OracleBulkCopy - strange behaviour ("invalid number")

Question

Hi there,
I'm using the .NET Oracle Data Client. I use the "OracleBulkCopy"-class for copying a datatable to the database. This worked for at least one table for me very well and really fast. But now I recreated the table. The previous table had among some other columns :
LSL FLOAT
USL FLOAT
TARGET FLOAT
CHARTID NUMBER(24,0)
The new table now has :
LSL BINARY_DOUBLE
USL BINARY_DOUBLE
TARGET BINARY_DOUBLE
CHARTID NUMBER(24,0)
Now the strange things began to happen, not only for this table, also for another where I changed some FLOATs to BINARY_DOUBLE.
I get this error on the insert which works on the old table with exactly the same data :
Error in row '1' column '18'
ORA-01722: invalid number
Column '18' is the column "CHARTID" which was changed in no way, the schema definition is exactly the same as before. The column has a valid number in it, to be exact it's a 201 .
Now the next really interesting thing is, if I copy the data from the old table where the insert worked to the new table, using this command
INSERT INTO NEWTABLE (SELECT * FROM OLDTABLE WHERE PK=48311)
I don't get any errors. The values inserted are correct.
I even tried casting all double-values to type double and all number-values to int before inserting the datatable - it didn't change anything.
I would appreciate any help.
Does anyone have an idea what is happening here?
Thanks & best regards,
Compu
P.S.: The code used for the insert :
// It does not make any difference according to the error if I comment the following 3 lines
OracleGlobalization glob = OracleGlobalization.GetThreadInfo();
glob.DateFormat = "yyyy-mm-dd hh24:mi:ss";
OracleGlobalization.SetThreadInfo(glob);
OracleBulkCopy copy = new OracleBulkCopy(conn);
copy.ColumnMappings.Add("PK", "PK");
// some other ColumnMappings, all correct
copy.DestinationTableName = tableName;
copy.BulkCopyTimeout = 10000;
copy.WriteToServer(dt); // the error occures on this line
copy.Close();
Edited by: Compufreak on 27.09.2011 01:55

Answer 1

Wow, I think I found the problem. I wrote a little test :
CREATE TABLE TEST1 (FLOATCOL FLOAT,NUMCOL NUMBER)
CREATE TABLE TEST2 (FLOATCOL BINARY_FLOAT,NUMCOL NUMBER)
CREATE TABLE TEST3 (FLOATCOL BINARY_DOUBLE,NUMCOL NUMBER)
       static void Main(string[] args)
            DataTable dt = new DataTable();
            dt.Columns.Add("FLOATCOL"); //  dt.Columns.Add("FLOATCOL", typeof(double))
            dt.Columns.Add("NUMCOL"); //  dt.Columns.Add("FLOATCOL", typeof(int))
            DataRow newRow = dt.NewRow();
            newRow["FLOATCOL"] = "1.337"; //  newRow["FLOATCOL"] = 1.337
            newRow["NUMCOL"] = "5"; //  newRow["NUMCOL"] = 5
            dt.Rows.Add(newRow);
            OracleConnection conn = new OracleConnection(Properties.Settings.Default.OracleString);
            conn.Open();
            OracleBulkCopy copy = new OracleBulkCopy(conn);
            copy.ColumnMappings.Add("FLOATCOL", "FLOATCOL");
            copy.ColumnMappings.Add("NUMCOL", "NUMCOL");
            copy.DestinationTableName = "TEST1";
            copy.WriteToServer(dt);
            copy.DestinationTableName = "TEST2";
            copy.WriteToServer(dt);
            copy.DestinationTableName = "TEST3";
            copy.WriteToServer(dt);
            copy.Close();
            conn.Close();
        }With the commented lines, all 3 statements work. With the uncommented lines, it only works for table "TEST1". I don't know why, but Oracle seems to cast the strings to numbers if FLOAT in combination with NUMBER is used - but not with BINARY_FLOAT and NUMBER .