SQL Server 2008: Error converting data type nvarch

问题:

Presently troubleshooting a problem where running this SQL query:

UPDATE tblBenchmarkData 
SET OriginalValue = DataValue, OriginalUnitID = DataUnitID, 
    DataValue = CAST(DataValue AS float) * 1.335 
WHERE 
    FieldDataSetID = '6956beeb-a1e7-47f2-96db-0044746ad6d5' 
    AND ZEGCodeID IN 
             (SELECT ZEGCodeID FROM tblZEGCode 
              WHERE(ZEGCode = 'C004') OR 
                   (LEFT(ZEGParentCode, 4) = 'C004'))

Results in the following error:

Msg 8114, Level 16, State 5, Line 1
Error converting data type nvarchar to float.

The really odd thing is, if I change the UPDATE to SELECT to inspect the values that are retrieved are numerical values:

SELECT DataValue 
FROM tblBenchmarkData 
WHERE FieldDataSetID = '6956beeb-a1e7-47f2-96db-0044746ad6d5' 
AND ZEGCodeID IN 
         (SELECT ZEGCodeID 
          FROM tblZEGCode WHERE(ZEGCode = 'C004') OR 
                               (LEFT(ZEGParentCode, 4) = 'C004'))

Here are the results:

DataValue
2285260
1205310

Would like to use TRY_PARSE or something like that; however, we are running on SQL Server 2008 rather than SQL Server 2012. Does anyone have any suggestions? TIA.

回答1:

It would be helpful to see the schema definition of tblBenchmarkData, but you could try using ISNUMERIC in your query. Something like:

SET DataValue = CASE WHEN ISNUMERIC(DataValue)=1 THEN CAST(DataValue AS float) * 1.335 
                     ELSE 0 END

回答2:

Order of execution not always matches one's expectations.

If you set a where clause, it generally does not mean the calculations in the select list will only be applied to the rows that match that where. SQL Server may easily decide to do a bulk calculation and then filter out unwanted rows.

That said, you can easily write try_parse yourself:

create function dbo.try_parse(@v nvarchar(30))
returns float
with schemabinding, returns null on null input
as
begin
  if isnumeric(@v) = 1
    return cast(@v as float);

  return null;
end;

回答3:

So starting with your update query that's giving an error (please forgive me for rewriting it for my own clarity):

UPDATE B
SET
   OriginalValue = DataValue,
   OriginalUnitID = DataUnitID,
   DataValue = CAST(DataValue AS float) * 1.335
FROM
   dbo.tblBenchmarkData B
   INNER JOIN dbo.tblZEGCode Z
      ON B.ZEGCodeID = Z.ZEGCodeID
WHERE
   B.FieldDataSetID = '6956beeb-a1e7-47f2-96db-0044746ad6d5'
   AND (
      Z.ZEGCode = 'C004' OR 
      Z.ZEGParentCode LIKE 'C004%'
   )

I think you'll find that a SELECT statement with exactly the same expressions will give the same error:

SELECT
   OriginalValue,
   DataValue NewOriginalValue,
   OriginalUnitID,
   DataUnitID OriginalUnitID,
   DataValue,
   CAST(DataValue AS float) * 1.335 NewDataValue
FROM
   dbo.tblBenchmarkData B
   INNER JOIN dbo.tblZEGCode Z
      ON B.ZEGCodeID = Z.ZEGCodeID
WHERE
   B.FieldDataSetID = '6956beeb-a1e7-47f2-96db-0044746ad6d5'
   AND (
      Z.ZEGCode = 'C004' OR 
      Z.ZEGParentCode LIKE 'C004%'
   )

This should show you the rows that can't convert:

SELECT
    B.*
 FROM
    dbo.tblBenchmarkData B
    INNER JOIN dbo.tblZEGCode Z
       ON B.ZEGCodeID = Z.ZEGCodeID
 WHERE
    B.FieldDataSetID = '6956beeb-a1e7-47f2-96db-0044746ad6d5'
    AND (
       Z.ZEGCode = 'C004' OR 
       Z.ZEGParentCode LIKE 'C004%'
    )
    AND IsNumeric(DataValue) = 0
    -- AND IsNumeric(DataValue + 'E0') = 0 -- try this if the prior doesn't work

The trick in the last commented line is to tack on things to the string to force only valid numbers to be numeric. For example, if you wanted only integers, IsNumeric(DataValue + '.0E0') = 0 would show you those that aren't.