Jython

Hi,

Who is responsible for updating the Maven2 repository at ibiblio?  With
the talk of a beta1 nearing, I'd really like to see something more
up-to-date get published.  This would help get a newer version in
people's hands.  Moreover, I'm working on a jython plugin for Maven2 and
I'd like to be targeting the more modern version.

Thanks,
Kevin

--=20
Kevin Menard
Servprise International
WebReboot -- Remote Reboot Without Pulling the Plug
800.832.3823

Samuele Pedroni wrote:
> Samuele Pedroni wrote:
>   
>> Charlie Groves wrote:
>>   
>>     
>>> A few of the failures don't make sense to me yet, especially in the
>>> rich comparison operators.  The tests seem to expect the comparison
>>> method on the subclass to be called no matter what when compared to
>>> the base class.  i.e. from rich_comparisions
>>>
>>> class ZZ(complex):
>>>     def __eq__(self, other):
>>>        try:
>>>             return abs(self - other) <= 1e-6
>>>        except:
>>>             return NotImplemented
>>>
>>> zz = ZZ(1.0000003)
>>> assert zz == 1+0j
>>> assert 1+0j == zz
>>>
>>> fails on the second assertion.  Somehow __eq__ should always be called
>>> on the subclass but I haven't seen anything about that in what I've
>>> read about new style classes.  What am I missing?
>>>   
>>>     
>>>       
>> that we all forgot to change PyObject to implement the new style rules 
>> for picking __xxx__
>> vs. __rxxx__, I think I told Brian that this was still missing but this 
>> bit of information has got lost.
>>
>> We are still simply trying __xxx__ and then __rxxx__ in this order. Now 
>> if the right hand side
>> operand is an instance of a subclass  __rxxx__ should be tried first, 
>> for __eq__ it means
>> trying __eq__(other, self) before __eq__(self, other).
>>
>> The real rule is even more complex and subtle (although  this version of 
>> the rule was introduced only in 2.3
>> if I recall correctly), and depends on where the methods are found in 
>> the hierarchy exactly.
>>
>> This needs to be done for all binary methods.
>>   
>>     
> Also this is not going to be particularly pleasant to implement in a 
> sane way, I may try to think about it
> a bit during the weekend.
>
>   
for now I fear we are going to need to do something like PyPy, based on 
a lookup that also reports
where the method was found, plus probably some shortcuts for when the 
types are equal or are not
user defined types.

Long term it would be good to make the core dispatching for binary 
operators really based on double dispatching,
and put complicated logic only where is needed, on the other hand the 
maintainability of such a solution
is not too clear and even generating code would involve touching many 
classes which is not completely nice.

regards.

Frank Wierzbicki wrote:
> On 11/19/06, Charlie Groves <charlie.groves@...> wrote:
>> Hi all,
>>
>> I committed an implementation of slots earlier today, and with that
>> all of the language features for 2.2 are in.

So, where do we stand on a beta? If it's possible to get something in 
the next week, I'd love to tell the folks at JavaPolis (Dec 11-15) about 
a new Jython beta release.

-- 
Charles Oliver Nutter, JRuby Core Developer
Blogging on Ruby and Java @ headius.blogspot.com
Help spec out Ruby today! @ http://www.headius.com/rubyspec
headius@... -- charles.nutter@...

I just uploaded my first Jython patch - itertools.

The test coverage for the module is quite extensive, so I feel pretty
confident that it works "as designed". One test fails, but this is a test of
an implementation detail i CPython (reusing tuples). As I mentioned this is
my first module port, so some things may not be "optimal" when it comes to
the usage of internal Jython APIs. Feedback is very welcome. Oh, and I've
actually never used svn before, so my file properties may be a bit off...
Feel free to point me in the right direction if you find anything weird.

'Henrik

BTW There are couple of issues:
I tried to run tests that are in jython repository (bugtests) and
there are several failures.
Is there a list of showstoppers for beta version?

I suspect that some test failures that I encounter are due to some
misconfiguration.
So I'd like to see more detailed testing instructions. In particular:
1. What versions of JRE shoud be used for building Jython and running
tests (should it be samne version or versions can differ)?
2. Should tests be run by Jython? If so/ Should Jython that runs tests
be the same that is being tested?

2. Ant build target that is used for distribution (full-build) every
time cleans checked out sources and pulls them afresh (if I'm at home
it costs me about $2 for internet traffic :) so i'd like to have more
fine-grained control over build. That is separate targets, something
like "clean", "refresh-sources"  and "build-installer". Is that
feasible?

-- 
Petr Gladkikh

Hi Lars,

jython-dev@... is the best place for these sorts of
quieries, since there are actually several active developers lately,
and if you look at Jython-dev or jython-checkins@...
you will see that there has been a surge of activity over the last
couple of months.  The sys bug is fixed on the tip, and we are making
pretty good progress towards a beta release, though I can't say
exactly when it will happen.  We are also talking about reviving
snapshot releases, which should help.

-Frank

On 12/3/06, Lars Kellogg-Stedman <lars@...> wrote:
> Hi there,
>
> First, apologies if you see this email multiple times -- I wasn't able
> to find an authoritative source for your address.
>
> I've recently started working with Jython -- after spending some time
> looking at different Python web frameworks, I thought I'd try the
> Tomcat/Jython route.  Unfortunately, I've discovered that:
>
> (a) The latest production release (2.1) doesn't support the Iterator
> protocol, so too much recent code fails (mostly with NameError
> exceptions on StopIteration).
>
> (b) The latest development release (2.2a1) has a serious import bug
> that renders it essentially useless on anything with a /sys directory
> -- i.e., most modern Linux systems.  The symptoms are:
>
>   AttributeError: java package 'sys' has no attribute 'modules'
>
> Strace reveals that Jython is trying really hard to import the "/sys"
> directory as a module.  There's no __init__.py file here anyway, so
> I'm not sure what's going on.
>
> (c) The latest news post on the jython.org site is from early 2005.
>
> (d) All the recent email in the announce list on Sourceforge appears to be spam.
>
> All this makes me wonder if there is still life left in this otherwise
> tremendously interesting project.  Your blog shows some recent
> activity, so if I've simple been looking in the wrong place please let
> me know.  I'm really quite excited about the possibilities of a Jython
> + Tomcat setup.
>
> Thanks,
>
>   -- Lars
>
> --
> Lars Kellogg-Stedman <lars@...>
>