[SciPy-user] numpy array in ctype struct

[Rob Hetland]

On Jan 15, 2008, at 7:04 PM, Paul Kienzle wrote:
>
> Instead of creating the ctypes struct I just used numpy to create a  
> scalar
> of the correct structure.  I wrapped it in a class so that I could  
> reference
> the fields directly as e.g., instance.A.  I used a factory function
> for generating the class from the structure definition since I will  
> need
> to wrap several structures.
>
> I'm attaching cstruct.py and embedarray.c where I demonstrate this.
>
> I'm particularly pleased that I can assign a 4x4 array to instance.A
> and it just works!
>
> The ctypes docs talk about possible alignment issues for the  
> structs on
> some architectures.  They say it the ctypes follows the conventions of
> the compiler which created it.  I haven't checked if this will be a
> problem with the numpy solution you outline above.

> <embedarray.c><cstruct.py>

How do you pass ndarrays of arbitrary size.  I would like to have  
some C code like:

extern void bar(double **data, int L, int M) {
     ...
}

The code needs to pass the array on to another library as a **double,  
and I cannot seem to get ctypes to pass a 2 dimensional array to the  
subroutine.  Arrays if fixed size seem to work fine, and single  
dimensional *doubles also work fine, even when passing  
multidimensional numpy.ndarrays.

But, how to pass an arbitrary (two-dimensional) numpy.ndarray as a  
**double?

-Rob


----
Rob Hetland, Associate Professor
Dept. of Oceanography, Texas A&M University
http://pong.tamu.edu/~rob
phone: 979-458-0096, fax: 979-845-6331