Thread Subject: How to use the double array for indexing in MEX file

For you MEX-experts this must be easy: How do I use an
array, which in Matlab is of double precision, for indexing
into another array inside a Mex function? And what if the
array was created as 'int64', how do I use it then? There
does not seem to be any equivalent to mxGetPr for this
datatype...

If you have any source of information that you can recommend
of these aspects, that would be much appreciated.

Petr

"Petr Krysl" <pkryslNOSP@Mucsd.edu> writes:

> For you MEX-experts this must be easy: How do I use an
> array, which in Matlab is of double precision, for indexing
> into another array inside a Mex function? And what if the
> array was created as 'int64', how do I use it then? There
> does not seem to be any equivalent to mxGetPr for this
> datatype...
>
> If you have any source of information that you can recommend
> of these aspects, that would be much appreciated.

If the array really is double, then extract each element in the usual
way, then cast it to int:

double *index_array_double;
int index;

index_array_double = mxGetPr(prhs[0]);

for(i=0; i<N; i++) {
  index = (int) index_array_double[i];
  // use index
}


If the array is some other datatype, you can cast the pointer output
from mxGetData:

long long int *index_array_int64;

index_array_int64 = (long lont int*) mxGetData(prhs[0]);

index_array_int64[0] is the first element


mxGetData() does the same thing as mxGetPr(), it just returns a void *
type instead of a double *. Both are pointers to the data buffer of the
variable. You just have to tell C how to interpret the pointer.

-Peter

Peter Boettcher <boettcher@ll.mit.edu> wrote in message
<muyy79o20pw.fsf@G99-Boettcher.llan.ll.mit.edu>...
> "Petr Krysl" <pkryslNOSP@Mucsd.edu> writes:
>

> If the array really is double, then extract each element
in the usual
> way, then cast it to int:

Thanks, much obliged.

Petr

> long long int *index_array_int64;
>
> index_array_int64 = (long lont int*) mxGetData(prhs[0]);
>
> index_array_int64[0] is the first element
>
>
> mxGetData() does the same thing as mxGetPr(), it just
returns a void *
> type instead of a double *. Both are pointers to the data
buffer of the
> variable. You just have to tell C how to interpret the
pointer.
>
> -Peter

I have tried this method (casting) to send an uint32 data
from Matlab to MEX but then when I want to access my data in
MEX it's just garbage. If I use mxGetPr and convert the
input uint32 to double in Matlab then everything works.

Any other steps I need to take before or after doing the
explicit casting in MEX files.

"Hamid " <spam.wax@gmail.com> wrote in message <g2k5s2
$4on$1@fred.mathworks.com>...
>
> > long long int *index_array_int64;
> >
> > index_array_int64 = (long lont int*) mxGetData(prhs
[0]);
> >
> > index_array_int64[0] is the first element
> >
> >
> > mxGetData() does the same thing as mxGetPr(), it just
> returns a void *
> > type instead of a double *. Both are pointers to the
data
> buffer of the
> > variable. You just have to tell C how to interpret the
> pointer.
> >
> > -Peter
>
> I have tried this method (casting) to send an uint32 data
> from Matlab to MEX but then when I want to access my
data in
> MEX it's just garbage. If I use mxGetPr and convert the
> input uint32 to double in Matlab then everything works.
>
> Any other steps I need to take before or after doing the
> explicit casting in MEX files.
>

Can you post some code to show us what did and didn't work
for you? I agree with Peter, you should only use mxGetPr
for inputs that are known to be double, and you should use
mxGetData for all the other classes. Using mxGetData with
a uint32 class should work just fine as long as you do
your casting correctly.

James Tursa