Minimal JS Hello World

[Nicolas Cannasse]

Hi,

With the latests patches (r4771+), the JS output for a
trace("Hello World") simple class is now the following :

-----
var $_;
var Test = function() { }
Test.main = function() {
console.log("Hello World");
}
Test.main();
-----

(don't forget compile with --dead-code-elimination !)

This required quite important changes in the JS code generator, some
testing might be require to ensure that we didn't break anything !

Best,
Nicolas

[fintan boyle]

I was hopeful that maybe hello world in haxe js would be <50 lines by the end of the summer!  I have to revise my expectations :-)

brilliant

--
To post to this group haxe...@googlegroups.com
http://groups.google.com/group/haxelang?hl=en

[Mihail Ivanchev]

That's awful lot of code, it takes 3 minutes to parse in the webkit browser of Android 2.3.5. Back to the drawing board!

[Philippe Elsass]

I suppose you're back to 200 lines as soon as you use reflection.
And with one iterator?

[Nicolas Cannasse]

Le 04/06/2012 17:31, Philippe Elsass a �crit :

> I suppose you're back to 200 lines as soon as you use reflection.
> And with one iterator?

You're welcome to make your own tests ;)

As for iterator it depends, most of the for...in should not add any
overhead.

Best,
Nicolas

[clemos]

Hi Nicolas,

I've updated try-haxe:
http://try-haxe.aaaliasing.net/

It seems to break when you only use externs.
(js is not defined...)
which breaks my example :p

Cheers
Clément

On Mon, Jun 4, 2012 at 5:16 PM, Nicolas Cannasse
<ncan...@motion-twin.com> wrote:

[Franco Ponticelli]

@Nicolas: Amazing!

[clemos]

PS:
On a slightly un-related topic, reading this change:
http://code.google.com/p/haxe/source/detail?r=4769
I was planning to file a issue about that,
but I would rather expect the opposite, @:runtime becoming the default
behaviour...
inline members would always have their runtime counterpart (to go
along well with Reflect by default, for instance)
these could eventually be removed by --dead-code-elimination
and kept using @:keep.

What do you think ?

Cheers,
Clément

[Nicolas Cannasse]

Le 04/06/2012 17:43, clemos a �crit :

> Hi Nicolas,
>
> I've updated try-haxe:
> http://try-haxe.aaaliasing.net/
>
> It seems to break when you only use externs.
> (js is not defined...)
> which breaks my example :p

Fixed in r4773, thanks for the report ;)

Best,
Nicolas

[Nicolas Cannasse]

Le 04/06/2012 18:00, clemos a �crit :

> PS:
> On a slightly un-related topic, reading this change:
> http://code.google.com/p/haxe/source/detail?r=4769
> I was planning to file a issue about that,
> but I would rather expect the opposite, @:runtime becoming the default
> behaviour...
> inline members would always have their runtime counterpart (to go
> along well with Reflect by default, for instance)
> these could eventually be removed by --dead-code-elimination
> and kept using @:keep.
>
> What do you think ?

@:runtime require explicit support in the backend. For instance in genJS
we will replace all potential accesses to .iterator by $iterator() call
that will add a runtime type check for Array.

I'm not sure we want to generalize such tricks.

Best,
Nicolas

[clemos]

On Mon, Jun 4, 2012 at 6:12 PM, Nicolas Cannasse
<ncan...@motion-twin.com> wrote:

Mh I didn't read it was about extern methods, so I think I don't
really get what @:runtime is about...

Still, about "normal" inline methods,
I just tested, and it appears that (in JS at least) inline methods are
created just like normal methods.
That's a good behaviour, but as these method are very likely to be
unused, I'd expect them to be removed when using DCE.

Do you want me to file a report for that ?

BTW I updated try-haxe to latest SVN.

Regards,
Clément

[clemos]

On Mon, Jun 4, 2012 at 6:10 PM, Nicolas Cannasse
<ncan...@motion-twin.com> wrote:

> Le 04/06/2012 17:43, clemos a écrit :
>
>> Hi Nicolas,
>>
>> I've updated try-haxe:
>> http://try-haxe.aaaliasing.net/
>>
>> It seems to break when you only use externs.
>> (js is not defined...)
>> which breaks my example :p
>
>
> Fixed in r4773, thanks for the report ;)

Thanks,
Same thing for js.SWFObject...
I didn't test with other externs yet, but...
are we supposed to add this @:initPackage everywhere ?
:S

Regards,
Clément

[Nicolas Cannasse]

Le 04/06/2012 18:21, clemos a �crit :

> On Mon, Jun 4, 2012 at 6:12 PM, Nicolas Cannasse
> <ncan...@motion-twin.com> wrote:

>> Le 04/06/2012 18:00, clemos a �crit :

>>
>>> PS:
>>> On a slightly un-related topic, reading this change:
>>> http://code.google.com/p/haxe/source/detail?r=4769
>>> I was planning to file a issue about that,
>>> but I would rather expect the opposite, @:runtime becoming the default
>>> behaviour...
>>> inline members would always have their runtime counterpart (to go
>>> along well with Reflect by default, for instance)
>>> these could eventually be removed by --dead-code-elimination
>>> and kept using @:keep.
>>>
>>> What do you think ?
>>
>>
>> @:runtime require explicit support in the backend. For instance in genJS we
>> will replace all potential accesses to .iterator by $iterator() call that
>> will add a runtime type check for Array.
>>
>> I'm not sure we want to generalize such tricks.
>
> Mh I didn't read it was about extern methods, so I think I don't
> really get what @:runtime is about...

Yes, it's only to prevent printing an error saying that the method is
actually not present at runtime.

