Bug with backtick

[MissileSilo]

Consider this code:

class jQueryCalendar
  saveDate: (inputField) ->
    alert 'Saving Date' + $(inputField).val()

  initCalendarObjects: ->
    params =
      onSelect: (dateText, inst) =>
        @saveDate `this`
    $('.timelineCalendar').datepicker params

After running coffee -c -b on this file, the following javascript is generated:

var jQueryCalendar;

jQueryCalendar = (function() {

  function jQueryCalendar() {}

  jQueryCalendar.prototype.saveDate = function(inputField) {
    return alert('Saving Date' + $(inputField).val());
  };

  jQueryCalendar.prototype.initCalendarObjects = function() {
    var params,
      _this = this;
    params = {
      onSelect: function(dateText, inst) {
        return _this.saveDate(_this);
      }
    };
    return $('.timelineCalendar').datepicker(params);
  };

  return jQueryCalendar;

})();

Notice that what was in backticks incorrectly replaced. "this" was replaced with "_this"

A workaround is to put a space after the backtick: this instead of this

But it seems like this is a bug that needs to be fixed.

[michaelficarra]

This actually seems like a desirable behaviour to me. I know it doesn't insert what was literally inside the backticks, but it does reference the value you would think you'd be referencing when writing it. By that, I mean the bound context, referenced as this or @ in the CoffeeScript and referenced as _this in the output JS. Regardless, this likely wasn't an intended compilation, and we should change it to be consistent with the rule that backticks are inserted literally.

[codelahoma]

A bare this within the onSelect method would just reference the params object, wouldn't it?

Looks like CoffeeScript creates _this under the assumption that references to this or @ inside a class declaration are intended to refer to the class instance, no matter how they're nested.

Whether I'm right or wrong, I agree that backtick behavior should always be to literally insert, since the whole point is to be able to do something CoffeeScript can't or doesn't do.

[satyr]

An outcome of using Literal node for too many things.

$ coffee -bsc
  class A
    `this`
    'constructor': ->

var A;

A = (function() {

  A.name = 'A';

  function A() {}

  A;


  A.prototype['constructor'] = function() {};

  return A;

})();

[michaelficarra]

Yeah, that's what I figured.

[shesek]

@codelahoma,

Looks like CoffeeScript creates _this under the assumption that references to this or @ inside a class declaration are intended to refer to the class instance, no matter how they're nested.

CoffeScript doesn't assume anything, it only does that when you're using fat-arrow functions.

A bare this within the onSelect method would just reference the params object, wouldn't it?

No. Without the fat-arrow function, it would refer to the jQuery object.

@MissileSilo While it is a bug that should be fixed, (ab)using backticks like that to access the "unbound" value of this is something that you really shouldn't be doing. If you chose to use a fat-arrow function, you should accept the fact that you're losing access to the "original" this value instead of using workarounds like those.

Other than making the code extremely hard to understand (two different values for this, depending on backticks being used or not), its also very implementation-specific (The _this thingy is somewhat new, CoffeeScript used to bind the function instead - in which case both this and this would have the same value).

Just manually define an that = this, use that when referring to your jQueryCalendar instance and this when referring to $('.timelineCalendar').

[bato3]

As @shesek said:

Other than making the code extremely hard to understand (two different values for this, depending on backticks being used or not)

I'm not convinced, because we have escaped this and @.

I find hack:

  @saveDate ` this` #extra space

Now what do you say? Now it's a mess. :D

[satyr]

I prefer (this) for readability. Bugs breed creativity.