #myInt=(select count(*) from #tmp_x)
select p.patientid, count(*)
from patient as p
left join #tmp_x x on x.patientid=p.patientid
where x.patientid is null
group by p.patientid
Thanks Jon Egerton

Answer 3

You can't use a single CTE for two distinct statements - they go out of scope.
So either you need to remove the need for the variable, or you can just declare a table variable and then keep selecting on that as many times as you like.
You can declare table var as follows:
declare #x table (rowno int, patientid varchar...
Then you can fill that using the select that you've currently got in your CTE.

Answer 4

The problem is with this line:
set #myInt=(select count(*) from x)
You are trying to do the select on x before it has been declared!
UPDATE
In this case it would be best to use a temporary table or table variable rather than a CTE. For example:
declare #myInt int
select row_number() over(partition by c.patientid order by c.admissiondate) as rn
,c.patientid,c.admissiondate
,max(c.claimsfromdate) as maxHemiDate
,min(c.claimsfromdate) minHemiDate
,(
select max(c2.claimsfromdate)
from claims as c2
where c2.patientid=c.patientid
group by c2.patientid
) as maxClaimsDate
,p.drgCode
,datediff(dd,min(c.claimsfromdate),max(c.claimsfromdate)) /7 as weeksWithHemi
INTO #XTable
from claims as c inner join icdclaims as ci on ci.id=c.id
inner join tblicd as t on t.icd_id=ci.icd_id
inner join patient as p on p.patientid=c.patientid
and p.admissiondate = c.admissiondate
and p.dischargedate = c.dischargedate
where t.icdText like '%X%' and p.statecode='21'
group by c.patientid, c.admissiondate, p.drgCode
set #myInt=(select count(*) from #XTable)
select p.patientid, count(*)
from patient as p
left join #XTable x on x.patientid=p.patientid
where x.patientid is null
group by p.patientid
This is the quick and dirty method, but you could obviously declare your table earlier in the script.

Answer 5

A CTE may be plural. It doesn't solve your stated problem, but it is often clearer to break a complex CTE into several steps, each building on the prior definitions. In this example there is a recursive CTE and a second CTE that computes statistics based on the first. The final SELECT combines the results:
with CTE
as (
select 1 as Number
union all
select Number + 1
from CTE
where Number < 10
),
Summary
as (
select Count( 42 ) as HowMany, Min( Number ) as Least, Max( Number ) as Most
from CTE
)
select *
from CTE cross join
Summary