Book mentioned in podcast:
Code Craft
The practice of writing excellent code
by Pete Goodliffe
ISBN 1-59327-119-0
Rough transcript of podcast:
Writing code like a story
Hi, my name is Kris Coppieters from Rorohiko. This is the third Lightning Brain podcase - writing code like a story.
I've been programming for over 30 years now - and my programming style has changed over time, to suit new languages, to suit new ideas, to suit programming styles.
If I look back at code I wrote, say, ten years ago, the changes are not all that great when compared to code I wrote earlier; I seem to have settled into a number of habits that work well, and have not felt much need to change.
Many of the things I do also seem to coincide with the advice given in many books about coding style - most of it is common sense.
I want to make it clear that some of my preferences are just that - preferences; sometimes, as a coder, you get 'locked in' to a certain approach. There are often many other approaches that are at least as valid, but there is no clear benefit to switch between them, so you stay 'locked' into a particular way of doing things.
When I was younger, I would let myself be enticed into endless debates - things like where to put the braces {} and whether to use tabs or spaces, that kind of stuff.
Through experience, I've learned that those things are highly irrelevant. I have found that consistency is more important rather than what your are consistent in.
So, if you like the braces one way or another, you won't get any argument from me. But you will get an argument if your braces are sprinkled one way here, another way there. To me that's a bit the same as having to read a book where the font size jumps up and down at random: it makes reading the stuff harder for no reason.
I often have to work on other people's code - and currently, my approach is to follow whatever consistent coding convention is used in each individual source file. I have no trouble with a project that consists of source files created by individuals who all had different ideas on how to structure their code - as long as there is a consistency to it within each source code file.
I know of developers that cannot stand working on source code that is not exactly structured 'their' way. I consider such inability a major disadvantage: these people tend to waste large amounts of time on 'restructuring' some perfectly consistent piece of code.
I strive to write code with as few comments as possible. Yes - you heard that right. I think comments are a last-resort type of thing.
I try to write the code first and foremost in such a way that it is as clear and as easy to understand as possible; I think this might be more of an art than a technique.
I only use comments if I cannot make everything crystal clear by means of clean code. You do need as little comments as possible, but no less. Gratuitous comments that simply re-state what the code does are a no-no - things like:
// Increment i
i++;
// This is the constructor
make me cringe.
Comments are dangerous - they have a tendency to get outdated as the code around them evolves, and instead of being helpful, they often become a liability. So it is important to try hard to make the code so clear it does not need comments.
Instead, I restructure and rewrite my code until it is as close to self-explanatory as possible.
My priorities are always to write readable and understandable code first, and efficient code second. When someone else (often me) needs to pick up the code in a few months or years, the most important goal is to make it easy on the developer to grasp the meaning of the code. Not doing so is inviting bugs to be introduced during maintenance by a developer who only half-grasps or half-remembers what is going on.
Often, if the code cannot be made self-explanatory, that is a symptom that something is fundamentally wrong with the approach used, and you need to take a step back and rethink things.
Writing good code is almost the same as writing a good book or a good story: it has good content, is easy to understand, is consistent in as many respects as possible, is nicely laid out. Code also needs those traits.
I also pay a lot of attention to the code layout and neatness. Small example: I will often order all of the functions, procedures and declarations and alike in alphabetical order.
I know that many IDEs make it easy to 'jump' around to functions in a source code file - but sometimes I find myself separated from my finely honed development environment, working in an unfamiliar debugger, or staring at my source code in printed form, or using WordPad to view the code. Having everything alphabetical makes it easy to guess which direction to scroll to find something in the source code.
Just recently I also figured out a way to describe how I like to pick names for things.
One basic rule of thumb is: short scope allows short names. For example, if I need an index variable, and it will be needed for just two or three lines, I am perfectly happy to call it 'idx' or even 'i'. If the scope gets larger, and spans a few tens of lines, I will make the index name more descriptive - for example 'spreadIdx' or 'pageIdx'. If there are similar variables around, I find what sets them apart and name them accordingly, making sure they are not easily confused - unless in rare circumstances, I'll never use variables like 'idx' and 'idx2'.
Names that have large scope (e.g. span multiple source files) are often longer, and often tell a little story (but again - not too long; it's easy to go overboard).
When creating names, I also try to be consistent in how a name gets formed, and I will often build name that first states the more general and then the more specific.
Similar names have a similar 'lead-in', so in case they get ordered alphabetically (e.g. in a debugger or some IDEs), similar things end up 'close to' other similar things. For example, I'll use
const kFileName_Template = "bla.indt";
const kFileName_MainDocument = "yaya.ind";
instead of
const kTemplate_FileName = "bla.indt";
const kMainDocument_FileName = "yaya.ind";
Till next time!
Tuesday, August 7, 2007
Subscribe to:
Post Comments (Atom)
No comments:
Post a Comment