标签:des blog http io ar os 使用 sp for
nginx源码剖析数据结构篇(二) 双向链表ngx_queue_t
Author:Echo Chen(陈斌)
Email:chenb19870707@gmail.com
Date:October 20h, 2014
ngx_queue作为顺序容器链表,它优势在于其可以高效地执行插入、删除、合并操作,在插入删除的过程中,只需要修改指针指向,而不需要拷贝数据,因此,对于频繁修改的容器很适合。此外,相对于STL list,它还具有以下特点:
头文件:http://trac.nginx.org/nginx/browser/nginx/src/core/ngx_queue.h
源文件:http://trac.nginx.org/nginx/browser/nginx/src/core/ngx_queue.c
1: typedef struct ngx_queue_s ngx_queue_t;
2:
3: struct ngx_queue_s {
4: ngx_queue_t *prev;
5: ngx_queue_t *next;
6: };
可以看到,它的结构非常简单,仅有两个成员:prev、next,这样对于链表中元素来说,空间上只增加了两个指针的消耗。
1: //q 为链表容器结构体ngx_queue_t的指针,将头尾指针指向自己
2: #define ngx_queue_init(q) \
3: (q)->prev = q; \
4: (q)->next = q
初始状态的链表如图所示:
判断方法非常简单,即判断链表的prev指针是否指向自己,如上图所示
1: #define ngx_queue_empty(h) \
2: (h == (h)->prev)
1: //h为链表指针,x为要插入的元素
2: #define ngx_queue_insert_head(h, x) \
3: (x)->next = (h)->next; \
4: (x)->next->prev = x; \
5: (x)->prev = h; \
6: (h)->next = x
标准的双链表插入四步操作,如图所示:
与头部插入类似,只是第一步给的h->prev ,即为最后一个结点:
1: #define ngx_queue_insert_tail(h, x) \
2: (x)->prev = (h)->prev; \
3: (x)->prev->next = x; \
4: (x)->next = h; \
5: (h)->prev = x
x为要删除的结点,将x的下一个的结点的prev指针指向x的上一个结点,再将x的前一个结点的next指针指向x的下一个结点,常规链表双链表结点删除操作,不处理内存释放
1: #define ngx_queue_remove(x) \
2: (x)->next->prev = (x)->prev; \
3: (x)->prev->next = (x)->next
4:
1: #define ngx_queue_split(h, q, n) \
2: (n)->prev = (h)->prev; \
3: (n)->prev->next = n; \
4: (n)->next = q; \
5: (h)->prev = (q)->prev; \
6: (h)->prev->next = h; \
7: (q)->prev = n;
h为链表容器,q为链表h中的一个元素,这个方法可以将链表h以元素q为界拆分为两个链表h和n,其中h由原链表的前半部分组成(不包含q),而n由后半部分组成,q为首元素,操作也很简单,如图所示:
1: #define ngx_queue_add(h, n) \
2: (h)->prev->next = (n)->next; \
3: (n)->next->prev = (h)->prev; \
4: (h)->prev = (n)->prev; \
5: (h)->prev->next = h;
将链表n 合并到链表h的尾部,如图所示:
1: ngx_queue_t *ngx_queue_middle(ngx_queue_t *queue)
2: {
3: ngx_queue_t *middle, *next;
4:
5: middle = ngx_queue_head(queue);
6:
7: //头尾相等,空链表,返回头即可
8: if (middle == ngx_queue_last(queue)) {
9: return middle;
10: }
11:
12: next = ngx_queue_head(queue);
13:
14: for ( ;; ) {
15: middle = ngx_queue_next(middle);
16: next = ngx_queue_next(next);
17:
18: if (next == ngx_queue_last(queue)) {
19: return middle;
20: }
21:
22: next = ngx_queue_next(next);
23:
24: if (next == ngx_queue_last(queue)) {
25: return middle;
26: }
27: }
28: }
这里用到的技巧是每次middle向后移动一步,next向后移动两步,这样next指到队尾的时候,middle就指到了中间,时间复杂度就是O(N),这是一道经典的面试题,今天在这里看到了源码,似成相识啊,果然经典面试题目都不是凭空而来。
可以看到,这里采用的是插入排序算法,时间复杂度为O(n),整个代码非常简洁。
1: void ngx_queue_sort(ngx_queue_t *queue, ngx_int_t (*cmp)(const ngx_queue_t *, const ngx_queue_t *))
2: {
3: ngx_queue_t *q, *prev, *next;
4:
