Top Posts & Pages
Error: Twitter did not respond. Please wait a few minutes and refresh this page.
Stephen Ritchie's offerings of ruthlessly helpful .NET practices.
Unreadable code with comments is inadequate code with comments you cannot trust. Code that is well written rarely needs comments. Only comments that provide additional, necessary information are useful.
Yesterday a colleague of mine told me that he lost 10 points on a university assignment because he did not comment his code. Today I saw a photo with a list of rules for commenting attributed to Tim Ottinger.
Ottinger’s three rules make a lot of sense. These rules are straightforward. In my experience, they are correct and proper. Here are Ottinger’s Comment Rules:
Comments are for things that cannot be expressed in code.
This is common sense. But, sadly, it is not common practice. Software is written in a programming language. A reader fluent in the programming language must understand the code. The code must be readable. It must clearly express what it is that the code does.
Only add comments when some important thing must be communicated to the reader, and that thing cannot be communicated by making the code any more readable. For example, a comment with a link to the MACRS depreciation method could be important because it helps explain the source of the algorithm.
Comments which restate code must be deleted.
Any restatement of the code is unlikely to maintained over time. If the comment is maintained, then it just adds to the cost. More importantly, when comments are not maintained they either end up substantially misrepresenting the code or end up being ignored. Reading comments that misrepresent code is a waste of time, at best. At worst, they cause confusion or introduce bugs. Remove any comments that restate the code.
If the comment says what the code could say, then the code must change to make the comment redundant.
Writing readable code is all about making sure that the compiler properly implements what the developer intended and making sure any competent developer can quickly and effectively understand the code. The code needs to do both: completely, correctly, and consistently. For example, a comment explaining that the variable x represents the principal amount of a loan violates the single truth rule. The variable ought to be named loanPrincipal. In this way the compiler uses the same variable to represent the same single true meaning that the human reader understands.
Tim Ottinger and Jeff Langr present more pragmatic guideance on when to write (and not write) comments: http://agileinaflash.blogspot.com/2009/04/rules-for-commenting.html