Fixing "Undefined Reference" Error in .o File

[NeoDevin]

I'm trying to call a function from a .o file, and it keeps telling me:

"undefined reference to ..." where the ... is whatever function I'm trying to call.

Do any of you have any idea what's wrong with this?

 

[jim mcnamara]

How did you compile your code - what command did you use?
A .o file has to be linked into your code (C example)

cc myfile.c anotherfile.o -o myfile

 

[NeoDevin]

I figured it out, I had wrong capitalization of the function name.

Next question:

If I have a pointer to the first element of a dynamically allocated 2 dimension array, is there any way to use two subscripts with that pointer.

//I create a dynamically allocated array:
int i;
int N = 100;
double **A;
double *p;
A = new double *[N]
for(i = 0; i < N; i++) A[i] = new double [N];

// and point a pointer to the first element.
p = A[0];

//now I want to be able to use it like this:

p[3][4] = 12;

 

[chroot]

C really has no concept of multi-dimensional arrays; all arrays are really stored in a one-dimensional format in memory. When you want to access element [3][4], the compiler actually performs some math to get a one-dimensional index: 3 * (number of elemts in row) + 4. If the compiler doesn't know how many elements are in a row, it cannot do this.

So, if all you're given is a pointer to the first element of a multidimensional array, you are effectively screwed.

- Warren

 

[NeoDevin]

There's no way to typecast it, if you know the number of elements in the row?

 

[rcgldr]

Your code creates an array of pointers, this is not a matrix.

In the matrix examples below, note that "double (*pmd)[5]" declares pmd to be a pointer to an array of 5 doubles,
as opposed to "double *pmd[5]" which is an array of 5 pointers to double; the parenthesis around *pmd is what makes the difference.

C++ matrix example:

    double (*pmd)[5] = new double[4][5];
    pmd[1][2] = 1.0;
    delete [] pmd;

C++ array with matrix typecast example:

    double *pad = new double[20];
    ((double (*)[5])pad)[1][2] = (double) 1.0;
    delete []pad;

C matrix:

    double (*pmd)[5] = (void *)malloc(4*5*sizeof(double));
    pmd[1][2] = 1.0;
    free(pmd);

C array with matrix typecast:

    double *pad = (void *)malloc(4*5*sizeof(double));
    ((double (*)[5])pad)[1][2] = 1.0;
    free((void *)pad);

Using Microsoft compiler, all 3 examples produce essentially the same code:
allocate 160 bytes,
set [ptr+56] = [ptr + (((1 x 5) + 2) x sizeof(double))] to 1.0
free ptr.

 

[NeoDevin]

    double (*pmd)[5] = new double[4][5];
    pmd[1][2] = 1.0;
    delete [] pmd;

Can you do this if you need both dymensions to be dynamically allocated?

C++ array with matrix typecast example:

    double *pad = new double[20];
    ((double (*)[5])pad)[1][2] = (double) 1.0;
    delete []pad;

This is what I was looking for. Thank you.

 

[rcgldr]

Can you do this if you need both dimensions to be dynamically allocated?

Not with the Microsoft compilers. Only the first dimension of a muti-dimensional array is allowed to be dynamic, the rest have to be compile time constants. You would need to use a vector of pointers to accomplish the equivalent, and would have to iterate for each allocation and/or freeing of a vector. The entity would be a mix of pointers and doubles.

In the examples above, what I really created was a pointer to an vector of 5 doubles, such as "double (*pv)[5]". C and C++ allow pv[j] or (*pv+i)[j], effectively promoting the pointer to a vector to a pointer to a matrix.

For a more familiar example, "double *ps" is a pointer to a scalar double (just one instance), and C and C++ allow ps or *(ps+i), effectively promoting the pointer to a scalar to a pointer to a vector.

Here is an example that actually uses a pointer to a matrix:

    static double *ps = new double[40];         // poitner to scalar
    double (*pv)[5] = (double (*)[5])ps;        // pointer to vector
    double (*pm)[4][5] = (double (*)[4][5])ps;  // pointer to matrix
    ps[6] = 1.0;                                // *(ps +  6) = 1.0
    pv[1][2] = 1.0;                             // *(ps +  7) = 1.0
    (*pm)[3][4] = 1.0;                          // *(ps + 19) = 1.0
    pm[1][2][3] = 1.0;                          // *(ps + 33) = 1.0
    delete [] ps;

 

[NeoDevin]

Not with the Microsoft compilers. Only the first dimension of a muti-dimensional array is allowed to be dynamic, the rest have to be compile time constants.

