Difference between revisions of "Writing Good Code"

Revision as of 19:12, 13 May 2008

Why this page

When tutoring students for their theses or projects, I often find many problems with the code they write. I'm not referring to bugs; bugs happen, as flu happens, although there are some things you can do to make them more unlikely. The problem discussed here is Bad code^TM, i.e., code that nobody can read or understand, not even its author. So I've decided to write this page with some advice on how to write Good code^TM. Please note that everything you find in this page should not be considered as strict rules, as it expresses the point of view of its author(s); but a reasonable piece of advice needs a good reason to counter it, in order not to follow it; so at least think about it. Contributions are welcome.

--Bernardo

General

Comments

A big problem is the use of comments. You can encounter projects with hundreds of lines of code and not a single line of comment, so that it is hard to guess even the general purpose of the program, or projects with ultra-verbose comments, like this:

k = k + 1; // increment k

Yeah, thanks, I thought it was extracting the square root of k.

Comments are important, and their correct use improves the quality of the code, even when there are no comments. Really! The source of the program is not only a way to obtain a binary that your computer can run, it is a way to express ideas. They should be clear to the compiler, so it compiles and you have your nice executable, but they should be clear also to anyone that reads your code, including your fellow students, your supervisor, and also you — even after a couple of months.

A tangled piece of code is not a way to clearly convey your ideas; and a comment added to a tangled piece of code doesn't make it much clearer. Besides, if you can't understand a piece of code, does some comment raise your confidence that the code is really working? Tangled code is more likely to be bugged, and more difficult to modify, so please write your algorithm in a way that it can be understood just be reading the instructions, not the comments. There good places where to write obfuscated code, and a thesis is not one of them.

Just an example (taken from real code! names have been changed to protect the innocent):

#define NUM_IT 3
for (j = 0; j < NUM_IT; ++j) {
    if (j == 0) { // First iteration
        // Do something
        ...
    } else if (j == 1) { // Second iteration
        // Do something else
        ...
    } else if (j == 2) { // Third iteration
        // Do things
        ...       
    }
}

This is a rather confusing way to write a simple sequence:

// Do something
...
// Do something else
...
// Do things
...       

So, what to comment? Good places for comments are functions (you tell what the function is about, what the parameters are, and describe the expected results and side effects), global or important variables, the beginning of a file, and a few others. Also, when you take a tricky decision, document it!

Indentation and spaces

In most languages, spaces and indentation are not part of the syntax, and they are mostly ignored by the compiler (Python is a significant exception), yet indentation and spaces are very useful to format your code and make it more readable. There are many way to use them, but the first rule is 'consistency'. Choose the style you like the most, and stick to it; particularly, choose if you want to use spaces or tabs for indentation, and be consistent, otherwise when someone else opens your project with a different editor with a different idea of tab length, your nice (you made it nice, didn't you?) indentation will be screwed up.

C and C++ code

The Linux kernel coding style is an interesting reading, particularly chapters 3 (Placing Braces and Spaces), 4 (Naming), 6 (Functions), 8 (Commenting).

Matlab

Matlab is a rather slow interpreted language, but it shines — as its name implies — at matrix manipulation. So try to avoid loops, and do operations in parallel on arrays. Many vectorized functions are built-in, so look in the help when you need simple operations like sums, means, maximum...