Convert numeric types: Which style to use?

Question

Let's say I want to convert a Double x to a Decimal y. There's a lot of ways to do that:

1. var y = Convert.ToDecimal(x);    // Dim y = Convert.ToDecimal(x)
2. var y = new Decimal(x);          // Dim y = new Decimal(x)
3. var y = (decimal)x;              // Dim y = CType(x, Decimal)
4. -- no C# equivalent --           // Dim y = CDec(x)

Functionally, all of the above do the same thing (as far as I can tell). Other than personal taste and style, is there a particular reason to choose one option over the other?

EDIT: This is the IL generated by compiling the three C# options in a Release configuration:

1. call valuetype [mscorlib]System.Decimal [mscorlib]System.Convert::ToDecimal(float64)
   --> which calls System.Decimal::op_Explicit(float64)
       --> which calls System.Decimal::.ctor(float64)
2. newobj instance void [mscorlib]System.Decimal::.ctor(float64)
3. call valuetype [mscorlib]System.Decimal [mscorlib]System.Decimal::op_Explicit(float64)
   --> which calls System.Decimal::.ctor(float64)

This is the IL generated by compiling the four VB options in a Release configuration:

1. call valuetype [mscorlib]System.Decimal [mscorlib]System.Convert::ToDecimal(float64)
   --> which calls System.Decimal::op_Explicit(float64)
       --> which calls System.Decimal::.ctor(float64)
2. call instance void [mscorlib]System.Decimal::.ctor(float64)  
3. newobj instance void [mscorlib]System.Decimal::.ctor(float64)
4. newobj instance void [mscorlib]System.Decimal::.ctor(float64)

So, it all ends up in System.Decimal::.ctor(float64)

Answer 1

I would suspect that using the unary casting operator would give the compiler a better chance to optimize, but I base this on absolutely nothing.

Answer 2

Convert.ToInt32() applies rounding to real numbers while casting to int just removes the fractional part. In my opinion typecasting method for "conversions" relies on .NET framework's magic too much. If you know that a conversion will have to take place, describing it explicitly is the easiest to understand. I would go for Convert option for most of the cases unless there is no conversion needed and a cast just works.

Answer 3

With non-object and non-string source data types, I prefer casts. Concise and easy to read.

For object and string source types, it depends on the source. For example, if the source is from the user, then I use Convert/Conversion methods.

I recommend downloading the source code for .NET at "Microsoft Reference Source Code Center" (http://referencesource.microsoft.com). Check out Convert.cs, Converstion.vb, and Conversions.vb. I previously would use the Convert/Conversion methods in all cases, but after reviewing these source files, my preference is casts.

Taking your example Convert.ToDecimal() and referencing Convert.cs, the call is really a simple cast:

    public static decimal ToDecimal(double value) { 
        return (decimal)value;
    }