5: q = ngx_queue_head(queue);
6:
7: //如果是空链表,直接返回
8: if (q == ngx_queue_last(queue)) {
9: return;
10: }
11:
12: for (q = ngx_queue_next(q); q != ngx_queue_sentinel(queue); q = next) {
13:
14: prev = ngx_queue_prev(q);
15: next = ngx_queue_next(q);
16:
17: ngx_queue_remove(q);
18:
19: //找到插入位置
20: do {
21: if (cmp(prev, q) <= 0) {
22: break;
23: }
24:
25: prev = ngx_queue_prev(prev);
26:
27: } while (prev != ngx_queue_sentinel(queue));
28:
29: //插入
30: ngx_queue_insert_after(prev, q);
31: }
32: }
1: #define ngx_queue_data(q, type, link) \
2: (type *) ((u_char *) q - offsetof(type, link))
其中q为ngx_queue_t* 类型,函数作用为根据q算出,算出链表元素的地址,其中linux接口offsetof是算出link在type中的偏移。
1: #define ngx_queue_head(h) \
2: (h)->next
3:
4:
5: #define ngx_queue_last(h) \
6: (h)->prev
7:
8:
9: #define ngx_queue_sentinel(h) \
10: (h)
11:
12:
13: #define ngx_queue_next(q) \
14: (q)->next
15:
16:
17: #define ngx_queue_prev(q) \
18: (q)->prev
1: #include <iostream>
2: #include <algorithm>
3: #include <pthread.h>
4: #include <time.h>
5: #include <stdio.h>
6: #include <errno.h>
7: #include <string.h>
8: #include "ngx_queue.h"
9:
10: struct student_info
11: {
12: long stu_id;
13: unsigned int age;
14: unsigned int score;
15: ngx_queue_t qEle;
16: };
17:
18: ngx_int_t compareStudent(const ngx_queue_t *a, const ngx_queue_t *b)
19: {
20: //分别取得a b 对象指针
21: student_info *ainfo = ngx_queue_data(a,student_info,qEle);
22: student_info *binfo = ngx_queue_data(b,student_info,qEle);
23:
24: return ainfo->score >binfo->score;
25: }
26:
27: void print_ngx_queue(ngx_queue_t *queue)
28: {
29: //遍历输出
30: for(ngx_queue_t *q = ngx_queue_head(queue);q != ngx_queue_sentinel(queue);q = ngx_queue_next(q))
31: {
32: student_info *info = ngx_queue_data(q,student_info,qEle);
33: if(info != NULL)
34: {
35: std::cout <<info->score << " ";
36: }
37: }
38:
39: std::cout << std::endl;
40: }
41:
42: int main()
43: {
44:
45: ngx_queue_t queue;
46: ngx_queue_init(&queue);
47:
48: student_info info[5];
49: for(int i = 0;i < 5;i++)
50: {
51: info[i].stu_id = i;
52: info[i].age = i;
53: info[i].score = i;
54:
55: if(i%2)
56: {
57: ngx_queue_insert_tail(&queue,&info[i].qEle);
58: }
59: else
60: {
61: ngx_queue_insert_head(&queue,&info[i].qEle);
62: }
63: }
64:
65: print_ngx_queue(&queue);
66:
67: ngx_queue_sort(&queue,compareStudent);
68:
69: print_ngx_queue(&queue);
70:
71: return 0;
72: }
输出结果:
ngx_queue设计非常精巧,基本涵盖了双链表的所有操作,建议需要面试的童鞋看一看,很多链表的题目都迎刃而解。此外,ngx_queue与其它nginx 代码耦合度低,有需要这种双向链表的实现时不妨直接拿过来使用。
-
菜鸟nginx源码剖析数据结构篇(二) 双向链表ngx_queue_t[转]
标签:des blog http io ar os 使用 sp for
原文地址:http://www.cnblogs.com/0x2D-0x22/p/4139784.html