> If you use binary serialization to clone, you're halfway to copy/paste, if
> that's a
> goal. If a class is properly clonable, it ought to be
> serializable/deserializable, it
> seems to me. Then you can kill two birds with one stone: implement Iclonable
> in
> terms of binary serialization. If a class derives from MarshalByRefObject, it
> probably shouldn't be clonable, but I don't want to be dogmatic about it.
There are two important drawbacks about cloning by serialization that
one should be aware of:
1- Performance: Using BinaryFormatter is much slower than a manual
implementation of Clone, and it's probably also much slower than a
reflection-based implementation. Of course, as "slow" is relative,
that's only an issue if you need to do a lot of cloning within tight
time constraints.
2- Semantics: A well-behaved implementation of Clone would be expected
to recursively call Clone on any ICloneable objects it contains. This
does not happen when you do binary serialization, which can be an
issue.
Regarding the original question (in case you don't want to use
serialization): You can use Reflection to check whether an object has
been cloned correctly, as others have suggested, but as you've
noticed, this only moves the problem to the issue of setting all
fields of an object prior to testing.
However, this is something you can ensure in your test: if you have a
reflection-based comparer, you can not only test whether the object
has been cloned correctly, but also that the source object is not
equal to a "default" one.
Illustrative code:
[Test]
public void Clone()
{
MyClass source = GetCloneTestSource();
MyClass defaultObject = new MyClass();
Assert.IsFalse (ReflectionBasedComparer.AreEqual (source,
defaultObject), "Forgot to initialize field for cloning.");
MyClass clone = source.Clone();
Assert.IsTrue (ReflectionBasedComparer.AreEqual (source, clone),
"Clone failed");
}
Fabian
===================================
This list is hosted by DevelopMentor® http://www.develop.com
View archives and manage your subscription(s) at http://discuss.develop.com
