What is the difference between this:
public class Foo {
private Bar bar;
public Foo() { bar = new Bar(); }
}
and this:
The CLR does not support initializing fields like this. To make it work, the C# compiler rewrites your constructor. The IL looks like this:
IL_0000: ldarg.0
IL_0001: newobj instance void ConsoleApplication1.Bar::.ctor()
IL_0006: stfld class ConsoleApplication1.Bar ConsoleApplication1.Foo::bar
IL_000b: ldarg.0
IL_000c: call instance void [mscorlib]System.Object::.ctor()
IL_0011: ret
Note that the Bar constructor is called before the Foo base constructor. Which is the difference with your other snippet, there the Foo base constructor is called afterward. This will rarely make a difference, unless the field is inherited and the base class constructor does something with it.
Or if field initializers depend on each other, they are initialized in strict textual order. You can control the order by doing it explicitly in the constructor.
There is no difference, because a compiler places all field definitions before constructor's code. So in IL-code both variants look equally.
Stating the obvious a bit, but with the second approach there's no need for a constructor.
The difference is that in the second case field initialization happens before this/base constructor while in the first case initialization happens inside the constructor.
To add to Darin Dimitrov's answer, initialising the fields in line is a good idea if you have several constructor overloads, and you don't want to call other overloaded constructors for some reason.