How to (efficiently) convert (cast?) a SqlDataReader field to its corresponding c# type?

First, let me explain the current situation: I'm reading records from a database and putting them in an object for later use; today a question about the database type to C# type conversion (casting?) arose.

Let's see an example:

namespace Test
{
    using System;
    using System.Data;
    using System.Data.SqlClient;

    public enum MyEnum
    {
        FirstValue = 1,
        SecondValue = 2
    }

    public class MyObject
    {
        private String field_a;
        private Byte field_b;
        private MyEnum field_c;

        public MyObject(Int32 object_id)
        {
            using (SqlConnection connection = new SqlConnection("connection_string"))
            {
                connection.Open();

                using (SqlCommand command = connection.CreateCommand())
                {
                    command.CommandText = "sql_query";

                    using (SqlDataReader reader = command.ExecuteReader(CommandBehavior.SingleRow))
                    {
                        reader.Read();

                        this.field_a = reader["field_a"];
                        this.field_b = reader["field_b"];
                        this.field_c = reader["field_c"];
                    }
                }
            }
        }
    }
}

This is (obviously) failing because the three this.field_x = reader["field_x"]; calls are throwing the Cannot implicitly convert type 'object' to 'xxx'. An explicit conversion exists (are you missing a cast?). compiler error.

To correct this I currently know of two ways (let's use the field_b example): number one is this.field_b = (Byte) reader["field_b"]; and number two is this.field_b = Convert.ToByte(reader["field_b"]);.

The problem with option number one is that DBNull fields are throwing exceptions as the cast is failing (even with nullable types as String), ant the problem with number two is that it's not preserving null values (the Convert.ToString(DBNull) yields a String.Empty), and I can't use them with enums too.

So, after a couple of lookups on the internet and here at StackOverflow, what I came up with is:

public static class Utilities
{
    public static T FromDatabase<T>(Object value) where T: IConvertible
    {
        if (typeof(T).IsEnum == false)
        {
            if (value == null || Convert.IsDBNull(value) == true)
            {
                return default(T);
            }
            else
            {
                return (T) Convert.ChangeType(value, typeof(T));
            }
        }
        else
        {
            if (Enum.IsDefined(typeof(T), value) == false)
            {
                throw new ArgumentOutOfRangeException();
            }

            return (T) Enum.ToObject(typeof(T), value);
        }
    }
}

This way I should handle every case.

开发者_Python百科Question is: Am I missing something? Am I doing a WOMBAT (Waste Of Money, Brain And Time) as there's a quicker and cleaner way to do it? It's all correct? Profit?

If a field allows nulls, don't use regular primitive types. Use the C# nullable type and the as keyword.

int? field_a = reader["field_a"] as int?;
string field_b = reader["field_a"] as string;

Adding a ? to any non-nullable C# type makes it "nullable". Using the as keyword will attempt to cast an object to the specified type. If the cast fails (like it would if the type is DBNull), then the operator returns null.

Note: Another small benefit of using as is that it is slightly faster than normal casting. Since it can also have some downsides, such as making it harder to track bugs if you try to cast as the wrong type, this shouldn't be considered a reason for always using as over traditional casting. Regular casting is already a fairly cheap operation.

don't you want to use the reader.Get* methods ? The only annoying thing is that they take column numbers so you have to wrap the accessor in a call to GetOrdinal()

using (SqlDataReader reader = command.ExecuteReader(CommandBehavior.SingleRow))
{
    reader.Read();

    this.field_a = reader.GetString(reader.GetOrdinal("field_a"));
    this.field_a = reader.GetDouble(reader.GetOrdinal("field_b"));
    //etc
}

This is how I've dealt with it in the past:

    public Nullable<T> GetNullableField<T>(this SqlDataReader reader, Int32 ordinal) where T : struct
    {
        var item = reader[ordinal];

        if (item == null)
        {
            return null;
        }

        if (item == DBNull.Value)
        {
            return null;
        }

        try
        {
            return (T)item;
        }
        catch (InvalidCastException ice)
        {
            throw new InvalidCastException("Data type of Database field does not match the IndexEntry type.", ice);
        }
    }

Usage:

int? myInt = reader.GetNullableField<int>(reader.GetOrdinal("myIntField"));

You can make a set of extension methods, one pair per data type:

    public static int? GetNullableInt32(this IDataRecord dr, string fieldName)
    {
        return GetNullableInt32(dr, dr.GetOrdinal(fieldName));
    }

    public static int? GetNullableInt32(this IDataRecord dr, int ordinal)
    {
        return dr.IsDBNull(ordinal) ? null : (int?)dr.GetInt32(ordinal);
    }

This gets a bit tedious to implement, but it's pretty efficient. In System.Data.DataSetExtensions.dll, Microsoft solved the same problem for DataSets with a Field<T> method, which generically handles multiple data types, and can turn DBNull into a Nullable.

As an experiment, I once implemented an equivalent method for DataReaders, but I ended up using Reflector to borrow an internal class from DataSetExtensions (UnboxT) to do the actual type conversions efficiently. I'm not sure about the legality of distributing that borrowed class, so I probably shouldn't share the code, but it's pretty easy to look up for oneself.

The generic hanlding code posted here is cool, but since the question title includes the word 'efficiently' I will post my less generic but (I hope) more efficient answer.

I suggest you use the getXXX methods that others have mentioned. To deal with the column number problem that bebop talks about, I use an enum, like this:

enum ReaderFields { Id, Name, PhoneNumber, ... }
int id = sqlDataReader.getInt32((int)readerFields.Id)

It's a little extra typing, but then you don't need to call GetOrdinal to find the index for each column. And, instead of worrying about column names, you worry about column positions.

To deal with nullable columns, you need to check for DBNull, and perhaps provide a default value:

string phoneNumber;
if (Convert.IsDBNull(sqlDataReader[(int)readerFields.PhoneNumber]) {
  phoneNumber = string.Empty;
}
else {
  phoneNumber = sqlDataReader.getString((int)readerFields.PhoneNumber);
}