[Numpy-discussion] numpy, swig and TNT-Arrays

Rolf Wester rolf.wester at ilt.fraunhofer.de
Tue Nov 11 05:28:21 EST 2008


Charles R Harris wrote:
> On Tue, Nov 11, 2008 at 2:19 AM, Rolf Wester
> <rolf.wester at ilt.fraunhofer.de>wrote:
> 
>> Charles R Harris wrote:
>>> On Tue, Nov 11, 2008 at 1:24 AM, Rolf Wester
>>> <rolf.wester at ilt.fraunhofer.de>wrote:
>>>
>>>> Hi all,
>>>>
>>>> I would like to wrap some C++ classes that use TNT-Arrays. Is it
>>>> possible to pass numpy arrays to C++ functions that expect TNT-Arrays as
>>>> function parameter? Does anybody know how the wrappers could be
>>>> generated using swig? I would be very appreciative for any help.
>>>>
>>>> With kind regards
>>>>
>>> IIRC, TNT does vectors and matrices, they have constructors, and they are
>>> contiguous. I think you can make wrappers, but it isn't going to be
>> anything
>>> straight forward unless you can reuse the memory from a numpy array and I
>>> don't recall that that sort of constructor is available.
>>>
>>> Is TNT still active? It looked pretty dead last time I looked several
>> years
>>> ago.
>>>
>>> Chuck
>>>
>>>
>>>
>>> ------------------------------------------------------------------------
>>>
>>> _______________________________________________
>>> Numpy-discussion mailing list
>>> Numpy-discussion at scipy.org
>>> http://projects.scipy.org/mailman/listinfo/numpy-discussion
>> TNT has constructors like:
>>
>> TNT::Array1D<double>(int n, double * data)
>>
>> which do not allocate a new C-array but that use "data" as their
>> data-array.
>>
> 
> I don't think there is any easy way to do what you want without writing some
> code somewhere along the line. You can expose the C++ functions and TNT to
> python, but to use numpy arrays you will need some way to get the data back
> and forth between TNT arrays and numpy arrays. I suspect you will end up
> just copying data into TNT arrays, calling your function,  and then copying
> data back out of the result. Cython might be an alternative to swig for
> that.
> 
> It would help to have a better idea of what you want to do. Do you just want
> to wrap an existing bunch of functions that use TNT?
> 
> Chuck
> 
> 
> 
> ------------------------------------------------------------------------
> 
> _______________________________________________
> Numpy-discussion mailing list
> Numpy-discussion at scipy.org
> http://projects.scipy.org/mailman/listinfo/numpy-discussion

It's my own code so I have the choice how to do it. Until now I used the
typemaps defined in numpy.i, so I had either to use the 1-dimensional
arrays even in case of multidimensional data or to copy the data. I
wondered wether there is a more elegant way of using numpy arrays on the
python side and TNT::Arrays on the C++ side without having to
explicitely write extra code.


Rolf

-- 
------------------------------------
# Dr. Rolf Wester
# Fraunhofer Institut f. Lasertechnik
# Steinbachstrasse 15, D-52074 Aachen, Germany.
# Tel: + 49 (0) 241 8906 401, Fax: +49 (0) 241 8906 121
# EMail: rolf.wester at ilt.fraunhofer.de
# WWW:   http://www.ilt.fraunhofer.de



More information about the NumPy-Discussion mailing list