Daniel's Blog

Couch Painting

Posted by Daniel Lyons Thu, 06 Sep 2007 03:52:00 GMT

Recently I had a kind-of debate with a kind-of friend about programming. It came to basically one thing: do you value readability?

I certainly do. I think one of the things that makes programming interesting is that we’re trying to express something to a computer and another human being at the same time. All programming failures fall on one side or the other of that balancing act.

The argument for readability cannot be absolute though. A hundred lines of Python may be more readable than 20 lines of Haskell. So we have to ask ourselves some more questions besides pure readabilit. We have to ask ourselves if we are following the elements of style. Can we “omit needless words?”

• Does this code communicate in 20 lines the same thing as 100 lines of Python?
• Do we gain from the conciseness?
• Does it take the same amount of time to understand?
• Do we want to constrain all of our developers to the same standard?

The allure of functional programming is that it is both more rigorous and more concise. When you understand a fold, you are tempted to use it even in languages where its application will be awkward or difficult, such as Ruby. This template substitution probably has appeared on my blog before; normally this would be coded as an imperative loop, but I chose to use Array#inject which is basically a fold:

vars.inject(@template.clone) do |template, (region, value)|
  template.gsub(/%#{region}%/, value.to_s)
end

In OCaml this would probably have looked a lot more concise than it does in Ruby, and it would be considered more readable. So you have to consider your audience too. Haskell’s audience is smarter than Java’s, or at least thinking in a more high-level way. Though, really, you should try to learn Haskell, it’ll make you smarter. (Or if you’re really vindictive you’ll at least bring the average down.)

So my operative concept is not readability anymore. It’s conciseness. Concise not like Perl or APL, concise like “omit needless words.” It may take you longer to figure out than comparable 5x larger Python code, but it won’t take you 5x longer, and you won’t be falling asleep.

And why do you believe your code is so important anyway? I’d bet 99.9% of code in this world is couch paintings. It seems so important right now. In five years, is it going to be important? Or is it going to already have been rewritten by then? So much code is treated like an investment when it is written as so much slapdash cut-and-paste filler. Other code is written elegantly and perfectly and winds up forgotten because nobody knew how to treat it like an investment. The entities most qualified to treat code as an investment seem to have practices that encourage the quality being the lowest. Hiring lots of developers, outsourcing, having corporate policies about language choice and a complete lack of code reviews have a strong negative effect on code quality.

If we really wanted to treat code as an investment, wouldn’t we want to revel in it a bit more? Read it, really take the time to enjoy it, pass the tricks around the organization? Treat the code like a member of the company rather than a process or, worse, the result of some process like so much insect shit?

And while we’re on the subject, I no longer believe any project really needs more than three developers.