对c库中qsort函数的探究
最近有朋友探讨qsort函数的问题，于是去C库中看看它是怎么实现的。
于是在glibc中找到Qsort.c略作修改，用里面的_quicksort函数顶替了C库里的qsort函数，还真能运行。
下面是我做测试的三个文件。
qst_bsech.h
#include <stdio.h>
typedef int (*__compar_d_fn_t) (__const void *, __const void *, void *);
typedef struct
{
char key[10];
int other_data;
} Record;
extern void //这一句要加上
_quicksort (void *const pbase, size_t total_elems, size_t size, __compar_d_fn_t cmp, void *arg);
qst_bsech.c (就是Qsort.c，原文件的注释还带着）
#include <alloca.h>
#include <limits.h>
#include <string.h>
#include "qst_bsech.h"
/* Byte-wise swap two items of size SIZE. */
#define SWAP(a, b, size) \
do \
{ \
register size_t __size = (size); \
register char *__a = (a), *__b = (b); \
do \
{ \
char __tmp = *__a; \
*__a++ = *__b; \
*__b++ = __tmp; \
} while (--__size > 0); \
} while (0)
/* Discontinue quicksort algorithm when partition gets below this size.
This particular magic number was chosen to work best on a Sun 4/260. */
#define MAX_THRESH 4
/* Stack node declarations used to store unfulfilled partition obligations. */
typedef struct
{
char *lo;
char *hi;
} stack_node;
/* The next 4 #defines implement a very fast in-line stack abstraction. */
/* The stack needs log (total_elements) entries (we could even subtract
log(MAX_THRESH)). Since total_elements has type size_t, we get as
upper bound for log (total_elements):
bits per byte (CHAR_BIT) * sizeof(size_t). */
#define STACK_SIZE (CHAR_BIT * sizeof(size_t))
#define PUSH(low, high) ((void) ((top->lo = (low)), (top->hi = (high)), ++top))
#define POP(low, high) ((void) (--top, (low = top->lo), (high = top->hi)))
#define STACK_NOT_EMPTY (stack < top)
/* Order size using quicksort. This implementation incorporates
four optimizations discussed in Sedgewick:
1. Non-recursive, using an explicit stack of pointer that store the
next array partition to sort. To save time, this maximum amount
of space required to store an array of SIZE_MAX is allocated on the
stack. Assuming a 32-bit (64 bit) integer for size_t, this needs
only 32 * sizeof(stack_node) == 256 bytes (for 64 bit: 1024 bytes).
Pretty cheap, actually.
2. Chose the pivot element using a median-of-three decision tree.
This reduces the probability of selecting a bad pivot value and
eliminates certain extraneous comparisons.
3. Only quicksorts TOTAL_ELEMS / MAX_THRESH partitions, leaving
insertion sort to order the MAX_THRESH items within each partition.
This is a big win, since insertion sort is faster for small, mostly
sorted array segments.
4. The larger of the two sub-partitions is always pushed onto the
stack first, with the algorithm then concentrating on the
smaller partition. This *guarantees* no more than log (total_elems)
stack size is needed (actually O(1) in this case)! */
void
_quicksort (void *const pbase, size_t total_elems, size_t size,
__compar_d_fn_t cmp, void *arg)
{
register char *base_ptr = (char *) pbase;
const size_t max_thresh = MAX_THRESH * size;
if (total_elems == 0)
/* Avoid lossage with unsigned arithmetic below. */
return;
if (total_elems > MAX_THRESH)
{
char *lo = base_ptr;
char *hi = &lo[size * (total_elems - 1)];
stack_node stack[STACK_SIZE];
stack_node *top = stack;
PUSH (NULL, NULL);
while (STACK_NOT_EMPTY)
{
char *left_ptr;
char *right_ptr;
/* Select median value from among LO, MID, and HI. Rearrange
LO and HI so the three values are sorted. This lowers the
probability of picking a pathological pivot value and
skips a comparison for both the LEFT_PTR and RIGHT_PTR in
the while loops. */
char *mid = lo + size * ((hi - lo) / size >> 1);
if ((*cmp) ((void *) mid, (void *) lo, arg) < 0)
SWAP (mid, lo, size);
if ((*cmp) ((void *) hi, (void *) mid, arg) < 0)
SWAP (mid, hi, size);
else
goto jump_over;
if ((*cmp) ((void *) mid, (void *) lo, arg) < 0)
SWAP (mid, lo, size);
jump_over:;
left_ptr = lo + size;
right_ptr = hi - size;
/* Here's the famous ``collapse the walls'' section of quicksort.
