MATLAB: Accessing a cell array of cell arrays in a C++ mex file

cc++ pointercell arraysMATLABmexmxarray

I'm writing a large code that requires me to use a cell array of cell arrays of single precision real matrices. This may seem odd, but there are good reasons centred on efficient storage, that I won't go into here. The simplest code that illustrates the problem that I have is:
#include "mex.h"
float fn(mxArray *strat)
{
    float *indstrat, x;
    indstrat = (float *)mxGetCell(mxGetCell(strat,0),0);
    x = *(indstrat+1);
    return x;
}
void mexFunction(int nlhs, mxArray *plhs[], int nrhs, mxArray *prhs[])
{
    (void) nlhs; (void) plhs;
    fn(prhs[0]);
}
There's one input, a cell array of cell arrays of the form, for example, strat{i}{j}(k). In this fragment, I've used mxGetCell twice to extract the pointer to the {1}{1} vector, and then extracted the second value from that using '*(indstrat+1)'. However, debugging this in Visual Studio shows that the values aren't being extracted properly. I must be doing something wrong with the pointer to the array of pointers implicit in a cell array like this, but I can't see what. Should there be a '**' somewhere for a pointer to a pointer?
Can anyone tell me what I'm doing wrong?

Best Answer

John - check the function signature for mxgetcell - it returns a pointer to an mxArray object. So you would have to do something like the following
 float fn(const mxArray *strat)
 {
    mxArray *tempMxArray = 0;
    float   *indstrat, x;
    tempMxArray = mxGetCell(mxGetCell(strat,0),0);
    indstrat    = (float *)mxGetPr(tempMxArray);
    x = *(indstrat+1);
    return x;
 }
It is very similar to what you had, we just assign the result of the double mxGetCell call to the tempMxArray pointer, and then use mxGetPr to get a pointer to the first element in your single data type matrix.
I tried the above with the following example where the mexFunction is defined as
 void mexFunction(int nlhs, mxArray *plhs[], int nrhs, const mxArray *prhs[])
 {
    float x;
    x=fn(prhs[0]);
    mexPrintf("%f\n", x);
 }
And in MATLAB, the following is executed from the Command Line
 cellArray    = {};
 cellArray{1} = {single(1:4)};
 cellTest(cellArray);    % where cellTest is the name of the c-file with above code
 2.000000                % 2 is the output of the mex function
Given that the input matrix is [1 2 3 4], the output of 2 makes sense. (I needed to add the const qualifier on the fourth input to the mexFunction.)
Related Question