> Still, about "normal" inline methods,
> I just tested, and it appears that (in JS at least) inline methods are
> created just like normal methods.
> That's a good behaviour, but as these method are very likely to be
> unused, I'd expect them to be removed when using DCE.
>
> Do you want me to file a report for that ?

Oh I thought it was done already, it's now fixed in r4776

Best,
Nicolas

[Nicolas Cannasse]

Le 04/06/2012 18:27, clemos a �crit :

> On Mon, Jun 4, 2012 at 6:10 PM, Nicolas Cannasse
> <ncan...@motion-twin.com> wrote:

>> Le 04/06/2012 17:43, clemos a �crit :

>>
>>> Hi Nicolas,
>>>
>>> I've updated try-haxe:
>>> http://try-haxe.aaaliasing.net/
>>>
>>> It seems to break when you only use externs.
>>> (js is not defined...)
>>> which breaks my example :p
>>
>>
>> Fixed in r4773, thanks for the report ;)
>
> Thanks,
> Same thing for js.SWFObject...
> I didn't test with other externs yet, but...
> are we supposed to add this @:initPackage everywhere ?
> :S

Well, the "js" package was always created by js.Boot before, and it's no
longer the case :)

Fixed for both remaining cases.

Best,
Nicolas

[clemos]

I updated to latest SVN,
now the example is only 11 lines :)
http://try-haxe.aaaliasing.net/
I've also cleaned a bit the presentation since it appears the word is
already spread... :S
https://plus.google.com/113704686911055424796/posts/hFsL4bpTmzK

Cheers,
Clément

On Mon, Jun 4, 2012 at 6:44 PM, Nicolas Cannasse
<ncan...@motion-twin.com> wrote:

[Franco Ponticelli]

@clemos, quick feature request: put the build status also below the code output ... I like to see the generated code (and errors) without having to switch back to the normal output. Thank you for the effort ... eager to see how it evolves!

[clemos]

Hi,

I did it and added some info...

Regards,
Clément

[Nicolas Cannasse]

Le 04/06/2012 18:54, clemos a �crit :

> I updated to latest SVN,
> now the example is only 11 lines :)
> http://try-haxe.aaaliasing.net/
> I've also cleaned a bit the presentation since it appears the word is
> already spread... :S
> https://plus.google.com/113704686911055424796/posts/hFsL4bpTmzK

Just got rid of var $_ in r4780 ;)

Best,
Nicolas

[Joshua Granick]

Nice!

Good work :)


On Mon, 04 Jun 2012 10:45:47 -0700, Nicolas Cannasse
<ncan...@motion-twin.com> wrote:

[Franco Ponticelli]

Thank you very much, appreciated! ... another suggestion: add SHIFT+ENTER (or any other combo) to CodeMirror to trigger the Run action. It is very useful to avoid switching from keyboard to mouse ;)

[clemos]

Disclaimer:
I'm (still...) not trying to compete with FlashDevelop
;)

Regards,
Clément

On Mon, Jun 4, 2012 at 7:51 PM, Franco Ponticelli

[Franco Ponticelli]

I know ... just throwing ideas ... feel free to ignore such suggestions ;)

[Andreas Mokros]

Hi.

On Mon, 04 Jun 2012 19:45:47 +0200
Nicolas Cannasse <ncan...@motion-twin.com> wrote:
> Just got rid of var $_ in r4780 ;)

Really great work, folks!
But Reflect.field gets not included...

@Nicolas: Did you see the bug, when compiling haxelib?

--
Mockey

[Hudson Ansley]

Wow, this is really impressive now getting that hello-world example
down to 10 lines (will be after Nicolas' last fix ;-)
and the js code is much less confusing this way also...
Regards,
Hudson

ps - just to make sure it was not going to bloat with something
slightly more complicated, this 11 line code only grows to 20

class Test {
  public static function main(){
    var str;
    var bodyHtml = "";
    var praise = ["amazing", "great", "time-saving", "coding joy"];
    for (str in praise) {
      bodyHtml += "Haxe is "+str+" :)<br>";
    }
  new js.JQuery("body").html(bodyHtml);
  }
}

[Philippe Elsass]

Looks like there's a bug in the metas generation.

When using Std.is(v,T) I see in the code: 

Test.__name__ = true;

instead of the usual:

Test.__name__ = ["Test"];

BTW I see trace() is nicely compiled as console.log(), but there isn't a sanity check for browsers not having a console - if it's undefined a do-nothing fallback should be created.

  }
}

--
To post to this group haxe...@googlegroups.com
http://groups.google.com/group/haxelang?hl=en

--
Philippe

[Nicolas Cannasse]

Le 04/06/2012 22:49, Philippe Elsass a �crit :

> Looks like there's a bug in the metas generation.
>
> When using Std.is(v,T) I see in the code:
> Test.__name__ = true;
>
> instead of the usual:
> Test.__name__ = ["Test"];

Yes it's normal. There are cases when we need to know if it's a Class
(has __name__), without needing the full name.

> BTW I see trace() is nicely compiled as console.log(), but there isn't a
> sanity check for browsers not having a console - if it's undefined a
> do-nothing fallback should be created.

As soon as you use haxe.Log.trace (for example by rebinding it) it will
use it instead of console.log

Best,
Nicolas

[Nicolas Cannasse]

Le 04/06/2012 20:36, Andreas Mokros a �crit :

> Hi.
>
> On Mon, 04 Jun 2012 19:45:47 +0200
> Nicolas Cannasse<ncan...@motion-twin.com> wrote:
>> Just got rid of var $_ in r4780 ;)
>
> Really great work, folks!
> But Reflect.field gets not included...

? uh ?

> @Nicolas: Did you see the bug, when compiling haxelib?

Yes, looks quite tricky, tomorrow maybe.

Best,
Nicolas