JPE - Java Python Extension Project

Steve Waterbury steve.waterbury at gsfc.nasa.gov
Tue Nov 21 05:57:58 CET 2000


Frederic, 

I don't think it's a good idea to have discussions on the 
python-announce-list, so I'm sending this reply to 
comp.lang.python, in case you want to continue ...  

Frederic Giacometti wrote:
> 
> Steve Waterbury wrote:
> 
> > Interesting.  Do you see any advantage to this versus ILU?
> >
> > ftp://ftp.parc.xerox.com/pub/ilu/ilu.html
> 
> As I know, ILU allows communication between different process, and thus
> requires packing/copying/unpacking over all arguments using a neutral
> representation, that needs to be defined through an IDL, and so on...

IDL (or ISL) yes, but not *necessarily* packing/copying/unpacking ... 

> In JPE, everything works in the same process, in the same memory space.
> pointers on Java objects are merely wrapped in Python objects, and
> vice-versa.  A lot less overhead, especially when dealing with complex data sets.

That's a pretty fair description of the way ILU works, too -- 
although both may not be in the same process, they could be in 
the same address space ... you may not be familiar with ILU 2.0, 
which supports multiple languages in one address space -- see:

ftp://ftp.parc.xerox.com/pub/ilu/2.0b1/manual-html/manual_19.html#SEC568

The reason I prefer the ILU paradigm is that it is multi-lingual:  
if one wants to integrate 3 languages in a point-to-point language 
integration paradigm, *3* point-to-point integrations are needed 
(x-y, y-z, x-z), whereas in the ILU paradigm, you need only one, even 
for several languages.  Note that ILU supports Python, Java, C, C++, 
Lisp (CLOS) and others.  

Also, when the next groovy language comes along, only one set 
of interfaces needs to be defined for it, rather than a plethora 
of integration solutions again.  

Okay, maybe folks just love to write these things, but I think 
it would be really great to see contributions to the ILU project 
rather than yet-another-language-x-language-y integration scheme, 
of which now there are (at least) 2 for Python with Java.   I'd 
wager that if all the effort that went into JPython and JPE had 
gone into adding improvements to ILU, it would probably be better 
than either one and would give Python the ability to call objects 
and modules from several other languages as well!  

I can't believe that folks think Python and Java are all they'll 
ever need ...  (On my project, we use Lisp, Java, and Python. ;^)  

</soapbox>  

For a nice frames version of the ILU docs, see:

http://www-db.stanford.edu/~testbed/ilu/ilu20doc/

Cheers,
-- Steve.

                                           oo _\o
                                            \/\ \
                                              /
____________________________________________ oo _________________
"Sometimes you're the windshield; sometimes you're the bug."
- Knopfler

Stephen C. Waterbury                       Component Technologies
Code 562, NASA/GSFC                  and Radiation Effects Branch
Greenbelt, MD 20771           Engineering Web/Database Specialist
Tel: 301-286-7557                              FAX:  301-286-1695        
WWW:          http://misspiggy.gsfc.nasa.gov/people/waterbug.html
_________________________________________________________________



More information about the Python-list mailing list