What you want to do is to encapsulate all low level API details, so that your library can be used anywhere. Right? Well, if that's the case, you should not create a light wrapper. What you are doing is coupling the API calls in one object. What you want to be doing is define a set of styles for the window. Define all the other stuff yourself also (windows messages, controls, etc..) and in your code, translate this stuff into windows API calls. This way, you can provide different implementations (makes your project portable). In the ideal case, you don't have to modify your application when you compile it under Linux or whatever GUI. The only thing that has to be done is to use a different implementation of your library! To get back on track here and to answer your question: I think that you should go for the two stage window construction. One rule is to never throw an exception from a constructor. So here you have almost no way of notifying the caller of your code that creation failed. One other suggestion though. For abstraction purposes, you should hide implementation details. The win API is an implementation detail. So hInstance variables should be hidden. The raw styles should be hidden. Try to hide as much handles as you can, etc, etc.. Behind every great black man... ... is the police. - Conspiracy brother Blog[^]
