Skip to content

Latest commit

 

History

History
37 lines (24 loc) · 2.75 KB

Enyo-and-Use-Strict.md

File metadata and controls

37 lines (24 loc) · 2.75 KB

Enyo and use strict

Why we don't work with "use strict"

Introduced in ECMAScript 5, the "use strict" pragma is a useful tool for limiting the code that will run in a JavaScript virtual machine to ensure compliance with certain practices that are considered to be safe. John Resig's article on strict mode provides some helpful background.

However, one of the language features that strict mode eliminates is the use of arguments.callee. arguments is a special variable, defined in every function, that provides access to the arguments used to call the function. This is often used to support variable-argument lists. JavaScript also provides properties on the arguments object, namely caller and callee, that are no longer accessible in strict mode. arguments.caller provides access to the function that called your code, while arguments.callee provides access to your own function object.

We currently do our "call our superkind's implementation" via the code

this.inherited(arguments, ...);

If you look in Enyo's Oop.js, the inherited method is implemented as follows:

enyo.kind.inherited = function(args, newArgs) {
    return args.callee._inherited.apply(this, newArgs || args);
};

This method looks at the callee property of the arguments object you passed in to find a special property called _inherited. When you create a new kind, one of the magic bits we do is iterate through all the methods you've defined on the kind and add this property pointing to the same-named method on the parent kind. If we didn't have access to arguments.callee, you'd have to change all of your method definitions to have a second name, e.g.,

onSetupItem: function onSetupItem(...) {

and the call to this.inherited would look like

this.inherited(onSetupItem, arguments);

That's a pretty big code change, and it's not one we're ready to make at this time.

If you're wondering why strict mode removes these things, there are some good reasons. One big one is that having access to the function that called you makes it easier to break out of your own code. This is important if you have sensitive code that's calling code that may be less trustworthy, so you don't want it to be able to inspect your state. JavaScript has a module pattern in which you can have variables only be visible to functions declared in the same scope, but with arguments.caller, you can pass a callback into one of those and that callback can then go and look around in the closure. arguments.callee is a problem because it lets you pass your own scope to other people; this message from es-discuss summarizes the problem.