I have a procedure which returns the error:
Must declare the table variable \"@PropIDs\".
But it is followed with the message:
You can avoid using dynamic sql by creating a sql function that use a CTE (I found the code below many years ago on sqlservercentral - Amit Gaur) :
Change the body of your procs with something like this :
SELECT p.WPRN AS ID,
p.Address AS Address,
p.Address AS Street
FROM [dbo].[Properties] AS p
WHERE
p.WPRN NOT IN ( SELECT item FROM dbo.strToTable(@NotNeededWPRNs, ','))
Below the sql code that transforms a string into a table :
CREATE FUNCTION [dbo].[strToTable]
(
@array varchar(max),
@del char(1)
)
RETURNS
@listTable TABLE
(
item int
)
AS
BEGIN
WITH rep (item,list) AS
(
SELECT SUBSTRING(@array,1,CHARINDEX(@del,@array,1) - 1) as item,
SUBSTRING(@array,CHARINDEX(@del,@array,1) + 1, LEN(@array)) + @del list
UNION ALL
SELECT SUBSTRING(list,1,CHARINDEX(@del,list,1) - 1) as item,
SUBSTRING(list,CHARINDEX(@del,list,1) + 1, LEN(list)) list
FROM rep
WHERE LEN(rep.list) > 0
)
INSERT INTO @listTable
SELECT item FROM rep
RETURN
END
The reason you get this error is that the scope of table variables is limited to a single batch, since sp_executesql
runs in its own batch, it has no knowledge that you have declared it in another batch.
It works when you @NotNeededWPRNs
is NULL
because concatenating NULL
yields NULL
(unless otherwise set), so you are just executing:
exec sp_executesql null;
I would also say, if you are using SQL Server 2008 or later please consider using table valued parameters instead of a delimited list of strings. This is much safer and more efficient, and validates the input, if I were to pass 1); DROP TABLE dbo.Prioperties; --
as @NotNeededWPRNs
, you could find yourself without a properties table.
First you would need to create the type (I tend to use a generic name for reusability):
CREATE TYPE dbo.IntegerList TABLE (Value INT);
Then you can add it to your procedure:
CREATE PROCEDURE [dbo].[GetNeededProperties]
@NotNeededWPRNs dbo.IntegerList READONLY,
@LastSynch DATETIME,
@TechCode VARCHAR(5)
AS
BEGIN
SELECT p.WPRN AS ID,
p.Address AS Address,
p.Address AS Street
FROM [dbo].[Properties] AS p
WHERE p.WPRN NOT IN (SELECT Value FROM @NotNeededWPRNs)
On an unrelated note, you should avoid using culture sensitive date formats where possible, 06/28/2013
is clearly supposed to be 28th June in this case, but what about 06/07/2013
, without setting DATEFORMAT
, or the language how do you know if this will be read as 6th July or 7th June? The best format to use is yyyyMMdd
, it is never ambiguous, even the ISO standard format yyyy-MM-dd can be interpreted as yyyy-dd-MM
in some settings.
Change you code to :
Declare @ProductsSQL nvarchar(max);
SET @ProductsSQL = 'DECLARE @PropIDs TABLE
(ID bigint);
Insert into @PropIDs (ID)
SELECT [WPRN] FROM [dbo].[Properties] WHERE(WPRN in (' + @NotNeededWPRNs + '))'
exec sp_executesql @ProductsSQL
Table variable declared outside the dynamic SQL will not be available to the dynamic SQL.
The issue is that you're mixing up dynamic SQL with non-dynamic SQL.
Firstly - the reason it works when you put NULL into @NotNeededWPRNs is because when that variable is NULL, your @ProductsSQL becomes NULL.
WHat you need to do is either make your @PropsIDs table a non-table variable and either a temporary table or a physical table. OR you need to wrap everything in dynamic SQL and execute it.
So the easy way is to do something like this:
Declare @ProductsSQL nvarchar(max);
SET @ProductsSQL = '
DECLARE @PropIDs TABLE
(ID bigint)
Insert into @PropIDs (ID)
SELECT [WPRN] FROM [dbo].[Properties] WHERE(WPRN in (' + @NotNeededWPRNs + '))
SELECT p.WPRN AS ID,
p.Address AS Address,
p.Address AS Street
FROM [dbo].[Properties] AS p
WHERE
p.WPRN NOT IN( SELECT ID FROM @PropIDs)
'
and execute that. OR as mentioned - change @ProdIDs to a temporary table. (The route you're approaching in the CREATE #ProdIds, but then you need to use #ProdIDs instead of @ProdIDs everywhere in the sproc).