converting to/from C++ arrays to mxArrays

Hi everyone, I'm trying to write some code in c++ to speed up a computationally difficult portion of my matlab code. Could anyone lend some pointers on the best way to convert from mxArray/vector to C++ and back?

I've looked at the examples, but unfortunately they seem to keep everything in their own variable type.

here's what ive been trying (input is a structure with the first element the vector I'm interested in exporting)

void mexFunction( int nlhs, mxArray *plhs[], int nrhs, const mxArray *prhs[])
{
mxArray       *tmp;
double        *pdata=NULL;

tmp = mxGetFieldByNumber(prhs[0],0,0);
pdata = mxGetPr(tmp);

// This part tries to convert to the N_Vector format I need for the sundials solver I use in the C++ routine. UserData is just a structure / class. Needs the number of values, NP (constant) and a pointer to the first value (in type double (realtype))

UserData *data = new UserData;
data->param = N_VMake_Serial(NP,static_cast<realtype*>(pdata));

I'm not able to have the parameter set get passed to the rest of the routine. If i output pdata, pdata[0], or *pdata, I get pointers for all of them. Where are the values? If I could even convert them to a regular c++ array I would know how to process them from there.

Answer 1

[James Tursa]

pdata is a regular C++ pointer and can be used as a regular C++ array in most contexts. Try checking the API results to see that you are getting what you expect before using them. e.g.,

if( !mxIsStruct(prhs[0] ) {
    // error, not a struct, take action
}
if( mxIsEmpty(prhs[0]) ) {
    // error, empty, take action
}
tmp = mxGetFieldByNumber(prhs[0],0,0);
if( !tmp ) {
    // error, field is NULL, take action
}
if( !mxIsDouble(tmp) ) {
    // error, not a double, take action
}
if( mxIsEmpty(tmp) ) {
    // error, empty, take action
}
NP = mxGetNumberOfELements(tmp);
pdata = mxGetPr(tmp);

Q: Why are you doing a static cast on pdata? Can't you just pass pdata directly since it is a double * and N_VMake_Serial is expecting a double * by default? I would think that using a static cast here would be a mistake (i.e., masking an actual coding error if pdata type did not match realtype)

[Peter]

the static cast was leftover from a previous iteration where I was returning a void pointer. Good point though, I removed it.

Ive simplified the code to simply passing the vector as an argument but I still don't see how the mxArrays are working (maybe its more that I dont understand how c++ arrays work?)

anyways, here is what i have:

double *vin1;

*vin1 = 0;

mexPrintf("\n%d",*vin1);

vin1 = (double*) mxGetPr(prhs[0]);

mexPrintf("\n%d",*vin1);

This returns 0, then -1717986918. Why doesn't it return the first value in my vector? (0.4)

[James Tursa]

%d is a decimal int format. Use %f or %g instead.

[Peter]

:)

thanks so much for your help!! Never used the printf command before having to use the matlab API, always used cout <<.