dolfin team mailing list archive

["A Navaei" <axnavaei@xxxxxxxxxxxxxx>]

Johan,

Thanks for the detailed explanation, I ended up with this:

def numpyArrayToDolfinMatrix(a):
   nx = a.shape[0]
   ny = a.shape[1]
   m = dolfin.Matrix(nx, ny)
   m.set(a, numpy.arange(0, nx, dtype=numpy.uint0), numpy.arange(0,
ny, dtype=numpy.uint0))
   m.apply();
   return m

Can you confirm that from the c++ side, just a buffer pointer is
passed and there is no loop involved?

I'm actually trying to convert a c++ image data structure to dolfin
matrix, why shouldn't dolfin have a design to dynamically accept this
sort of input? Here is how the c++ image data is converted from/to
numpy in our project:

http://code.google.com/p/wrapitk/source/browse/trunk/ExternalProjects/PyBuffer/itkPyBuffer.txx

I'm trying to solve some variational image processing problems where
the data term in the PDE comes from the image buffer. Just to make
sure, is dolfin matrix the right way of inputing the image data or am
I missing something?


-Ali

2009/1/14 Johan Hake <hake@xxxxxxxxx>:
> On Wednesday 14 January 2009 14:07:27 A Navaei wrote:
>> Just to add, numpy.array is of type of numpy.ndarray, I don't
>> understand why this shouldn't work:
>>
>> a = numpy.array([3., 4., 5.])
>> m = dolfin.Matrix()
>> m.set(a, 1, 3)
>
> First the error message corresponds to the type numpy.ndarray, even if it says
> Numpy array.
>
> The PyDOLFIN matrix interface is not intended to be a dynamical highlevel
> interface a là MATLAB or Numpy. We have provided some functionality for basic
> linear algebra, add, sub, mult with scalar and vector, and its corresponding
> ifoo operators. You can multiply a PyDOLFIN.Matrix with a Numpy array,
> triggering an usuall matrix vector multiplication. This operation returns a
> Numpy array. You can also convert a matrix or vector to a numpy equivalent by
> foo.array().
>
> You can use v.set(a) where v is a PyDOLFIN.Vector and a is a corresponding
> numpy array. For now you have to take care of the dimension being suitable
> for the PyDOLFIN.Vector, otherwise you can trigger a segfault. (Not your
> fault but rather ours...)
>
> With PyDOLFIN.Matrix it is a bit more tricky. As these structurs are intended
> to be the result of an assemble(form). Think of these objects as quite static
> ones, only mutable by low level library operations.
>
> All this said, you can do what you want above by:
>
>  >>> a = numpy.array([3., 4., 5.])
>  >>> m = dolfin.Matrix(1,3)
>  >>> m.set(a, numpy.array([0],numpy.uint0), numpy.array([0,1,2],numpy.uint0))
>  >>> m.apply()
>  >>> m.disp()
>  row 0: (0, 3)  (1, 4)  (2, 5)
>
> Note the strange requirement of numpy.uint0.
>
> Good luck!
>
> Johan
>
>
>> Traceback (most recent call last):
>>  File "<console>", line 1, in <module>
>>  File "/usr/lib/python2.5/site-packages/dolfin/cpp.py", line 4434, in set
>>    return _cpp.Matrix_set(*args)
>> TypeError: Numpy array expected
>>
>>
>> 2009/1/14 A Navaei <axnavaei@xxxxxxxxx>:
>> - Hide quoted text -
>>
>> > Hi,
>> >
>> > Is it possible to convert numpy ndarray from/to dolfin matrix? It
>> > seems dolfin matrix only accepts numpy array while ndarray is a better
>> > choice.
>> >
>> >
>> > -Ali
>>
>> _______________________________________________
>> DOLFIN-dev mailing list
>> DOLFIN-dev@xxxxxxxxxx
>> http://www.fenics.org/mailman/listinfo/dolfin-dev
>
>
>