experimental commit: derive names for function expressions when we can

[michaelficarra]

Related issues:

• Named Functions jashkenas/coffeescript#15
• script/coffeescript tags don't work in IE 8 jashkenas/coffeescript#366
• The function keyword jashkenas/coffeescript#758

and many more. See also https://github.com/jashkenas/coffee-script/wiki/FAQ

Any objections before I merge this into master?

[vendethiel]

So I guess it doesn't harm on IE ? Could be possible to change arguments.callee inside the function body?

[michaelficarra]

@Nami-Doc: only the non-standard name property, about which I don't care. @satyr actually recommended something like this for CS back in the day, I believe.

[epidemian]

Looks good to me. It will probably help a lot with debugging; making stack traces more readable :)

It's also good to know that a minifier like UglifyJS can get rid of the unnecessary function name for production code:

$ echo 'foo = -> 99' | bin/coffee --js
// Generated by CoffeeScript 2.0.0-dev
void function () {
  var foo;
  foo = function foo$() {
    return 99;
  };
}.call(this);

$ echo 'foo = -> 99' | bin/coffee --js | uglifyjs --compress --mangle
void function(){var n;n=function(){return 99}}.call(this);

The built-in Redux minifier leaves it as a named function expression though (maybe a missing option to esmangle?):

$ echo 'foo = -> 99' | bin/coffee --js --minify
void function(){foo=function a(){return 99}}.call(this)

Now that i look at it, the minified output also omits the var declaration inside the auto-invoked function, making foo leak into the global scope. Is that an intended behavior or a bug?

[mark-hahn]

foo = function foo$() {

I thought jeremy vetoed this because it didn't work in IE.

[vendethiel]

He veoted function foo(){}

[mark-hahn]

I hope I don't have to try to find something in coffeescript issues,
because that is impossible. But I distinctly remember foo = function foo$() { being vetoed. I know for a fact it doesn't work in old IE.

On Sun, Dec 30, 2012 at 4:30 PM, Nami-Doc [email protected] wrote:

He veoted function foo(){}

—
Reply to this email directly or view it on GitHubhttps://github.com//pull/131#issuecomment-11770659.

[michaelficarra]

@mark-hahn: The related issues are all linked in the linked FAQ. I'm pretty confident I understand the IE bug. It has been around and well known for quite a while.

var f = function g(){};

behaves like this in IE6:

function g(){}
var f = g;

So the name g is exposed to the containing scope in IE6, not just the function body's scope.

edit: Also, f !== g, so memory use will be higher in IE6. But nothing will break because of the change.

[mark-hahn]

Sorry. I somehow missed the beginning of this thread.

[michaelficarra]

@epidemian: I think you've stumbled upon an esmangle bug. I've reported it and it is linked above.

[vendethiel]

@mark-hahn actually, it's fairly easy :p

[michaelficarra]

I'm going to close this because of general FUD and because I don't want this change affecting the merge into jashkenas/coffee-script.

[devinrhode2]

Hey Micheal, FWIW I thought of a cleaner way to name object methods:

Module.method = function Module_method() {};

The Module_ prefix is important so you never mistake one function for another