In the examples above, what I really created was a pointer to an vector of 5 doubles, such as "double (*pv)[5]". C and C++ allow pv[j] or *(pv+i)[j], effectively promoting the pointer to a vector to a pointer to a matrix.

As an analogy, "double *ps" is a pointer to a scalar double (just one instance), and C and C++ allow ps or *(ps+i), effectively promoting the pointer to a scalar to a pointer to a vector.

Here is an example that actually uses a pointer to a matrix:

    static double *ps = new double[40];         // poitner to scalar
    double (*pv)[5] = (double (*)[5])ps;        // pointer to vector
    double (*pm)[4][5] = (double (*)[4][5])ps;  // pointer to matrix
    ps[6] = 1.0;                                // *(ps +  6) = 1.0
    pv[1][2] = 1.0;                             // *(ps +  7) = 1.0
    (*pm)[3][4] = 1.0;                          // *(ps + 19) = 1.0
    pm[1][2][3] = 1.0;                          // *(ps + 33) = 1.0
    delete [] ps;

My code from post 3 works to create a two dimensional dynamically allocated array. Try it. I don't understand why you tell me that it doesn't work.

//next two lines create a dynamically allocated N by M array A[N][M]
double **A = new double *[N];
for(i = 0; i < N; i++) A[i] = new double[M];

//we can then use it like this
A[1][3] = 14;
cout << A[1][3];

What I was asking was not how to create the array (though if there is a better way, I'd be glad to hear it, this method always seemed inelegant to me), but if I later have a pointer pointing to the first element of the array, how to use it with two subscripts.

//make a pointer and point it to the first element of A
double *p;
p = A[0];

//I wanted to be able to use p something like p[1][3]
//you gave me the answer by typecasting the pointer

((double (*)[M])p)[1][3] = 14;
cout << ((double (*)[M])p)[1][3];

 

[D H]

You are talking past one another, thanks to how C (and C++) treats multi-dimensional and ragged arrays. Jeff Reid is creating multi-dimensional arrays while NeoDevin is creating ragged arrays. A multi-dimensional array occupies a monolithic block of memory and is allocated by a single new/malloc, while a ragged array comprises multiple chunks of memory and requires multiple new/malloc invocations. Both are accessed the same way: [noparse]multi_dim_array[index1][index2][/noparse] versus [noparse]ragged_array[index1][index2][/noparse].

 

[rcgldr]

If I have a pointer pointing to the first element of the array, how to use it with two subscripts.

You can't. You need a pointer to one of the pointers in A in order to use two subscripts. If you just copy the first pointer in A to a pointer to double, "p = A[0]", there's no way to get to A[1], A[2], ..., from p, because the vector of pointers only exists in A, while p only contains a single pointer. You'd have to create a duplicate of A, a pointer to a pointer to a double. You can implement this with just two allocated chunks of memory, by allocating a large vector, then setting A[] to sequentially higher locations in the large vector:

int main()
{
    int i0, i1;                         // indexes
    int d0, d1;                         // dimension sizes

    d0 = 4;
    d1 = 5;

    double **apd = new double *[d0];    // allocate vector of pointers
    double *pd = new double[d0*d1];     // allocate large vector of doubles

    double *pt = pd;                    // temp pointer to vector of doubles
    for(i0 = 0; i0 < d0; i0++){         // set pointers to vectors of doubles
        apd[i0] = pt;
        pt += d1;}

    double **ppd = apd;                 // copy of apd;

    for(i0 = 0; i0 < d0; i0++)          // access the data
        for(i1 = 0; i1 < d1; i1++)
            ppd[i0][i1] = 1.0*i0*i1;

    delete [] pd;                       // delete large vector of doubles
    delete [] apd;                      // delete vector of pointers
    return(0);
}

You are talking past one another, thanks to how C (and C++) treats multi-dimensional and ragged arrays. Jeff Reid is creating multi-dimensional arrays while NeoDevin is creating ragged arrays.

I already mentioned this in post #6: "Your code creates an array of pointers, this is not a matrix."