Hacker News Reader

Comment by windsurfer | Hacker News Reader

The best tip I've ever received is "Try not to write comments. Make your code speak for itself."

And I write a lot of perl.

I came here to say exactly the opposite.

I read a lot of code these days where people seem to think that descriptive variable and function names replace the need for any comments. But, even the most well written code is only going to make it clear _what_ the code it doing. The _why_ is likely to remain obscure. And without knowing why the code does what it does, it will often be impossible to know whether you can make certain changes.

The _why_ is likely to remain obscure.

The why might remain obscure, but that depends on two very important things: your documentation, as you noted; and your data structures.

Show me your flow charts and conceal your tables and I shall continue to be mystified, show me your tables and I won't usually need your flow charts; they'll be obvious. - Fred Brooks

As much as you can organize your data to make the purpose of it obvious, do that. It will make the system far easier for a new-comer to comprehend, and will make far more obvious to do you the sort of code you need to write to make the wheels turn. Where that is not enough, then use comments, but don't assume that prose is an adequate replacement for precise and logical data representation. Often times, I've found that the latter is a far better choice.

This is the reason why I am kind of laughed at, and why my comments are kind of feared, because I either don't need comments at all, but once I need comments, they usually end up being a page of things you need to know about this unit.

You make a good point about why vs. what. Still, my experience has been that if I ask myself when adding a comment, "Could I make the code say this instead?" the answer is "yes" 9 times out of 10. It is preferable to do so because code is much likelier to be read and maintained than comments are. Also, this question often makes me see flaws in how the code was expressing itself, so I end up with better code.

I think the best comment is one which explains an ugly edge case which forces the code to look like it does.

//window.update; is not used because of API bug #12345 see: bugs.mysql.com/12345 so use xyz instead.

Nine times out of ten I can read and understand good code with little effort, but code often starts out looking clean and becomes messy for a reason.

I have another version that I prefer: "Try not to write comments. Make your code speak for itself. Then comment it."

Best of both worlds.

Bravo if you can accomplish that with perl. I haven't hacked in perl in a long time but, from what I remember, you would have to resist using a lot of perl's shortcuts in order to make it speak for itself.

That depends a lot on _who_ you're expecting the code to "speak" too. If I were writing code and expected non-perl programmers to read it, I might avoid using Schwartzian Transforms, but the reductio ad absurdum arguement for that point of view says you could only ever write code in plain english[1] because you might someday need it to be fixed by someone who doesn't know language-de-jour.

If a Perl program needs fixing, you need to get someone who knows Perl to fix it, which means they _should_ know the standard set of Perl idioms. You _can_ write Perl in a way that Cobol programmers can easily understand and change, but you should never have too...

big

[1] or perhaps chinese to maximise the probability of being understood by a random maintenance programmer?

Reply on news.ycombinator.com