Please introduce explicit shadowing

[a-square]

A (simplified) real life example of why it's important: https://gist.github.com/Kallikanzarid/4736201

I see no reason to take away a feature of JavaScript that's useful, not broken, and cannot be misused.

Edit: also, for some reason many people claim that the same problem exists in Scheme. It doesn't, as the code in my gist illustrates.

[vendethiel]

That's Ruby scoping for you. See #1121 etc. If you really need shadowing, use do (or backticks).

[a-square]

Ruby scoping is one bad thing about Ruby, why imitate it for the sake of imitating it, especially when you can introduce explicit shadowing as an option?

Edit: to counterweight Ruby, consider that fact that pretty much every successful language has support for explicit shadowing: C, Pascal, Fortran 90, Java (as much as it has lexical scope), C#, Scheme, ML, Haskell, Erlang, Prolog - in short, pretty much any language I know and pretty much any language people care about, except Ruby, non-strict Perl and CoffeeScript.

There are good reasons why explicit shadowing should be in every lexically scoped language, mainly because with minimal effort it allows us to make sure we don't need to know anything about preceding code to reason about a function's behavior (unless it has free variables, in which case we obviously can't). As a consequence, it ensures that most functions can be rearranged without breaking anything, thus the code is less brittle.

[mehcode]

I'm all for allowing the let keyword to prefix variable declarations so they are enforced to be block-local (including shadowing any preceding declarations).

x = 42
foo = ->
  let x = 12
  x

# x === 42
# foo === 12

[jashkenas]

Sorry -- avoiding of shadowing is a choice we want to pursue as a design decision ... as I'm sure you're well aware. Take a look at Coco or Livescript for alterna-versions that embrace shadowing.

[a-square]

But you do shadow variables, it's lexical scoping! You just for some reason decided to gut our ability to do explicit shadowing.

[jashkenas]

Lexical scoping is about which variables are possible to reference across scopes, not about whether you tend to shadow your variables or not. It's well possible to program without shadowing, by choice, in a lexically scoped language.

You just for some reason decided to gut our ability to do explicit shadowing.

I did indeed. For the reasons why, see previous tickets.

[a-square]

It's well possible to program without shadowing, by choice, in a lexically scoped language

In the land of magic elves, yeah.

[epidemian]

In the land of magic elves, yeah.

That's not a very compelling nor constructive argument 😕

Also, i fail to see what would the problem be in your example. If the content variable would be global, i.e. a property of window, then that assignment wouldn't touch it (you have to explicitly do window.content = ...). The only way in that that assignment would not be declaring a new variable in that scope is if the content would be declared in an enclosing scope; but that's not the global scope, it is, at most, the scope of the file (and it would be a very poor decision to use such a generic name as "content" for a variable with very little context IMHO).

[a-square]

Some files have more than one author over time.

[juliankrispel]

I'd love to learn more about the reasoning behind that design decision @jashkenas. Is that in writing anywhere?

[vendethiel]

Probably in one of the thousands issues about this topic :p.

[juliankrispel]

@Nami-Doc not really, couldn't find anything comprehensive in the issues. I'm not really that keen on reading thousands of emotional comments about why coffeescript should behave like this or that, I'm just interested in the decision-making, would appreciate any pointers.

[michaelficarra]

Explicit shadowing is available through CoffeeScript's version of let: do.

x = 'outer'
do (x = 'inner') =>
  x is 'inner' # true
  x = 'still inner'
x is 'outer' # true

[vendethiel]

@juliankrispel I'm always the one who get to go through the issues! Fine. #712 and #1121

PS : github issue autocompletion seriously sucks :/