[SAGA-RG] SAGA python bindings...

[Ole Weidner]

All,

On Aug 16, 2011, at 5:58 AM, Andre Merzky wrote:

> Hi Sylvain,
> 
> On Tue, Aug 16, 2011 at 12:19 PM, Sylvain Reynaud
> <Sylvain.Reynaud at in2p3.fr> wrote:
>> 
>>> On 8/8/11 11:20 , "Sylvain Reynaud"<Sylvain.Reynaud at in2p3.fr>  wrote:
>>>> 
>>>> Yes, JPySAGA implements the same python binding as JySAGA (the python
>>>> wrapper on top JavaSAGA), but as far as I remember Julien also made a
>>>> few compatibility tests with the wrapper that implements this binding on
>>>> top of SAGA-C++.
>>>> We were not able to get everything working because this wrapper was
>>>> still under development, but we were at least able to implement the
>>>> simple use-cases that we shown in our demo at Brussels last year, in
>>>> order to show that SAGA-C++ and JSAGA could be used together in the same
>>>> application.
>>> 
>>> One of the issues in the JPySAGA is the usage of the "create" class
>>> methods. Do you know the history of that?
>> 
>> No, I don't know the history, but since JySAGA has been developed on top of
>> JavaSAGA, I can imagine that it was influenced by the "create" methods of
>> the Java binding.
> 
> I Cc'ed Steve Fisher, who kept better track of the python API evolution than
> most I think - he might be able to shed light on this.
> 
> 
>>>>> Either way, as we do have a significant community of python users for
>>>>> SAGA, we are still interested in driving a uniform python bindings
>>>>> forward.  Would you be interested in discussing that topic again,
>>>>> or is that at the moment out of scope for your group?
>>>> 
>>>> Yes, we are still very interested in discussing that topic.
>>>> I put in CC Julien (in holidays this week) because he has developed
>>>> JPySAGA and he knows Python far better than I do.
>>> 
>>> Good. As said, we are interested particularly in JSAGA because of gLite
>>> support.
>>> Our language choices are Java and Python, so I already proposed to assist
>>> in sorting out the python differences between the various implementations.
>> 
>> I know that Julien is also very concerned with having "pythonic" APIs, and I
>> think he would be probably interested in contributing to select the most
>> "pythonic" parts of each SAGA Python binding, in order to converge on a SAGA
>> binding as "pythonic" as possible.
>> 
>> Andre, do you think this would be the right way to synchronize the
>> implementations, or is it already too late to do such changes in the binding
>> (considering the existing SAGA-python users community) ?
> 
> Honestly, our group has mixed feelings.  Of course it would be nice if
> the python bindings were unified, but we are also somewhat scared of
> breaking code which is in heavy use already, since years.  It would have
> been much better to sync the python bindings way earlier - but well,
> that is just wishful thinking... ;-)

We can always create an alternative set of Python bindings for our SAGA implementation (i.e., forking the current python bindings & change the API). Changing the existing API is not an option due to strong application dependencies. 

Developing an alternative set of Python bindings would only require minimal effort on our side. Once we have reached that point, we can still think about a gentle, non-intrusive migration strategy.  

> 
> If we find reasonable technical procedures to mitigate the transition
> pain for our end users, we would certainly be willing to migrate to
> a common binding.  The biggest motivation for us would be if (a) our
> users could seamlessly experiment with other SAGA implementations, and
> (b) we could that way increase the acceptance of SAGA as a standardized
> and widely available solution, and thus increase adoption in general.
> 
> 
> My $0.02, I'd love to hear other people's opinion on that topic
> (Ole?  Shantenu?).
> 
> 
> BTW, in terms of group procedure: we already came to a consensus about
> what python bindings are to be standardized as OGF specification.

Andre, could you send around a link to that document, please? If all parties can agree with the binding specification, I'd say we should just go with it. If not, I think it would be worth it to have another iteration. 

My $0.02. 

Cheers,
Ole

>  From
> this thread, and some offline discussions, it seems that the opinions
> though vary on that topic.  We can certainly re-iterate the python bindings
> on that level, but I would hate to see us spending another year on it.
> 
> Since most interested parties are in Lyon, I'll try to book another set
> of sessions, so that we can come to a closure on the specification side,
> and can focus on the technical aspects, if that's ok with everybody...
> 
> Best, Andre.
> 
> 
>> 
>> 
>>>>> PS.: you are likely aware thet OGF-33 is being held n Lyon, in mid
>>>>> September.  Do you plan to attend, by any chance?
>>>> 
>>>> Yes, I am already registered. I really have no excuse for not doing the
>>>> "travel" since I can see the place of the conference through the window
>>>> of my desktop!  ;-)
>>> 
>>> Note that I will also be in Lyon, so I would be happy to meet up with you.
>> 
>> I will be happy too.
>> 
>> Regards,
>> Sylvain
>> 
>> 
>>> Regards,
>>> 
>>> Mark
>>> 
>>> 
>>> 
>>> 
>> 
>> 
> 
> 
> 
> -- 
> Nothing is ever easy...
> --
>  saga-rg mailing list
>  saga-rg at ogf.org
>  http://www.ogf.org/mailman/listinfo/saga-rg