C# Switch expressions returning different result

Question

I've switched to C# 8 on one of my projects. And I've been moving all of my switch statements to expressions. However I found out that my project started working differently and I've found out that it was because of the switch expression. Lets get this code for example

class Program
{
    public enum DataType
    {
        Single,
        Double,
        UInt16,
        UInt32,
        UInt64,
        Int16,
        Int32,
        Int64,
        Byte
    }

    static void Main(string[] args)
    {
        dynamic value1 = 5;
        dynamic value2 = 6;

        var casted = CastToType(value1, DataType.Int16);
        var casted1 = CastToTypeExpression(value2, DataType.Int16);


        var type = casted.GetType(); // Int16
        var type1 = casted1.GetType(); // Double
        var bytes = BitConverter.GetBytes(casted); // byte arr with 2 el => [5, 0] <- expected behavior 
        var bytes1 = BitConverter.GetBytes(casted1); // byte arr with 8 el => [0, 0, 0, 0, 0, 0, 24, 64]
    }

    public static dynamic CastToType(dynamic value, DataType type)
    {
        switch (type)
        {
            case DataType.Byte:
                return (byte)value;
            case DataType.Double:
                return (double)value;
            case DataType.Int16:
                return (short)value;
            case DataType.Int32:
                return (int)value;
            case DataType.Int64:
                return (long)value;
            case DataType.Single:
                return (float)value;
            case DataType.UInt16:
                return (ushort)value;
            case DataType.UInt32:
                return (uint)value;
            case DataType.UInt64:
                return (ulong)value;
            default: throw new InvalidCastException();
        }
    }

    public static dynamic CastToTypeExpression(dynamic value, DataType type)
    {
        return type switch
        {
            DataType.Byte => (byte)value,
            DataType.Double => (double)value,
            DataType.Int16 => (short)value,
            DataType.Int32 => (int)value,
            DataType.Int64 => (long)value,
            DataType.Single => (float)value,
            DataType.UInt16 => (ushort)value,
            DataType.UInt32 => (uint)value,
            DataType.UInt64 => (ulong)value,
            _ => throw new InvalidCastException(),
        };
    }
}

I've wrote the result as a comment, but tl;dr when classic switch is used casting the value returns the value in the expected Type, but when switch expression is used it returns type of "Double", resulting to different byte[] when getting the bytes of the value.

What's the difference between the two? What do I miss?

Answer 1

In your switch statement form, each arm is returning a value directly. It's converting from the numeric type directly to object, as that's effectively the return type of the method.

Your switch expression form is slightly different. It first extracts a result from the switch expression, then converts that result to the declared return type. So what's the type of the switch expression? It's the "best" type from all of the types of the individual expressions in the arms of the switch expression.

All of those types can be implicitly converted to double (which is one of the types itself), so that's the best type. So your switch expression method is equivalent to:

public static dynamic CastToTypeExpression(dynamic value, DataType type)
{
    double result = type switch
    {
        DataType.Byte => (byte)value,
        DataType.Double => (double)value,
        DataType.Int16 => (short)value,
        DataType.Int32 => (int)value,
        DataType.Int64 => (long)value,
        DataType.Single => (float)value,
        DataType.UInt16 => (ushort)value,
        DataType.UInt32 => (uint)value,
        DataType.UInt64 => (ulong)value,
        _ => throw new InvalidCastException(),
    };
    return result;
}

You can see this "best type" without using a switch expression, using implicitly typed arrays:

var array = new[]
{
    (byte) 0, 0.0, (short) 0, 0,
    0L, 0f, (ushort) 0, 0U, 0UL
};

Here the type of array is inferred to be double[].