[dojo-contributors] dojo.lang.inherits

Bill Keese bill at dojotoolkit.org
Wed Apr 5 20:24:49 EDT 2006


>> 2. there is no separation between constructor initialization and instance
>> initialization.

The suggestion is something similar to this, right?

dojo.inherits(subclass, superclass, fnInit);
dojo.extends(subclass, { param1: ..., func1: ... } );

(where fnInit is the constructor in the java sense of the word)

This is really abusing the meaning of "inherit".  Wouldn't it be better 
to combine these two functions into one function like this?

dojo.defineClass(subclass, superclass, fnInit,
  { param1: ..., func1: ... });

As a matter of fact, Alex has already done this for widgets (see 
http://archive.dojotoolkit.org/nightly/src/widget/Dialog.js )

PS: alternately you could use this syntax:

dojo.defineClass(subclass, superclass,
	{ constructor: ..., param1: ..., func1: ... });


> I'd be interested in your thoughts on this:
> 
> http://dean.edwards.name/weblog/2006/03/base/
> 
> I wrote it to alleviate some of the problems you refer to below.
> 
> -dean

Yeah, there was some discussion about this and some objections about not 
staying true to javascript, but I didn't really understand it.  (The 
current dojo stuff is just a thin wrapper on top of javascript and 
people wanted to keep it that way, but not sure why.  Maybe just 
philosophical reasons.)  Hopefully someone can explain in more detail.

Bill

> On 5 Apr 2006, at 23:51, Scott J. Miles wrote:
> 
> At the Dojo meeting today (4/5/2006), dojo.inherits came up and we 
> agreed to
> open up the discussion to the general contribs list.
> 
> There are two (arguable) faults in the dojo.inherits implementation:
> 
> 1. invoking super class methods is wordy and inconvenient.
> 2. there is no separation between constructor initialization and instance
> initialization.
> 
> There is general agreement that we want to 'let JS be JS'.
> 
> With regard to (2), my personal opinion is that JS doesn't have an axe to
> grind with respect to initialization, and there is nothing un-JS about
> supporting a systemic instance initialization function (that is to say, a
> standard way of specifying an instance initializer).
> 
> Alex seemed right-away to have an idea of how tweak dojo.inherits to add
> such a feature (yay). We briefly discussed adding additional parameters to
> inherits vs. a key-word style signature. I prefer the former, but there was
> no general agreement.
> 
> There was some talk about the term 'inherits' implying a non-JS idiom, but
> this is not a problem IMO. Any suggestions for new nomenclature are
> appreciated.
> 
> Re (1), nobody seemed to have a handle on how to make calling
> inherited/parent/base-class methods easier. Morris has developed his own
> inheritence system that may provide a solution.
> 
> Thanks for listening. :)
> 
> Regards,
> Scott J. Miles
> TurboAjax Group
> http://www.turboajax.com
> 
> _______________________________________________
> dojo-contributors mailing list
> dojo-contributors at dojotoolkit.org
> http://dojotoolkit.org/mailman/listinfo/dojo-contributors



More information about the dojo-contributors mailing list