no way of getting constructor name

[frank06]

I believe it's currently not possible to obtain a class' constructor name.

I'd like to do the following:

class Persistence
  save: () =>
    # apply some string manipulation wizardry and persist to database
    console.log "Saving #{extractTableName @constructor.toString()}"

class Animal extends Persistence
class Horse extends Animal

(new Animal).save()
(new Horse).save()

Right now @constructor.toString() will not yield the name because the constructor is an anonymous function assigned to a variable - and therefore doesn't have a name.

Would it be feasible to name the constructor function, in addition to the assignment, like so?

Horse = function Horse() {
  return Animal.apply(this, arguments);
};

[michaelficarra]

I do recall someone bringing up a scoping issue regarding this in another issue, but I can't seem to remember which one. If #640 (executable class bodies) is ever implemented, my proposal should fix this problem.

[zmthy]

Yeah. We had this and it was really cool, but there's a bug in IE6 with scoping when you have a named function, so it was removed.

[frank06]

@michaelficarra: Cool, i'll be following #640 (+1 to executable class bodies)

@Tesco: Aww. I hate IE6 even more.

I really hope this gets reintroduced into coffee.

[StanAngeloff]

We shouldn't seriously support IE6 any more, should we? I mean the beast is 9 years old and the last time I recall seeing someone even in a corporate environment using it was two or three years ago. Time to move on... I wonder if IE7 still has the leaky implementation of named functions.

[zmthy]

Sadly, yes.

It's also quite sad that even last version of JScript — 5.8 — used in Internet Explorer 8, still exhibits every single quirk described below.
http://kangax.github.com/nfe/

[zmthy]

I've had a crack at a fix for this. #729
Haven't tested if it fixes the IE problem, but it should do because it ensures the constructor is always a declaration and never an expression.

[danielribeiro]

This and proposals that solve this have been discussed and shot down way too many times (add #684, #675 do the list)

Using the createDisplayName function on #675, tweaked to check if the variable is a prototype and its prototype is not empty, will actually name all classes. Coffescript or not. On the example, "clas.includes && clas.includes(JS.Kernel)" check is for JS.Class only.

[sstephenson]

Here's an idea. CoffeeScript can call the extended hook with a second argument—the subclass' identifier as a string. Then you can write a base class and keep track of subclasses' names however you want.

Implemented here: https://github.com/sstephenson/coffee-script/compare/extended-child-name

[danielribeiro]

Great and simple tweak sstephenson. I'd normally argue that explicitely requiring to sublcass a base class is a problem, however, this is not so on my project (most classes already have to extend a base class).

[jashkenas]

I've merged in a variant of Tesco's fix for class names, and you can now access the "name" property of the constructor function, at the cost of having slightly more verbose compiled output... closing the ticket.

[frank06]

Great! Will start using it asap.

[jashkenas]

Sorry folks, I had to revert this commit because (surprise, surprise) it didn't work in Internet Explorer. Things are back as before on master at the moment.

[jashkenas]

And one more update. Tesco was quite right about the pure-function-declaration-ness being critical for IE to work. Here's a patch that puts it back that way, and works in internet explorer:

https://github.com/jashkenas/coffee-script/commit/acc06d772a731e65abdd572c9c5a0e452712df5f

[danielribeiro]

I've been working around by using the plugin mechanism, and changing the source code. On 0.9.5, function.name should work ok. With one caveat. function.name is non-standard, and will not work on IE. Guess I'll still stick with metaprogramming to make coffeescript do things they way I wish it did.

[satyr]

@danielribeiro: Coco amends that problem by explicitly setting name.
$ coco -bpe 'class C'
var C;
C = (function(){
function C(){} C.name = "C";
return C;
}());

[ghost]

This can be achieved with a base class if required, it uses a RegExp which isn't ideal but should work in IE.

class Base
  @getName: ->
    this.toString().match(/^function\s(.+)\(/)[1]

class Foo extends Base

alert Base.getName() # alerts 'Base'
alert Foo.getName()  # alerts 'Foo'

[gavJackson]

(For future Googlers...)

Expanding on soniciqs answer above, here's how I determine the class name in IE without base classing:

if clazz.constructor.name?
    classRef = clazz.constructor.name
else
    classRef = clazz.constructor.toString().match(/^function\s(.+)\(/)[1]

This can be single-lined (if thats the way you roll) to:

classRef = if command.constructor.name? then command.constructor.name else command.constructor.toString().match(/^function\s(.+)\(/)[1]