實(shí)用排序算法（復(fù)雜度小于等于O(n^2)）中效率最低但實(shí)現(xiàn)并不是最簡(jiǎn)單的的兩個(gè)，C、C++教材卻總喜歡拿來(lái)大講特講，非常不利于初學(xué)者養(yǎng)成“程序效率”的思維。

實(shí)際上，各種排序算法里，除了堆排序?qū)崿F(xiàn)較為復(fù)雜外，從代碼量的角度，大多數(shù)算法都不比冒泡、選擇算法復(fù)雜多少。

舉幾個(gè)高速的排序算法的例子，這些才適合進(jìn)入教材

鴿巢排序，排序字節(jié)串、寬字節(jié)串最快的排序算法，計(jì)數(shù)排序的變種（將計(jì)數(shù)緩沖區(qū)大小固定，少一次遍歷開銷），速度是STL中std::sort的20多倍，更重要的是實(shí)現(xiàn)極其簡(jiǎn)單！缺點(diǎn)是需要一個(gè)size至少等于待排序數(shù)組取值范圍的緩沖區(qū)，不適合int等大范圍數(shù)據(jù)
C/C++ code

void PigeonholeSort(BYTE *array, int length)
{
    int b[256] = {0};
    int i,k,j = 0;
    for(i=0; i<length; i++)
        b[array[i]]++;
    for(i=0; i<256; i++)
        for(k=0; k<b[i]; k++)
            array[j++] = i;
}

多一次遍歷的計(jì)數(shù)排序，排序字節(jié)串的話速度約是鴿巢排序的一半
C/C++ code

void CountingSort(BYTE *array, int length)
{
    int t;
    int i, z = 0;
    BYTE min,max;
    int *count;
    min = max = array[0];
    for(i=0; i<length; i++)
    {
        if(array[i] < min)
            min = array[i];
        else if(array[i] > max)
            max = array[i];
    }
    count = (int*)malloc((max-min+1)*sizeof(int));
    for(i=0; i<max-min+1; i++)
        count[i] = 0;
    for(i = 0; i < length; i++)
        count[array[i]-min]++;

    for(t = 0; t <= 255; t++)
        for(i = 0; i < count[t-min]; i++)
            array[z++] = (BYTE)t;
    free(count);
}

快速排序，快排最標(biāo)準(zhǔn)的遞歸實(shí)現(xiàn)，速度約是std::sort的一半
C/C++ code

void swap(BYTE *a,BYTE *b)
{
    BYTE tmp;
    if ( a != b )
    {
        tmp = *a;
        *a = *b;
        *b = tmp;
    }
}

int partition(BYTE *arr,int left, int right)
{
    int i = left - 1, j = right;
    BYTE v = arr[right];
    while(1)
    {
        while(arr[++i] < v);
        while(arr[--j] > v)
            if(j == 1)
                break;
        if(i >= j)
            break;
        swap(&arr[i],&arr[j]);
    }
    swap(&arr[i],&arr[right]);
    return i;
}

void quicksort(BYTE *arr, int left, int right)
{
    if (left < right)
    {
        int i = partition(arr,left,right);

        quicksort(arr,left,i-1);
        quicksort(arr,i+1,right);
    }
}

void QuickSort(BYTE *array,int length)
{
    quicksort(array,0,length-1);
}

這是速度與std::sort相當(dāng)?shù)娜穭澐挚炫?br> C/C++ code

void swap(BYTE *a,BYTE *b)
{
    BYTE tmp;
    if ( a != b )
    {
        tmp = *a;
        *a = *b;
        *b = tmp;
    }
}

void quicksort(BYTE *arr, int left, int right)
{
    if (left < right)
    {
        BYTE v = arr[right];
        int i = left - 1,j = right,p = left - 1,q = right,k=0;
        while (1)
        {
            while (arr[++i] < v);
            while (arr[--j] > v)
                if(j==left)
                    break;
            if (i >= j)
                break;
            swap(&arr[i], &arr[j]);
            if(arr[i] == v)
            {
                p++;
                swap(&arr[p],&arr[i]);
            }
            if(arr[j] == v)
            {
                q--;
                swap(&arr[q],&arr[j]);
            }
        }
        swap(&arr[i],&arr[right]);
        j = i - 1;
        i++;
        for(k=left; k<=p; k++,j--)
            swap(&arr[k],&arr[j]);
        for(k=right-1; k>=q; k--,i++)
            swap(&arr[k],&arr[i]);
        quicksort(arr,left,j);
        quicksort(arr,i,right);
    }
}

void QuickSort(BYTE *array,int length)
{
    quicksort(array,0,length-1);
}

相當(dāng)簡(jiǎn)單的梳排序，效率是std::sort的三分之一
C/C++ code

void CombSort(BYTE *arr, int size)
{
    UINT gap = size, swapped = 1, i = 0;
    BYTE swap = 0;
    while ((gap > 1) || swapped)
    {
        if (gap > 1)
            gap = gap / 1.3;
        swapped = 0;
        i = 0;
        while ((gap + i) < size)
        {
            if (arr[i] - arr[i + gap] > 0)
            {
                swap = arr[i];
                arr[i] = arr[i + gap];
                arr[i + gap] = swap;
                swapped = 1;
            }
            ++i;
        }
    }
}

LSD基數(shù)排序，與std::sort速度相當(dāng)，但是需要一個(gè)與輸入緩沖一樣大的緩沖區(qū)
C/C++ code

#define R 256

#define digit(a, d)    ( a >> 8*d )

static BYTE *aux;

void radix_sort(BYTE *arr, int left, int right)
{
    if(left < right)
    {
        int d = 0;
        for(d=3; d>=0; d--)
        {
            int i=0, j=0, count[R+1];
            for(j=0; j<R; j++)
                count[j] = 0;
            for(i=left; i<=right; i++)
                count[digit(arr[i],d) + 1]++;
            for(j=1; j<R; j++)
                count[j] += count[j-1];
            for(i=left; i<=right; i++)
                aux[count[digit(arr[i],d)]++] = arr[i];
            for(i=left; i<=right; i++)
                arr[i] = aux[i-1];
        }   
    }
}

void RadixSort(BYTE *array,int length)
{
    aux = (BYTE*)malloc(length);
    radix_sort(array,0,length-1);
    free(aux);
}

歸并排序，效率越是std::sort的六分之一，通常的實(shí)現(xiàn)是遞歸，但和快排不同，歸并改循環(huán)極其容易
C/C++ code

void merge(BYTE *array, int low, int mid, int high)
{
    int i, k;
    BYTE *temp = (BYTE *) malloc(high-low+1);
    int begin1 = low;
    int end1 = mid;
    int begin2 = mid + 1;
    int end2 = high;

    for (k = 0; begin1 <= end1 && begin2 <= end2; ++k)
        if(array[begin1]<array[begin2])
            temp[k] = array[begin1++];
        else
            temp[k] = array[begin2++];   
    while(begin1<=end1)
        temp[k++] = array[begin1++];
    while(begin2<=end2)
        temp[k++] = array[begin2++];
    for (i = 0; i < (high-low+1); i++)
        array[low+i] = temp[i];
    free(temp);
}

void merge_sort(BYTE *array, UINT first, UINT last)
{
    UINT mid,i;
    for(mid=1; mid<=last-first; mid += mid)
        for(i=first; i<=last-mid; i+=mid+mid)
            merge(array,i,i+mid-1,min(i+mid+mid-1,last));
}

void MergeSort(BYTE *array, UINT length)
{
    merge_sort(array,0,length-1);
}