I don't know why I didn't show you this before.
I have done some tests AS3 vs JS a while ago. Pretty much on the
beginning of my journey with JS when started to investigate possibility
of moving towards html5.
http://www.flaemo.com/bixbite/
you can also review about 86 different specs collected up there
http://flaemo.com/bixbite/gettable.php?id=*86*
Just change the id to review them all.
When comes to speed of execution of generic parts, raw loop iterations,
object creations etc, Flash is no longer a king on desktop.
But still better in some crucial elements. Take a look at Object vs
Class creation and speed of accessibility.
The only thing that Flash is 10x faster with, is the very last part of
the test. Display List vs DOM. However is not true for Canvas or even
svg could get closer if not better.
Number of iteration is quite small up there to see differences between
Scope Object 2 and 3 lvl because this test has been tweaked to run on
Android devices as well.
But it will give some picture of where we are.
Dan
On 12/14/2012 11:49 AM, Michael Schmalle wrote:
Quoting Daniel Wasilewski <devudes...@gmail.com>:
Well, thanks for clarification.
And I love you keep AST abstract as it should be :)
Imagine in the future JS 2.0 will came out...
Well any good design is not dependent on it's output. This design is
no different, the only thing it's dependent on is it's AST structure. ;-)
'goog' seems to be the reasonable way to go. I would be a bit concern
about name-space implementation yet.
org.apache.data.blah.foo is the major bottleneck. And currently not
that different from haxe approach. Even after closure compiler heavy
lifting it is still
a.b.c.d.e.f. It starts with "org.apache.data.blah.foo" as a string ,
so why not ns_orgApacheDataBlahFoo that will get complied to 'abcdef'
? Just an idea.
You know, a week or two ago when this was brought up, I was going to
suggest doing the same thing?
Why don't people do it in other frameworks? If it's a performance
issue to use nested object prototypes than;
org_apache_flex_foo.MyClass.prototype should never get collisions. I
don't know enough JavaScript theory to be useful though. :)
That is why I took a week of MY time to get up in svn a valid
alternate cross compiler, I want to give people hope in a new
direction so it can galvanize a new effort.
And I am glad people like you are here. I can step in at some point
but I am ridiculously busy at the moment... and towards the end of
2012 :(
Isn't it the end of 2012 right now? ;-)
Mike
