3.4. Documentation
We know, real programmers don't document. “It was hard to write; it should be hard to understand.” We feel your pain. However, the sanity you save may be your own. You could hop between jobs sufficiently rapidly that you never have to revisit a program you wrote more than six months ago.[2] Or you could document it so you'll understand what you did.
[2] But if you never write programs for your personal use, you can't consider yourself a true programmer.
Reams of valuable material have been written on the subject of program documentation, most of it by people who aren't programmers or who would be considered fringe programmers at best by people who think of themselves as “real” programmers. If documentation is your nemesis, we don't intend to try and emulate the worthy people who can tell you exactly how and why you should write scads of hierarchical documents.
Because we hate it too.
So here, instead, is the lazy person's guide to documentation, based purely upon selfish motives. (Of course, in some programming environments, you have no choice but to write copious documentation, because management requires it. These environments are rare, in our experience.)
Imagine this is a program written by someone else, and you've been asked or ordered to take it over. What help do you need to understand it? As crazy as this scenario sounds, it's as close to the truth as anything else. Many times, you will have to do something with a program you wrote but haven't seen for more than six months, and many times, you won't remember nearly enough about how you wrote it to get back up to speed.
If it was quick to write and easy to understand, don't waste time documenting it. There is such a thing as self-documenting code, and the better you are at writing it, the less time you'll have to waste explaining yourself. Not all code is self-documenting, of course. Nevertheless, writing the following calls for some comment and scrutiny:
for (my $i = 0; $i <= $#a; $i++) { $b[$i] = $a[$i] * 86400; }Whereas in writing
@seconds = map $_ * $SECS_PER_DAY, @days
it probably wouldn't.
 | I once spent about a month getting one number right in a program. It was a driver for a communications board, and the number represented the combined flag settings for a rather important register. The vendor documentation was inadequate for divining the correct settings, and each attempt at ringing the combinations required many hours of exhaustive regression testing. That one number received a lot of commenting when we got it right. |
It's not enough to say how things work, you also have to say why. Any time you make a choice that isn't obvious, explain it. Otherwise when you come back to it in six months, you'll be staring at that spot wondering, “Why did I do it that way? There's a much more obvious way…”
Comments are not enough. They may explain the how and the why of the code, but they don't say why you were doing it in the first place. Whether or not you put it in the same file as the code, somewhere you need a user guide for the person using your code, and you also need to capture the thoughts that went into the code's design. Why is there a module at this point in your class hierarchy? If you read your data from a flat file, instead of an Oracle database that everyone else uses, because your data file contains one more attribute than the database captures, say so. If it took thought to come up with, the thought needs to be captured somewhere.
Perl provides a couple of ways of documenting as you code, and in the ideal place for documentation—right next to the code itself.[3] The first way is the old dependable comment-to-end-of-line, used for informing a code maintainer of the operation of tricky code or why certain nonobvious choices were made. The second way is POD—plain old documentation—a rudimentary system of embedding interface documentation in the program itself, which can later be extracted into Unix man pages, HTML, or other formats by programs included with perl.[4]
[3] Brad Appleton terms the reason that this is advantageous the locality principle of reference documentation and has an extensive discussion on it at http://c2.com/cgi/wiki?LocalityOfReferenceDocumentation.
[4] Look for the programs pod2man, pod2html, or pod2anything-else-you-might-want-to-use.
 | Comment the hard parts. Use POD. |
This documentation is for the user of the program or module, someone who might be lucky enough never to need to look at your actual code. The advantage of putting this documentation in the same file as the code is twofold: it reduces the number of files you need to ship, and it enables you to put external documentation as close as you like to the actual code it refers to. (Although not out of sequence—POD is like a gap-toothed backwoods cousin of WEB, Donald Knuth's “literate programming” tool which was flexible enough to do that reordering and capable of producing wonderfully typeset annotated versions of your program. Unfortunately, the WEB user also had to learn TEX and learn how to embed TEX and program code in a hieroglyphic soup that made TECO macros read like Dr. Seuss. But enough nostalgia.)
POD is almost certainly the best choice for your user documentation, because people expect to find it in Perl files.
 | To be blunt, a proper program is defined by the set consisting of the executable code, comments that describe the how and why of the code, and documentation describing the function of the program (with examples). Real men and women comment their code! |