Ask HN: Inheritance? | Hacker News Reader

Ask HN: Inheritance?

3 pointsBigZaphod5 comments

Inheritance is just an implicit form of delegation that comes with a bunch of default behaviors that can totally screw you.

Of course I use it constantly and spend a ton of time "designing" everything "properly" to have a sweet class hierarchy with all kinds of unnecessary power and flexibility to solve the problem at hand.

Does anyone else find themselves sucked into that trap? It seems like when using straight C, I can still think and program in an OO style, but I don't so readily waste as much effort trying to come up with a "perfect" tree of objects. Is this simply an issue of discipline or is there a drug-like effect when using inheritance?

View on HN

It seems natural to overdo a new tool until you've found its best uses.

But I think the only drug is perfectionism, and it applies to anything reasonably complicated. The cure is brutal honesty. How much will you really be able to reuse in a sophisticated design? (Not much.) How likely is it to change? (Very likely.) Will an unanticipated new feature completely destroy your formerly-perfect object design? (Often, yes.)

One of the original thoughts behind inheritance was to manage complexity. But since code is read much more than it is written, I think of my eyes and not my fingers. Sure, I may have saved some time writing the code, and even had fun coming up with its unique design. But the next 40 or 50 times I have to go back and understand the code, and read 4 or 5 other files just to figure out what a function does, I see the error of my ways.

That doesn't mean I avoid inheritance, but I strictly limit it to things where it has immediate and obvious value. I may have one parent class, and the class isn't very big. It's also not something I'm likely to share in a public header; I'll inherit from it myself, and not ask others to do the same, which avoids a whole category of hurt. :)

The effect is this: it looks like real work but doesn't require you to hit compile and run the app and find that it doesn't work. It's the same with refactoring. People love doing it because you seem busy but don't have to take the risk of revealing (if only to yourself) that your code doesn't do what you think it ought.

It's one reason for the success of Java in the marketplace, there's so much scaffolding to be done that it keeps loads of people busy-looking.

Yeah, I hear ya. I think everyone has that to a certain degree. Once I heard of YAGNI (http://en.wikipedia.org/wiki/You_Ain%27t_Gonna_Need_It), I was able to let go and just code more efficiently.

http://en.wikipedia.org/wiki/Is-a versus http://en.wikipedia.org/wiki/Has-a. Usually the "perfect" tree isn't really all that perfect.

You should almost never use implementation inheritance. In fact, until you have mastered designing with composition, the 'extends' keyword should be banned from your vocabulary.

Reply on news.ycombinator.com