I was banging my head against the wall in SSIS trying to copy data from SQL2000 to SQL2005, with error messages like these:
Description: “OLE DB provider ‘STREAM’ for linked server ‘(null)’ returned invalid data for column ‘[!BulkInsert].column_name’. “.

Turned out it was due to NaN being inserted by a third party. And SSIS did not like them.

Anyhoo… I modified Antonio’s function to return Nulls instead NaN and used them on the offending columns. Works like a charm. So thanks again everybody

Thanks for the Article !!!

In Our System it was Insereting Double.NaN into a float column that cause the problem

You can see the value “NaN” in SQL 2005 ,but not in SQL 2000!

Thanks again

I ran a trace and narrowed down the affected tables. Next, I ran dbcc show_statistics ‘table name’ and statement consistently failed on one table in particular. I dropped the statistics from the table and recreated the stats using sp_autostats ‘table name’. Afterwards, the error was corrected. The article below led me in the correct direction. It should be mentioned that I DID NOT install SP4 to resolve the problem.

http://support.microsoft.com/kb/816503

Hope this helps.

I don’t have a current test case to try this out on, and am disinclined to go creating one now if you can already confirm that this works. :)

CREATE FUNCTION IsNaN
(
@number float
)
RETURNS int
AS
BEGIN
return case isnumeric(convert(varchar(100), coalesce(@number,0))) when 1 then 0 else 1 end
END
GO

CREATE FUNCTION IsNaN
(
@number float
)
RETURNS int
AS
BEGIN
return isnumeric(convert(varchar(100), @number))
END
GO

and then …

UPDATE tblName SET floatCol = @someOkValue where dbo.IsNan(floatCol) = 1

UPDATE tblName SET floatCol = @someOkValue WHERE id IN (x,y,z, ...)

where you fill in the id list by hand, by eyeballing the bad rows in Enterprise Manager. If your table is very big, then I don’t imagine you’ll enjoy that much. But I don’t know any automated way to match these broken values in a WHERE clause, sorry.