Friday, February 13, 2009

Clever Programming

There are certain programming language features that my Principles of Programming Languages professor disapproves of. For instance, in C (and, to a slightly lesser extent, Java), the operator == is used to compare two things. For instance,

if (a == b) {
(do something)
}


will do the something if a is equal to b, and not otherwise. On the other hand, the operator = sets one value to be equal to another. For instance, a = b will take the value in b and store it in a.

However, the = operator also returns the value of b, so if you want to be clever, you can do something like

if (a = b) {
(do something)
}


This sets a equal to b and does something if b was true. Unfortunately, it also leads to rampant errors, because when people type if (a = b) they almost always meant if (a == b), and if = did not return a value, then the compiler would let you know you were making a mistake.

So, my professor is against using = in an if statement; it's the kind of clever thing that leads to mistakes. He also inveighs against things like using the ++ operator in an expression. If you write a++ it will increment a by 1 (so it is the same as a = a + 1). No problem there. ++a does the same thing. But where they are different is that they each return a value, but ++a increments a before returning it, and a++ increments it after returning. So that if I said something like

if (k++ == 3)

I would be saying "If k is equal to three (and, by the way, after you do that, increment k by 1)..", but if I wrote

if (++k == 3)

I would be saying "Increment k, then if it's equal to 3..."

Clever! It saves you a whole line of code, and makes your code harder to maintain and more error-prone at all once!

A guy in my class (who looks like a young, evil Vincent D'Onofrio) argued back, saying that he always uses such methods when possible. He argued that anyone who knows the language should be able to read a statement like

if (a = ++k)

so that he is not worried that it won't be understood. He doesn't work with CS1 students, after all. Someone who wouldn't understand these basic language features is worthy only of contempt, he suggested.

What I think he's missing is that maintenance makes up most of the costs associated with code, and every time that it takes someone 10% longer to read a line of your code, or even 5% more likely to make an error in maintaining it (not to mention your own increased chances of error in writing it), it un-pays for itself. And what was the benefit to begin with? Occasionally getting to skip some typing? Feeling clever by combining statements?

I think it's foolish.

4 comments:

Sally said...

Your prof's arguments (and yours) seem quite reasonable to me.

The Evil V.D'O. also doesn't seem to realize that one does, unfortunately, frequently have to work with people who one may deem "worthy of contempt" for not being quite up to snuff; many other coworkers will simply think of things in a different way and find one's own "clever" shortcuts confusing. Indeed, there are days one will look at one's own work and say "What the hell was I trying to do here again?" Clarity and a minimization of opportunities for error benefit everyone.

Tam said...

Yes. I think this guy has probably underestimated how easy it is to be confused by one's own code, despite knowing the language. Can't you write a sentence so abstruse that you would have a slightly hard time read it back?

I like the quote - perhaps by Dijkstra - something to the effect that, "Debugging is harder than coding, so if you write code as cleverly as you can, you won't be able to debug it."

Anonymous said...

I don't think this guy would have any trouble debugging this part of the program. He knows that code as well as everyone else has to learn to know what n = n + 1 means. Just how ignorant do you have to expect the next people to be?

I agree that you have to expect people to be rather ignorant if for no other reason than that the next person may be used to the current common programming languages and not your by-then-archaic old thing.

Tam said...

See, I don't think the problem is that people might be ignorant and not know what the different operators actually mean. I think the problem is that decoding something like

if (a = ++k)

takes a lot more processing power and thus takes longer and is more error-prone than decoding

k++;
a = k;
if (a)

"Shorter and tricker to write, longer and trickier to read" is not a recipe for success, I think.

Your point about legacy programming languages is also a good one.