Economics & Eye Candy

Tue Jun 7 19:51:03 EDT 2005

Dear Dan

Thanks for the reply. Alas, I am one of those
colleagues you describe, since the major part of what
I want to do simply doesn't work well with rev.

Firstly, rev is far too slow for numerical
computation. I did some tests and even byte-optimized
Python beats rev for speed in most cases and that's
hardly a "Gold Standard" for performance. Java, long
lambasted for being slow, absolutely leaves rev in the
dust as far as performance goes. Try doing a matrix
inversion or sorting a huge list in both languages and
you'll see what I mean. Granted, the Java development
path is somehwat less fun than rev's, but Java is
stable mature and truly cross (cross cross cross)
platform if you use Swing components (Dan :-D)

Now imagine you want to render some complex objects in
3D or visualize 2D or 3D data, both of which are
things one might like to do even if you're not a geeky
science coder like me (e.g. for games, financial
analysis, GIS etc. etc.) All of this kind of stuff
would be excruciating and impractical to code in rev
unless ... (and here's my second major gripe with rev)
... you're willing to postpone your project for 6
months and write wrappers for a visualization DLL in C
... which is just what you wanted to avoid when you
started using Transcript in the first place! 

With Java or Python, I import OpenGL or whatever
visualization package I'm using and my workflow
continues in my language of choice. I have access to
the collective effort of thousands before me in a huge
shared community of code. Java and Python may not be
as much fun as rev, but they're like flying a plane
compared to knuckle-walking in C.

But as I said, I like Transcript and would like to see
the evolution of rev steer more towards making the
Transcript language really useful. So here's my
wishlist for the good folks in Edinburgh ...

1. More comprehensive collection of datatypes
(real arrays with ability to nest arrays etc.)
2. Much better performance 
(should also be enhanced by point 1)
3. Easier access to externals
4. A soild 2D vector graphics package

And finally, inspired by Richard Gaskin's suggestion
... Compilation to Java bytecodes and integration of
Java class libraries into rev ... Ooooh, you might
have something there Richard ...

Keep language you love, just swap interpreters (kind
of like Jython)

I like it :-D

Best

Gordon

--- Dan Shafer <revdan at danshafer.com> wrote:

> Gordon.....
> 
> While I sympathize with some of your feelings (while
> I continue to be  
> my most productive self when using Rev and grinning
> from ear to ear  
> watching my colleagues who use Python, Java, and
> Basic go through  
> excruciating pains to do what I can do in seconds or
> minutes in my  
> favorite tool!), this notion of integrating
> externals is one on which  
> I have pretty strong opinions. (Yes, I know. That's
> unusual for me. NOT)
> 
> The problem, IMNSHO, isn't with Rev, it's with the
> fact that as far  
> as I know there is no single method of creating and
> implementing  
> externals that runs on all platforms. I may be wrong
> about that; DLLs  
> may in fact be more cross-platform than they were
> last time I looked,  
> particularly with the emergence of OS X. But I am
> pretty strongly  
> opposed to Rev spending significant time and
> resources extending the  
> capabilities of the program for *any*
> platform-specific  
> functionality. And, FWIW, that includes when they do
> so for MY  
> favorite platform, OS X. I'd have been delighted to
> see 2.6 fix a  
> bunch more bugs and not implement CoreImage and
> DeepMask stuff. Now,  
> I realize that in saying that I'm in a small
> minority and I also  
> realize that the DeepMask stuff works cross-platform
> and will be a  
> real boon as Longhorn emerges from its near-decade
> in cold storage.  
> But my position remains the same: of all the
> development tools from  
> which I can choose, only two that I know of -- Rev
> and Flash -- give  
> me true cross-platform capability. And because I
> choose to work on a  
> minority platform but want to be able to reach the
> majority platform,  
> cross-platform is my personal number one feature in
> Rev. Anything  
> that detracts from that is just in my way or an
> unnecessary appendage.
> 
> So if there is a way to facilitate the incorporation
> of cross- 
> platform external routines relatively transparently
> and to give me a  
> Transcript-level way of dealing with them, I'm happy
> to see it  
> implemented even though I have yet to need such
> capability in the  
> dozens of things I've written. But if supporting
> externals must be  
> done for a specific platform -- or, almost as badly,
> differently for  
> each platform -- then I'm in favor of Rev passing on
> that and fixing  
> more of the bugs we still have to work around.
> 
> 
> On Jun 7, 2005, at 1:10 PM, Gordon Webster wrote:
> 
> > still no real
> > arrays, still no bridge to the vast world of
> shared
> > libraries that would allow me to integrate
> external
> > functionality into rev and save me having to
> either
> > reinvent the wheel or spend my time writing C
> wrappers
> >
> 
> 
> 
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> Dan Shafer, Co-Chair
> RevConWest '05
> June 17-18, 2005, Monterey, California
> http://www.altuit.com/webs/altuit/RevConWest
> 
> _______________________________________________
> use-revolution mailing list
> use-revolution at lists.runrev.com
>
http://lists.runrev.com/mailman/listinfo/use-revolution
> 