[Pyrex] pyrex for parrots?

[Bob Ippolito]

On May 24, 2005, at 7:48 PM, Michal Wallace wrote:

> On Tue, 24 May 2005, Bob Ippolito wrote:
>
>
>> On May 24, 2005, at 6:15 PM, Greg Ewing wrote:
>>
>>
>>> Michal Wallace wrote:
>>>
>
>
>>>> It seems like pyrex could benefit from the same kind of
>>>> refactoring... Then you could have separate backends to create
>>>> modules for for cpython, jython (using JNI), pirate, IronPython,
>>>> and whatever else crops up. Have you thought about doing something
>>>> like that?
>>>>
>
>
>>> I hadn't. It's an interesting idea, although changing
>>> the target language is likely to have implications for
>>> the Pyrex language itself, not just the compiler --
>>> currently it's heavily oriented towards C.
>>>
>
>
>> Sounds more like the kind of effort that should be put into PyPy,
>> rather than Pyrex, since it more or less boils out of what they're
>> already doing.
>>
>
> I think pirate is a good candidate for
> a pypy backend...
>
>
> But I was actually talking about creating extension modules that  
> interface with C. You'd still need the "cdef" and other type  
> information that pyrex supports.
>
> The backends would still be outputting C
> code, but styled to work with that particular
> virtual machine.
>
> In jython's case you might have to output both C *and* java  
> wrappers. I'm not sure.
> But maybe this is just duplicating swig...

Well, obviously, at some point PyPy is going to need a FFI for C (and/ 
or Java, IL, etc.)... to do things like allocate memory and do I/O :)

-bob