Re: How to access underlying C datatype of octave [u]int{8|16|32|64}NDArray?

[John Swensen]

That works fine (e.g. that is how I am doing it), but remember that the 
fortran_vec is column major, so you might want to do
sigout.tranpose()
if your are doing a memcpy() from a C array to the fortran_vec.

Here is a snippet of my code.
<<BEGIN>>
uint8NDArray m( dim_vector(settings.width, settings.height));
octave_uint8* tmp = m.fortran_vec();
memcpy( tmp, capturebuffer, settings.height*settings.width );
return octave_value( m.transpose() );
<<END>>
where 'capturebuffer' is a uint8_t array and 'settings' is a struct with 
camera parameters in it.

Hope this helps.

John Swensen

Benjamin Lindner wrote:
> Hello list,
>
> What is the preferred (or best) way to get access to the underlying
> C Datatypes of the octave integer Arrays?
> I am using an int32NDArray and would like to retrieve a pointer to
> int32_t which is the internal datatype. The member functions data() and
> fortran_vec() return pointers to the octave_int32 type, which does not
> behave like a C integer (which is in my case necessary).
>
> One solution that (at least syntactically) works is to use
> reinterpret_cast<>, but it there a more elegant solution?
>
> Imagine C++ Code like:
>
> #include <octave/oct.h>
> #include <octave/oct-intytpes.h>
>
> template <class T> int foo( const T* pIN, size_t len_in, T* pOUT ) {
>    // do something in here...
>    // T must be a built-in C integer type
> }
>
> DEFUN_DLD( foo, args, , "out=foo(in)" ) {
>    int32NDArray sigin( args(0).array_value() );
>    unsigned int len_sigin = args(0).length();
>    
>    int32NDArray sigout( dim_vector(len_sigin, 1), 0 );
>    
>    // IS THIS THE PREFERRED WAY ??
>    const int32_t* pIN = reinterpret_cast<const int32_t*>( sigin.data() );
>    int32_t* pOUT = reinterpret_cast<int32_t*>( sigout.fortran_vec() );
>    
>    // call template 
>    unsigned int t = foo( pIN, len_sigin, pOUT );
>    
>    ...
> }
>
> thanks for hints
> benjamin
>