DataReader returning incorrect .net datatype for database table float column

Question

I have 1 Table in Oracle SQL Developer which containts 1 column as Float.Data reader have should return Decimal for oracle float datatype as per the table given here : https://learn.microsoft.com/en-us/dotnet/framework/data/adonet/oracle-data-type-mappings

But the problem is datareader returns double as datatype for Float Column as shown below :

But the problem is datareader returns double as datatype for NREAL as Float and NFLOAT1 as float and surprisingly datareader returns Decimal for both the column as shown below :

Code :

static void Test()
        {
            using (OracleConnection connection = new OracleConnection("connection string")
            {
                connection.Open();
                using (OracleCommand command = connection.CreateCommand())
                {
                    command.CommandText = "select id , NFLOAT from Numeric_Table";
                    using (OracleDataReader reader = command.ExecuteReader())
                    {
                        for (int i = 0; i < reader.FieldCount; i++)
                        {
                            var columnName = reader.GetName(i);
                            var dotNetType = reader.GetFieldType(i);
                            var sqlType = reader.GetDataTypeName(i);
                        }
                    }
                }
            }
        }

I am using : Oracle.ManagedDataAccess.Client

Is that a bug inside Oracle.ManagedDataAccess.Client library or I am doing something wrong?

Update : Based on comments I want to mention something :

Although I might have reference to different source documentation not applicable to oracle library I am using,bt still I am getting decimal datatype for my other 2 columns i.e NREAL and NFLOAT1 so why this behaviour is not consistent ?

Answer 1

The FLOAT data type is not the whole picture here. SQL Developer is not showing you that NFLOAT has a precision that's less than 50 while NREAL and NFLOAT1 have precisions that are 50 or higher. Left unspecified when created, the precision defaults 126, which is the highest possible precision for the FLOAT data type.

A simple test query will demonstrate the difference:

SELECT CAST(0 AS FLOAT(49)), CAST(0 AS FLOAT(50)) FROM DUAL

The first column will be returned as System.Double. The second will be returned as System.Decimal.

In SQL Developer, right-click on the table and select "Edit..." That dialog does show the defined precision for the column.

If you want NFLOAT to be returned as System.Decimal increase the precision to at least 50.

---

I've been combing the documentation and running my own tests lately. It's a pet project and I don't use Oracle regularly so there's been a lot of trial and error. The DUMP function has been quite helpful in determining if I have the right data type in an expression before seeing what ODP .NET does to it.

For another example of how ODP .NET returns different data types depending on the precision, consider NUMBER(p,0).

Type     Min p    Max p
---------------------------
Int16    1        4
Int32    5        9
Int64    10       18
Decimal  19       38

I was hoping that Byte or SByte would be returned when p is 1 or 2 but that didn't pan out.