loading...

re: Declaring JS Variables in 2019 VIEW POST

TOP OF THREAD FULL DISCUSSION
re: I’m in the “preventing re-declaration doesn’t really matter, and let is two characters shorter than const, so…” camp. People have suggested const ...
 

Do you mean mutable, vs re-declarable?

var is re-declarable in that I can write var x = 123; var x = 456; and that's totally valid. I suppose let is re-declarable in a lower scope (inside {}).

 

Ah yes, I meant re-assignable, not re-declarable.

The argument for const over let is that const “protects” you from unintended reassignment, but I don’t see that as a problem that requires extra measures.

let everywhere is simple. const and let together is more complex, so there needs to be a good reason for the added complexity to justify it. I don’t like making my code more complex just because something sounds like a cool feature.

People who argue for const haven’t convinced me yet that accidental reassignment is something that I need to guard against.

I hope that makes sense.

Would you use const for say, defining magic numbers at the top of your program?

const POST_TIMEOUT = 60 * 1000;


// ... much later
function blah() {
  setTimeout(foo, POST_TIMEOUT);
}

Just trying to understand if it fits into your world at all :)

const is most likely is a very useful feature in certain situations. But what are those situations? I can only guess. I personally don’t think that I need const because I don’t think that accidental reassignments are an issue in my own code.

So what are those situations? That’s what people arguing for const should focus on explaining, in my option.

I think I have a good analogy. In CSS, I argue for using this code:

[hidden] {
  display: none 1important;`
}

The !important part is similar to const. It protects the display property from being re-assigned if the hidden attribute is present. Why is that important? Because a <div class="foo" hidden> element could get accidentally rendered with something that is as simple as .foo { display: flex; }. That .foo style may not have intended to style that specific hidden element but it did match it, and now the hidden attribute is ignored and the element is rendered.

So I think I make a good case for adding !important in this particular situation.

I would like someone to make a good case for why POST_TIMEOUT needs to be const. What concrete problem can be prevented by that?

It's about two things: semantics - code is not just meant for the computer, it's also for humans who read it. Using const better signifies to team members and maintainers that this variable shouldn't change. Another is safety by default. Yes, maybe your code doesn't have issues with constants. But it's about baking in safety to your code to reduce the possibility of any errors. It's like how Java class variables are package-private by default (like @flame10a said) and it's highly recommended to make them private, and variables in Rust are actually immutable (non-reassignable) by default. So, in the future, if for example you're coding while running low on sleep, or one of your coworkers decides to work on it and he hasn't worked on the project before, you are a lot less likely to make a mistake.

I guess const makes (more) sense in large teams that work on large code bases (emphasis on I guess). I would like to see some practical evidence. If somebody was in a concrete situation where const saved them (and their team) from a nasty bug, I would like to read about it, so that I can get a feel of const’s usefulness.

If I started using const, I would have to decide every time when I declare a new variable, whether or not it should be re-assignable. This mental task seems unnecessary and annoying to me. I would rather spend my brain power on the actual problems that I’m trying to solve in my code.

Since I've switched to using const, I find that I almost never use let. In fact, the rarity of let makes it stand out in the code when it is used; it feels like a beacon to anyone reading the code that this variable is special, it's going to be updated later.

But I almost never update variables anyway. There are basically two cases where I use it:

  1. In an old-fashioned for (let i=0; i<num; i++) loop
  2. In a case where I'm setting the value inside of a block, either a complicated conditional situation or a try ... catch situation:
let x;
switch (some_var) {
  case 0:
    x='a';
    break;
  case 1:
    x='b';
    break;
  default:
    x='c';
}
let x;
try {
  x = methodThatMightThrow();
}
catch(_) {
  x = fallbackVal;
}

Neither of those are very common in my code, so I basically never use let.

I'm glad to hear you're on the same page as me! I like the shining beacon analogy.

For me, both those examples often warrant extracting behavior to some helper function, but I understand how you're using them there, too.

That sounds like a solid approach. I’m surprised that const works fine in (the header of) for-of statements but not in for.

Yeah, the difference seems to be that in a for ... of loop like this:

for (const thing of myThings) { /*...*/ }

you're essentially grabbing an iterator at the start of the loop, i.e. myThingsIter = myThings[Symbol.iterator](), then at the beginning of each loop running const thing = myThingsIter.next(). It seems to desugar to something like this:

const myThingsIter = myThings[Symbol.iterator]()
while (true){
  const {done, value: thing} = myThingsIter.next();
  if (done) break;

  /* ... */
}

You're always assigning a whole new value to the loop variable as opposed to modifying the loop variable as is typical in a classic for (let i=0; i<n; i++) loop.

Really the only reason you need a let in that loop is because you're reassigning the value with i++, equivalent to i = i+. In theory you can use a const if you're doing something super unusual, like using an object as your loop variable and modifying a property of it:

for (const o = {count: 0}; o.count < n; o.count++) {
    /* ... */
}

I don't know why anyone would ever want something like this, and there would undoubtedly be cleaner ways to write the code, but it works. You're never reassigning o, so no problem with it being a const.

code of conduct - report abuse