Smalltalk and Java

Dear Squeaker,

I am working on a new Smalltalk intended to be embedded in Java  
programs. Athena is an offspring of Smallworld.
You can embed Ruby, Python, Lua, Scheme in Java. You can achieve  
similar goals with Athena.

http://bergel.eu/athena

Comments will be welcomed...

cheers,
Alexandre

--
_,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:
Alexandre Bergel  http://www.bergel.eu
^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;.

For whom it might be of interest:
http://bugs.squeak.org/view.php?id=6739

I have accessed the subversion tree of squeak but found out that this
patch seem not to be intergated. I do not know of any other special
sources for 64-bit Linux systems so I can't tell whether this is
solved there or not

Regards
Friedrich

why I would like to have Smalltalk over Java? I can't understand it.


A question from a newbie

El 09/11/2007, a las 17:08, Bergel, Alexandre escribió:

On Fri, 2007-11-09 at 18:37 +0100, Giuseppe Luigi Punzi Ruiz wrote:
> why I would like to have Smalltalk over Java? I can't understand it.
>
Cell phones and embedded devices of that kind are very restricted.
To bring your own piece of software on the phone you would have to
code it native. After having coded your native program you must
install it on the phone. This is not easy as cell phones aren't
open platforms.
On modern phones there is a second option. They bring a java enabled
layer with them. That means there is a vm available where you can
install software. Then you are able to run a java program in a
sandbox which restricts access to the devices like gsm for sms
sending etc. That means to have a smalltalk vm on top of java you
enable every cell phone to use your software written in smalltalk.
It would even be possible to send it via MMS it to a friend.

hope this helps,

Norbert

> why I would like to have Smalltalk over Java? I can't understand it.
>
>
> A question from a newbie

The same reason you'd build any language over the JVM.  Portability.
Of course, being able to interact with the existing body of
JVM-.targeted code is also a nice bonus.

Brett.

Portability is one good reason. I have found another good reason few  
weeks ago when I saw the talk of Jim Hugunin on Iron Python (Python  
for .Net). Jim said that he experiences a significant improvement of  
Iron Python at each new version of .Net.

In the case of Athena this is not true (yet) since there is no JIT.  
But it would not be surprising if a smalltalk program compiled for a  
JVM has a better performance than the same program on the standard  
squeak VM.

The link of Jim's talk: http://www.swa.hpi.uni-potsdam.de/dls07/ 
presentations/JimHugunin_2007_DynamicLanguagesOnDotNetWithTheDlr_Dls.pdf

Cheers,
Alexandre


On 9 Nov 2007, at 19:14, Brett Kosinski wrote:

--
_,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:
Alexandre Bergel  http://www.bergel.eu
^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;.

Bergel, Alexandre wrote:
> In the case of Athena this is not true (yet) since there is no JIT. But
> it would not be surprising if a smalltalk program compiled for a JVM has
> a better performance than the same program on the standard squeak VM.

Interesting. To me, it would be surprising, simply because all I've seen
in that direction like Jython, or Dan's Squeak-on-Java had significantly
worse performance compared when run on top of the JVM compared to their
"native" interpreter.

However, I would be *extremely* interested in having a data point to the
opposite.

Cheers,
   - Andreas

Andreas Raab schrieb:
> Interesting. To me, it would be surprising, simply because all I've seen
> in that direction like Jython, or Dan's Squeak-on-Java had significantly
> worse performance compared when run on top of the JVM compared to their
> "native" interpreter.
That might be because both python and squeak require more flexible
method dispatch than java provides (and the JVM supports)

For IronPython (and IronRuby, and a whole bunch of other languages)
there's DLR (dynamic language runtime) now for .NET, which provides
optimized/-izing code for dynamic languages.

So far, all I've seen in the java ecosystem was talk about a VM
extension to support dynamic languages ("invoke dynamic" or something
like that), but no decision or code. But I didn't look to hard for it
either.

Executive Summary: .NET's VM might be a more performant environment for
dynamic languages than JVM


Regards,
Patrick Georgi

Patrick Georgi wrote:
> Andreas Raab schrieb:
>> Interesting. To me, it would be surprising, simply because all I've seen
>> in that direction like Jython, or Dan's Squeak-on-Java had significantly
>> worse performance compared when run on top of the JVM compared to their
>> "native" interpreter.
> That might be because both python and squeak require more flexible
> method dispatch than java provides (and the JVM supports)

Indeed, that was my impression.

> Executive Summary: .NET's VM might be a more performant environment for
> dynamic languages than JVM

Yes, I understand the same. And I wish it were different...

Cheers,
   - Andreas

Hi Alexandre,

Congrats and thanks.

It is even much more appreciated if you can add an option to bootstrap from
(Little) Smalltalk source files.

I know, only retards would talk about source code and files ;-)

NO, don't mean Timothy Budd's original idea of bootstrapping from source is
retarded ;-)

Maybe he would regret that he would never be able to claim that his Little
Smalltalk has been alive for decades since it was born! And no other images were
created from HIS image !

On the other hand, if Little Smalltalk has been alive since it was born then,
from real life experience, there would be a very slim chance "to invent the
future".

Oh, I almost forgot. My second request is to see some trait of Traits in Athena
;-)

To be honest, I don't really know about Traits but if Guy Steele put Traits in
his Fortress (with references to the Traitors as well ;-) then Traits is
something that deserves some good mouthing ;-)


Cheers,

PhiHo.

