Jay Kimble

[I really should stay out of these discussions... I really should, but I can’t resist.]

The other day I saw an article on Composition over Inheritance (I forget where I saw it). The article did a good job of explaining what "composition" is. This is my simplistic understanding (which means that someone will probably come in and tell me I have it ALL wrong... what else is new?): "composition" means essentially wrapping objects instead of direct inheritance (so instead of inheriting from X you instead have a private instance of X in your class that you use, but you create your own interface possibly something that looks totally different from the "wrapped" class).

I only really have one concern in all this; at least with what I have read (and not necessarily the last article I read on this subject). The advice almost comes across as "thou shalt avoid inheritance (if you can)." The implication is that inheritance is something bad. Now I understand that inheritance can complicate things in a unit testing scenario (be it TDD or POUT).

The problem is that implication that inheritance is bad. I think it might be better to discuss some things I recently did in the DLRScript source code. I recently used composition to build a compatible XMLHttpRequest object in my DLRScript environment (no, it’s in the unreleased bits which will be released as soon as I can test it properly... I was hoping for a JQuery compatibility, but it looks like that ain’t happening this iteration). I wrapped the SilverLight HttpWebRequest object. This let me create a Mozilla/Safari-compatible XmlHttpRequest object that has no extra features other than what those objects contain.

That said I also have some code I have had in place for some time that also works well. This code inherits from the SilverLight HtmlDocument, and creates a document object that is more in line with what we are used to seeing in client script in the browser. I also have a class that wraps HtmlElement to create an object that is also more familiar to JavaScript junkies. I recently added a style property to each of these elements, for instance. My style property simply utilizes the setStyle/getStyle methods (I forget the actual method names) which is already available in the HtmlElement (for instance) to get/set values of individual style properties. In this case having a few dangling methods doesn’t really hurt because I already need 90% of what is already there. I do have to override a few methods since I need to emit DomElements (my inherited form of the HtmlElement) from getElementById instead of an HtmlElement.

My point is this. Think. Think! THINK!!! THINK!! Don’t just blindly follow a rule. Look at your code. Take control of it. Make it do what you want it to do. Make sure that it makes sense. and if someone else is going to be using it, ask someone else if they think what you are doing makes sense. And then DOCUMENT IT!! At least provide an example of how to use it.