Gotta like those tight inner loops! They are the main reason
that this algorithm runs much faster than others. */
do
{
while ((*cmp) ((void *) left_ptr, (void *) mid, arg) < 0)
left_ptr += size;
while ((*cmp) ((void *) mid, (void *) right_ptr, arg) < 0)
right_ptr -= size;
if (left_ptr < right_ptr)
{
SWAP (left_ptr, right_ptr, size);
if (mid == left_ptr)
mid = right_ptr;
else if (mid == right_ptr)
mid = left_ptr;
left_ptr += size;
right_ptr -= size;
}
else if (left_ptr == right_ptr)
{
left_ptr += size;
right_ptr -= size;
break;
}
}
while (left_ptr <= right_ptr);
/* Set up pointers for next iteration. First determine whether
left and right partitions are below the threshold size. If so,
ignore one or both. Otherwise, push the larger partition's
bounds on the stack and continue sorting the smaller one. */
if ((size_t) (right_ptr - lo) <= max_thresh)
{
if ((size_t) (hi - left_ptr) <= max_thresh)
/* Ignore both small partitions. */
POP (lo, hi);
else
/* Ignore small left partition. */
lo = left_ptr;
}
else if ((size_t) (hi - left_ptr) <= max_thresh)
/* Ignore small right partition. */
hi = right_ptr;
else if ((right_ptr - lo) > (hi - left_ptr))
{
/* Push larger left partition indices. */
PUSH (lo, right_ptr);
lo = left_ptr;
}
else
{
/* Push larger right partition indices. */
PUSH (left_ptr, hi);
hi = right_ptr;
}
}
}
/* Once the BASE_PTR array is partially sorted by quicksort the rest
is completely sorted using insertion sort, since this is efficient
for partitions below MAX_THRESH size. BASE_PTR points to the beginning
of the array to sort, and END_PTR points at the very last element in
the array (*not* one beyond it!). */
#define min(x, y) ((x) < (y) ? (x) : (y))
{
char *const end_ptr = &base_ptr[size * (total_elems - 1)];
char *tmp_ptr = base_ptr;
char *thresh = min(end_ptr, base_ptr + max_thresh);
register char *run_ptr;
/* Find smallest element in first threshold and place it at the
array's beginning. This is the smallest array element,
and the operation speeds up insertion sort's inner loop. */
for (run_ptr = tmp_ptr + size; run_ptr <= thresh; run_ptr += size)
if ((*cmp) ((void *) run_ptr, (void *) tmp_ptr, arg) < 0)
tmp_ptr = run_ptr;
if (tmp_ptr != base_ptr)
SWAP (tmp_ptr, base_ptr, size);
/* Insertion sort, running from left-hand-side up to right-hand-side. */
run_ptr = base_ptr + size;
while ((run_ptr += size) <= end_ptr)
{
tmp_ptr = run_ptr - size;
while ((*cmp) ((void *) run_ptr, (void *) tmp_ptr, arg) < 0)
tmp_ptr -= size;
tmp_ptr += size;
if (tmp_ptr != run_ptr)
{
char *trav;
trav = run_ptr + size;
while (--trav >= run_ptr)
{
char c = *trav;
char *hi, *lo;
for (hi = lo = trav; (lo -= size) >= tmp_ptr; hi = lo)
*hi = *lo;
*hi = c;
}
}
}
}
}
main.c
#include <stdlib.h>
#include <string.h>
#include "qst_bsech.h"
#define N 5
/*之所以有两个compare就是为了让_quicksort和bsearch都有的用*/
static int r_compare (void const *a, void const *b, void *arg)
{
return strcmp ( ((Record *)a)->key, ((Record *)b)->key );
}
static int b_compare (void const *a, void const *b)
{
return strcmp ( ((Record *)a)->key, ((Record *)b)->key );
}
static void print ( Record *parray, int n )
{
int i;
for ( i=0; i<n; i++ )
printf( "%s", parray[i].key );
printf ( "\n" );
}
int main (int argc, char *argv[])
{
int i;
Record array[N];
Record key;
Record *ans;
for ( i=0; i<N; i++ )
{
fgets ( array[i].key, 9, stdin);
array[i].other_data = i;
}
_quicksort( array, N, sizeof(Record), r_compare, NULL);
print ( array, N );
strcpy ( key.key, "much\n" );
//注意这里容易犯错，fgets会在读入的字符串末尾加回车，而strcpy不会。
printf ("Looking for %s ...\n", key.key);
ans = bsearch ( &key, array, N, sizeof(Record), b_compare );
if (ans != NULL)
printf ( "%d: %s\n", ans->other_data, ans->key);
return 0;
}
编译直接gcc qst_bsech.h qst_bsech.c main.c
运行实例：
[12:14:46 @cfiles]$ ./a.out
I
love
you
very
much
I
love
much
very
you
Looking for much
...
4: much
排序和查找都正常工作，不过说实话_quicksort我没有完全看明白，大家看看，欢迎讨论，高手来讲解一下不胜荣幸。