On Fri, Nov 09, 2007 at 05:45:56PM +0100, Friedrich Dominicus wrote:
> For whom it might be of interest:
> http://bugs.squeak.org/view.php?id=6739
>
> I have accessed the subversion tree of squeak but found out that this
> patch seem not to be intergated. I do not know of any other special
> sources for 64-bit Linux systems so I can't tell whether this is
> solved there or not

Hi Friedrich,

If you are referring to the patch that is attached to Mantis #5688,
then it will not be part of the subversion sources. It is a change
set to be applied to VMMaker. I don't know when the patch will be
added to VMMaker, but in the mean time just load the patch into
your image after loading VMMaker, and all should be well.

The patch is here:
http://bugs.squeak.org/file_download.php?file_id=2981&type=bug

Someone has already linked #6739 and #5688 on Mantis, so I won't
add any further comments there.

HTH,
Dave

And one nice morning, suddenly Athena.Net can support multi-core CPU without
flexing a finger. ;-)

Cheers,

PhiHo.

Let's make it not just multi-core CPU, but multi-core 128bit-CPU ;-)

> Cheers,
>
> PhiHo.
>

Talking of Timothy Budd, I really like this:

</QUOTE>

Whenever I teach a course there always comes
a moment as I'm reading the textbook
where I say ``you know, I think
I could write a better textbook than this''.
The unfortunate thing is that this is true
even when I am using my own book!

</QUOTE>

To paraphrase this in context of (Little) Smalltalk:

<PARAPHRASED for Communication purpose>

Whenever I teach a course there always comes
a moment as I'm using a programming language
where I say ``you know, I think
I could write a better programming language than this''.
The unfortunate thing is that this is true
even when I am using my own programming language!

</PARAPHRASED for Communication purpose>

;-) ;-) ;-)

On Nov 10, 2007 12:13 AM, Andreas Raab <[hidden email]> wrote:
> Yes, I understand the same. And I wish it were different...
>
Why - because Microsoft is The Bad Guy and Sun is The Good Guy?

Personally, I'm happy to see this competition, especially in the light
of .NET actually having *two* viable full implementations, one of them
open source.

We have two power houses in a battle for who can produce the best
generic virtual machine. Dynamic languages are on the rise, I bet that
dynamic language support will be way way better in 1-2 years in the
JVM simply because the Jay guys can't sit still while the Dot guys are
extending the functionality of their platform. The major IDE out there
(I think), Eclipse, starts to look and feel more and more like a
Smalltalk environment, and in some respects even better (with all the
static type info, the RB on Eclipse rocks!).

I think we, the Dynamic OO Language Bunch, are doing quite well, actually ;)

But getting back on topic: I have seen a handful of Smalltalk-on-JVM
implementations, and they all sucked performance-wise. I wouldn't
spend a lot of time doing yet another attempt until JVM support got
better, and until that time if it is for phones, you're screwed, learn
Java, if it is for research, .NET might be the thing.

--
"Human beings make life so interesting. Do you know, that in a
universe so full of wonders, they have managed to invent boredom. " -
Death, in "The Hogfather"

me too because I have the same fear as you :)
And we should push dan to release Squeak on Java.

Cees de Groot wrote:
> On Nov 10, 2007 12:13 AM, Andreas Raab <[hidden email]> wrote:
>> Yes, I understand the same. And I wish it were different...
>>
> Why - because Microsoft is The Bad Guy and Sun is The Good Guy?

No, it's because I need to run on a variety of platforms and I need
cross-platform libraries to work with. The last time I checked both
wasn't the case for .NET and both was the case for Java. Other than that
I don't care very much although MS' stance on Mono and patents does
nothing to convince me that the patent suits won't hit the fan at some
point.

Cheers,
   - Andreas

2007/11/11, Andreas Raab <[hidden email]>:
> Cees de Groot wrote:
> > On Nov 10, 2007 12:13 AM, Andreas Raab <[hidden email]> wrote:
> >> Yes, I understand the same. And I wish it were different...
> >>
> > Why - because Microsoft is The Bad Guy and Sun is The Good Guy?
>
> No, it's because I need to run on a variety of platforms and I need
> cross-platform libraries to work with.

What's wrong with Squeak?

Cheers
Philippe

Philippe Marschall wrote:
For the situations I'm thinking about, mostly lack of basic acceptance.
We tried to get NASA to use Croquet for some of their next-gen stuff but
it was a complete no-go. They used Java for all the reasons mentioned
but if there would've been a way of integrating efficiently we might
have had a chance (they'll likely be using Jython or Scala instead now).

Cheers,
   - Andreas

On 11-Nov-07, at 1:19 PM, Andreas Raab wrote:
>
> For the situations I'm thinking about, mostly lack of basic  
> acceptance. We tried to get NASA to use Croquet for some of their  
> next-gen stuff but it was a complete no-go. They used Java for all  
> the reasons mentioned but if there would've been a way of  
> integrating efficiently we might have had a chance (they'll likely  
> be using Jython or Scala instead now).


Do I recall correctly that java allows native methods? So make the  
entire squeak vm a native method :-)

tim
--
tim Rowledge; [hidden email]; http://www.rowledge.org/tim
Strange OpCodes: NBR: Unconditional No BRanch