declaring a variable twice in IIFE
Asked Answered
G

2

20

I came through this fun quiz on the internet.

console.log((function(x, f = (() => x)){
  var x;
  var y = x;
  x = 2;
  return [x, y, f()]
})(1))

and the choices were:

  1. [2,1,1]

  2. [2, undefined, 1]

  3. [2, 1, 2]

  4. [2, undefined, 2]

I picked solution 2 TBH, basing that on that x has been redefined, y was declared and defined with no value, and that f has a different scope hence getting the global x memory spot than function x memory spot.

However, I tried it in jsbin.com

and I found it was solution 1, while I was not sure why that happened I messed with the function body and I removed var x from the function body, I found that the response changed to #3 which makes sense as x value changed and hence it showed x and f as 2 and y as 1 which was declared globally.

but still I can't get why it shows 1 instead of undefined.

Grays answered 7/4, 2019 at 14:28 Comment(8)
I find the best way to figure these things out is to step through them line by line with a debugger, and/or print out the values after each line.Juratory
var x; doesn't define a new variable within the function scope, where var x = somevalue; wouldEnsnare
@alex it does. ...Petrapetracca
@JonasWilms I see, would have expected x to be undefined then, implicitly copy the function formal value looks really odd to me.Ensnare
@alex it has its reasons.Petrapetracca
I'm glad this question was asked :)Houdon
Related: Scope of Default function parameters in javascriptHensley
I first read "declaring a variable twice in LIFE". Seems doable.Hydrozoan
P
24

but still I can't get why it shows 1 instead of undefined.

It's not just you. This is a deep, dark part of the specification. :-)

The key here is that there are two xs. Yes, really. There's the parameter x, and there's the variable x.

A parameter list containing expressions (like f's default value) has its own scope separate from the function body's scope. But prior to parameter lists possibly having expressions, having var x within a function with an x parameter had no effect (x was still the parameter, with the parameter's value). So to preserve that, when there's a parameter list with expressions in it, a separate variable is created and the value of the parameter is copied to the variable at the beginning of the function body. Which is the reason for this seemingly-odd (no, not just seemingly) odd behavior. (If you're the kind who likes to dive into the spec, this copying is Step 28 of FunctionDeclarationInstantiation.)

Since f's default value, () => x, is created within the parameter list scope, it refers to the parameter x, not the var.

So the first solution, [2, 1, 1] is correct, because:

  • 2 was assigned to the var x in the function body. So at the end of the function, the var x is 2.
  • 1 was assigned to y from the var x before x got the value 2, so at the end of the function, y is 1.
  • The parameter x's value has never changed, so f() results in 1 at the end of the function

It's as though the code were written like this instead (I've removed unnecessary parens and added missing semicolons):

console.log(function(param_x, f = () => param_x) {
  var var_x = param_x;
  var y = var_x;
  var_x = 2;
  return [var_x, y, f()];
}(1));

...I removed var x from the function body, I found that the response changed to #3...

#3 is [2, 1, 2]. That's correct, because when you remove the var x from the function, there's only one x, the parameter (inherited by the function body from the parmeter list). So assigning 2 to x changes the parameter's value, which f returns.

Taking the earier example with param_x and var_x, here's what it looks like if you remove the var x; from it:

console.log(function(param_x, f = () => param_x) {
  var y = param_x;
  param_x = 2;
  return [param_x, y, f()];
}(1));

Here's an annotated description of the original code (with the extraneous parentheses removed and missing semicolons added):

//                   /---- the parameter "x"
//                   v  vvvvvvvvvvv--- the parameter "f" with a default value
console.log(function(x, f = () => x) {
  var x;      // <=== the *variable* x, which gets its initial value from the
              //      parameter x
  var y = x;  // <=== sets y to 1 (x's current value)
  x = 2;      // <=== changes the *variable* x's value to 2
  //      +---------- 2, because this is the *variable* x
  //      |  +------- 1, because this is the variable y
  //      |  |   +--- 1, because f is () => x, but that x is the *parameter* x,
  //      |  |   |       whose value is still 1
  //      v  v  vvv
  return [x, y, f()];
}(1));

Final note regarding your title:

declaring a variable twice in IIFE

The variable is only declared once. The other thing is a parameter, not a variable. The distinction is rarely important...this being one of those rare times. :-)

Popular answered 7/4, 2019 at 14:37 Comment(5)
This is now in my list of "new weird things about JavaScript" alongside the semantics of (not yet really standard) instance property initialization expressions in class declarations.Houdon
@Houdon - Oh, this is much weirder than field declarations. :-DPopular
Well field declarations are jarring to me because this is the instance, not the this of the surrounding scope as with object initializers. I guess it's not really "weird" if you think of class declarations as being something akin to macro expansion.Houdon
@Houdon - Yeah. I think of them as code that's relocated to the beginning of the constructor (just after the super call if it's a subclass). That's what Java does with instance field initializers and instance initialization blocks (and they're literally copied to the beginning of each constructor in Java; thankfully in JavaScript we just have the one). JavaScript took effectively the same approach.Popular
@T.J.Crowder this is a great explanation, it is really weird that var x is assigned directly from the parameter x although there is no obvious assignments, thanks for clearing this out.Grays
H
2

The tricky part of that code is that the => function is created as part of a default parameter value expression. In parameter default value expressions, the scope includes the parameters declared to the left, which in this case includes the parameter x. Thus for that reason the x in the => function is in fact the first parameter.

The function is called with just one parameter, 1, so when the => function is called that's what it returns, giving [2, 1, 1].

The var x declaration, as Mr Crowder points out, has the (somewhat weird, at least to me) effect of making a new x in the function scope, into which is copied the value of the parameter x. Without it, there's only the one (the parameter).

Houdon answered 7/4, 2019 at 14:35 Comment(4)
@JonasWilms you're right; the var declaration must ... do something but it doesn't seem obvious to me what that is. It's as if the var declaration creates a new x in the function scope, but it clearly gets the value of the parameter x anyway (which I'd expect), leaving the parameter apparently in its own scope.Houdon
I guess the values are somewhat copied from the default initializers scope to the bodies scope. Im already digging into the specPetrapetracca
It makes sense. As noted in the spec: NOTE: A separate Environment Record is needed to ensure that closures created by expressions in the formal parameter list do not have visibility of declarations in the function body. thats reasonable.Petrapetracca
@Houdon - It's a very deep, dark part of the spec. :-)Popular

© 2022 - 2024 — McMap. All rights reserved.