Matlab API reading .mat file from c++, using STL container

Question

I have to read some .mat data files from c++, I read through the documentation, but I would like to know how to handle the data in a clean and elegant way, e.g. using std:vector(modest .mat file size(10M~1G), but memory issues should be taken seriously)

My function is sth like:

#include <stdio.h>
#include "mat.h"
#include <vector>

int matread(const char *file, const vector<double>& pdata_v) {

MATFile *pmat;

pmat=matOpen("data.mat","r");
if (pmat == NULL) {
    printf("Error opening file %s\n", file);
    return(1);
}
mxArray *pdata = matGetVariable(pmat, "LocalDouble");

// pdata -> pdata_v

mxDestroy pa1;  // clean up
return 0;
}

so, the question is, how can I make a copy from mxArray *pdata array to vector pdata_v efficiently and safely?

Answer 1

Here is an example of using the MAT-API:

test_mat.cpp

#include "mat.h"
#include <iostream>
#include <vector>

void matread(const char *file, std::vector<double>& v)
{
    // open MAT-file
    MATFile *pmat = matOpen(file, "r");
    if (pmat == NULL) return;

    // extract the specified variable
    mxArray *arr = matGetVariable(pmat, "LocalDouble");
    if (arr != NULL && mxIsDouble(arr) && !mxIsEmpty(arr)) {
        // copy data
        mwSize num = mxGetNumberOfElements(arr);
        double *pr = mxGetPr(arr);
        if (pr != NULL) {
            v.reserve(num); //is faster than resize :-)
            v.assign(pr, pr+num);
        }
    }

    // cleanup
    mxDestroyArray(arr);
    matClose(pmat);
}

int main()
{
    std::vector<double> v;
    matread("data.mat", v);
    for (size_t i=0; i<v.size(); ++i)
        std::cout << v[i] << std::endl;
    return 0;
}

First we build the standalone program, and create some test data as a MAT-file:

>> mex -client engine -largeArrayDims test_mat.cpp

>> LocalDouble = magic(4)
LocalDouble =
    16     2     3    13
     5    11    10     8
     9     7     6    12
     4    14    15     1

>> save data.mat LocalDouble

Now we run the program:

C:\> test_mat.exe
16 
5 
9 
4 
2 
11 
7 
14 
3 
10 
6 
15 
13 
8 
12 
1 

Answer 2

Here's another idea. If you're allergic to bare pointers in C++ code (nothing wrong with them, by the way), you could wrap the bare pointer in a boost or C++11 smart pointer with a deleter that calls the correct mxDestroyArray() when the pointer goes out of scope. That way you don't need a copy, nor does your user code need to know how to correctly deallocate.

typedef shared_ptr<mxArray> mxSmartPtr;

mxSmartPtr readMATarray(MATFile *pmat, const char *varname)
{
    mxSmartPtr pdata(matGetVariable(pmat, varname),
                     mxDestroyArray);  // set deleter
    return pdata;
}

int some_function() {
    mxSmartPtr pdata = readMATarray(pmat, "LocalDouble");
    ...
    // pdata goes out of scope, and mxDestroy automatically called
}

Idea taken from here: http://www.boost.org/doc/libs/1_56_0/libs/smart_ptr/sp_techniques.html#incomplete

Answer 3

You can first get the data pointer of the mxArray *pdata and then copy data to vector<double> pdata_v:

double *ptr = (double *) mxGetData(pdata);
pdata_v.resize(numOfData);
memcpy(&pdata_v[0], ptr, numOfData*sizeof(double));

---

ps1: Pay extra attention to that, in MATLAB, matrice are in col-major order. So if pdata stores [1 2 3; 4 5 6], pdata_v will be 1 4 2 5 3 6.

ps2: Change const vector<double>& pdata_v to vector<double>& pdata_v if you want to change its content.