Wednesday, October 31, 2007

Coding for maintainability

One important criteria for judging the value of any serious software product is how maintainable it is over its life cycle. It is important to notice that code is "write once" but "read many times" when people subsequently maintains it (enhancing features, fixing bugs, optimizing performance). Any cost spent in understanding any non-trivial code will be multiplied many times by itself and added to the overall maintenance budget of the software product.

It is so important to spend all the effort to make the code as readable as possible. Once you do it, you will start to get your dividends over the product's life cycle.

Lets set some criteria/characteristics of code quality from a maintainability standpoint.

Good code

Just by looking at its signature, you can guess what a class/method/function is doing without looking at its implementation. A very descriptive name of the class/method/function and parameter/variable are carefully chosen such that its associated semantics is very clear. When you open up its implementation, it is doing exactly what you guess.

Since the naming is so descriptive, it doesn't need a lot of comments to make the code readable. Comments are only appears for code segments that are implementing a complex algorithm or when optimizing performance of hotspots (where code readability is intentionally sacrificed).

The responsibility of a class/method/function is very clear. There is only one way to do a particular thing and is well-encapsulated within the responsible class/method/function. There is no duplicated code.

Responsibility tends to be distributed evenly so you don't have a package with too many classes or a class with too many instance variables/methods. And you won't have a method that span more than 2 pages of code lines.

Bad code

Similar code shows up in many places, it is hard to know which of these are "similar" and which are "exactly same"

Extremely long methods (couple hundreds lines of code within a method) and usually accompanied by deeply nested case statements

Only have trivial comments and lack of useful comments

There are many different ways to do one particular thing although they are supposed to achieve the same result

To do a particular thing, the caller need to make multiple calls to different objects in a certain sequence.

The class/method/function seems to be implying certain responsibilities but when you open up its implementation, you figure out it is doing something completely different. You quickly lose confidence in your understanding and has to trace down into all implementations details to see what it is actually doing in every place.

There are some places in the code where you cannot figure out what it is trying to do


In my next blog, I'll discuss how bad code is formed.

No comments: