Question

I’m using the SQL Server 2008 PIVOT function after I turn a CSV field into a table using a UDF.

For the purpose of showing this error, I’m eliminating the first step which is CROSS APPLYing the CSV field to a UDF to get the table I’m manually creating in the code example below.

CREATE TABLE #TCMDRES (
    SYS_NR BIGINT,
    DAT_TE VARCHAR(150),
    ID CHAR(1)
)

-- ,,,,,,WIN ASDF v2,20100406.BAK,32515325772,32514331136,
-- After being separated and put into columns 1-11...
INSERT INTO #TCMDRES
VALUES(1,'','X'),(2,'','X'),(3,'','X'),(4,'','X'),(5,'','X'),(6,'','X'),(7,'WIN ASDF v2','X'),
(8,'20100406.BAK','X'),(9,'32515325772','X'),(10,'32514331136','X'),(11,'','X')

SELECT *
FROM #TCMDRES

-- This errors out with:
-- Msg 248, Level 16, State 1, Line 16
-- The conversion of the nvarchar value '32514331136' overflowed an int column.
SELECT ID, [8] AS FIL_NA, [10] AS FIL_SZ_NR
FROM
    (
    SELECT DAT_TE, SYS_NR, ID
    FROM #TCMDRES a 
    ) a
PIVOT (MAX(DAT_TE) FOR SYS_NR IN ([8],[10])) AS pvt
WHERE [10] <> 0

DELETE FROM #TCMDRES

-- Try again but drop the last two digits from column 10...
INSERT INTO #TCMDRES
VALUES(1,'','X'),(2,'','X'),(3,'','X'),(4,'','X'),(5,'','X'),(6,'','X'),(7,'WIN ASDF v2','X'),
(8,'20100406.BAK','X'),(9,'32515325772','X'),(10,'325143311','X'),(11,'','X')

SELECT ID, [8] AS FIL_NA, [10] AS FIL_SZ_NR
FROM
    (
    SELECT DAT_TE, SYS_NR, ID
    FROM #TCMDRES a 
    ) a
PIVOT (MAX(DAT_TE) FOR SYS_NR IN ([8],[10])) AS pvt
WHERE [10] <> 0

DROP TABLE #TCMDRES

To summarize, if you attempt to PIVOT on a large value even when it’s stored as a VARCHAR, it fails trying to convert it to an INT.

Does anyone have any ideas on how to fix this or if it’s fixable? Seems to be a limitation in the PIVOT function…

Answer 1

It’s your WHERE clause. 0 is an INT. Try this (worked in testing for me):

WHERE [10] <> CAST(0 as bigint)

Alternatively, per Lamak:

another way would be WHERE [10] <> '0'