仅在大型数组上出现合并排序分段错误

Question

I'm working on implementing a few different sorting methods and for some reason my merge sort algorithm will not work on large data sets. The sort will work for 115,000 words but stops working when it reaches 135,000 words. Once I get this high I end up getting a segmentation fault. I do not understand where the seg fault is coming from. The sort works successfully for text files containing 5K to 125K strings.

The readFile array gets initialized with the number of words in the text file. When debugging it seems like the last numbers that get passed into the mergeSort() function are the following:

#0  0x0000000000402a87 in merge (inputString=0x7fffffbde790, from=0, mid=67499, to=134999) at mergeSort.cpp:102
    n1 = 67500
    n2 = 67500
    i = 0
    j = 0
    k = 32767
    L = <error reading variable L (value requires 2160000 bytes, which is more than max-value-size)>
    R = <error reading variable R (value requires 2160000 bytes, which is more than max-value-size)>
#1  0x0000000000402921 in mergeSort (inputString=0x7fffffbde790, from=0, to=134999) at mergeSort.cpp:88
    mid = 67499

void mergeSort(string readFile[], int from, int to) {
    if (from < to) {
        int mid = from + (to - from) / 2;
        mergeSort(readFile, from, mid);
        mergeSort(readFile, mid + 1, to);
        merge(readFile, from, mid, to);
    }
}

void merge(string readFile[], int from, int mid, int to) {
    int n1 = mid - from + 1;
    int n2 = to - mid;

    string L[n1];
    string R[n2];

    for (int i = 0; i < n1; i++) {
        L[i] = readFile[from + i];
    }
    for (int i = 0; i < n2; i++) {
        R[i] = readFile[mid + i + 1];
    }

    int i = 0;
    int j = 0;
    int k = from;

    while (i < n1 && j < n2) {
        if (L[i] <= R[j]) {
            readFile[k] = L[i];
            i++;
        } else {
            readFile[k] = R[j];
            j++;
        }
        k++;
    }
    while (i < n1) {
        readFile[k] = L[i];
        i++;
        k++;
    }
    while (j < n2) {
        readFile[k] = R[j];
        j++;
        k++;
    }
}

Answer 1

You allocate temporary arrays as automatic variables in the merge function. When the size of these arrays become too large, you lack stack space to allocate them and get undefined behavior (eg a stack overflow ).

To handle arbitrarily large arrays, you should allocate the temporary arrays with malloc or new and free them accordingly. To limit the number of allocations, you could allocate a temporary array in a wrapper and pass that recursively in the mergeSort function.

Here is a simple fix allocating temporary arrays in the merge function:

void merge(string readFile[], int from, int mid, int to) {
    int n1 = mid - from + 1;
    int n2 = to - mid;

    string *L = new string[n1];
    string *R = new string[n2];

    for (int i = 0; i < n1; i++) {
        L[i] = readFile[from + i];
    }
    for (int i = 0; i < n2; i++) {
        R[i] = readFile[mid + i + 1];
    }

    int i = 0;
    int j = 0;
    int k = from;

    while (i < n1 && j < n2) {
        if (L[i] <= R[j]) {
            readFile[k] = L[i];
            i++;
        } else {
            readFile[k] = R[j];
            j++;
        }
        k++;
    }
    while (i < n1) {
        readFile[k] = L[i];
        i++;
        k++;
    }
    while (j < n2) {
        readFile[k] = R[j];
        j++;
        k++;
    }
    delete[] L;
    delete[] R;
}

Here is a more elaborate version, possibly more efficient, allocating a single temporary array:

void merge(string readFile[], size_t from, size_t mid, size_t to, string aux[]) {
    size_t i, j, k;

    for (i = from; i < to; i++) {
        aux[i] = readFile[i];
    }

    i = from;
    j = mid;
    k = from;

    while (i < mid && j < to) {
        if (aux[i] <= aux[j]) {
            readFile[k++] = aux[i++];
        } else {
            readFile[k++] = aux[j++];
        }
    }
    while (i < mid) {
        readFile[k++] = aux[i++];
    }
    while (j < to) {
        readFile[k++] = aux[j++];
    }
}

void mergeSort(string readFile[], size_t from, size_t to, string aux[]) {
    if (to - from > 1) {
        size_t mid = from + (to - from) / 2;
        mergeSort(readFile, from, mid, aux);
        mergeSort(readFile, mid, to, aux);
        merge(readFile, from, mid, to, aux);
    }
}

void mergeSort(string readFile[], size_t n) {
    string *aux = new string[n];
    mergeSort(readFile, 0, n, aux);
    delete[] aux;
}