redis中动态字符串sds相关的文件为:adlist.h与adlist.c
一、数据结构
redis里定义的双向链表,与普通双向链表大致相同
单个节点:
1 typedef struct listNode { 2 struct listNode *prev; 3 struct listNode *next; 4 void *value; 5 } listNode;
链表:
1 typedef struct list { 2 listNode *head; 3 listNode *tail; 4 void *(*dup)(void *ptr); 5 void (*free)(void *ptr); 6 int (*match)(void *ptr, void *key); 7 unsigned long len; 8 } list;
链表以函数指针的方式,实现了复制、销毁与比较的方法的多态。
迭代器:
1 typedef struct listIter { 2 listNode *next; 3 int direction; 4 } listIter;
迭代器中有个成员变量direction,用于表示当前遍历的方向。
大致结构:
1 /* 2 +-------------------+ +----------------> +--------------+ <-------+ 3 |listNode *head |--------+ |listNode *prev|-->NULL | 4 +-------------------+ +--------------+ | 5 |listNode *tail |--------+ |listNode *next|----+ | 6 +-------------------+ | +--------------+ | | 7 |void *(*dup)(...) | | |void *value | | | 8 +-------------------+ | +--------------+ | | 9 |void (*free)(...) | | | | 10 +-------------------+ | | | 11 |int (*match)(...) | | | | 12 +-------------------+ +----------------> +--------------+ <--+ | 13 |unsigned long len | |listNode *prev|---------+ 14 +-------------------+ +--------------+ 15 |listNode *next|-->NULL 16 +--------------+ 17 |void *value | 18 +--------------+ 19 */
二、创建
redis中创建一个初始双向链表比较简单,只要分配好内存,并给成员变量赋初值就可以了
1 list *listCreate(void) 2 { 3 struct list *list; 4 5 if ((list = zmalloc(sizeof(*list))) == NULL) 6 return NULL; 7 list->head = list->tail = NULL; 8 list->len = 0; 9 list->dup = NULL; 10 list->free = NULL; 11 list->match = NULL; 12 return list; 13 }
redis中提供了头插法、尾插法以及指定位置插入节点三种方式向链表中添加节点,与普通双向链表无异,此处不做详细叙述。
三、销毁
因链表中每个节点的value可能指向堆空间,故不能直接把list结构体free,这样会造成内存泄露。需要先将每个节点的value释放,才可以free结构体
清空所有节点:
1 void listEmpty(list *list) 2 { 3 unsigned long len; 4 listNode *current, *next; 5 6 current = list->head; 7 len = list->len; 8 while(len--) { 9 next = current->next; 10 //若指定了销毁的函数,则使用指定的函数进行销毁value 11 if (list->free) list->free(current->value); 12 zfree(current); 13 current = next; 14 } 15 list->head = list->tail = NULL; 16 list->len = 0; 17 }
销毁链表:
1 void listRelease(list *list) 2 { 3 listEmpty(list); 4 zfree(list); 5 }
同样,redis的链表提供了与普通链表相同的删除单个节点的操作,此处也不做叙述。
四、迭代器操作
redis中提供了获取迭代器的接口
1 listIter *listGetIterator(list *list, int direction) 2 { 3 listIter *iter; 4 5 if ((iter = zmalloc(sizeof(*iter))) == NULL) return NULL; 6 if (direction == AL_START_HEAD) 7 iter->next = list->head; 8 else 9 iter->next = list->tail; 10 iter->direction = direction; 11 return iter; 12 }
以AL_START_HEAD为例,生成好的迭代器结构如下:
1 /* 2 +-------------------+ +---> +--------------+ <-------+----+ 3 |listNode *head |----+ |listNode *prev|-->NULL | | 4 +-------------------+ +--------------+ | | +--------------+ 5 |listNode *tail |----+ |listNode *next|----+ | +--|listNode *next| 6 +-------------------+ | +--------------+ | | +--------------+ 7 |void *(*dup)(...) | | |void *value | | | |int direction | 8 +-------------------+ | +--------------+ | | +--------------+ 9 |void (*free)(...) | | | | 10 +-------------------+ | | | 11 |int (*match)(...) | | | | 12 +-------------------+ +---> +--------------+ <--+ | 13 |unsigned long len | |listNode *prev|---------+ 14 +-------------------+ +--------------+ 15 |listNode *next|-->NULL 16 +--------------+ 17 |void *value | 18 +--------------+ 19 */
迭代器的next方法:
1 listNode *listNext(listIter *iter) 2 { 3 listNode *current = iter->next; 4 5 if (current != NULL) { 6 if (iter->direction == AL_START_HEAD) 7 iter->next = current->next; 8 else 9 iter->next = current->prev; 10 } 11 return current; 12 }
调用一次之后的结构:
1 /* 2 +-------------------+ +---> +--------------+ <-------+ 3 |listNode *head |----+ |listNode *prev|-->NULL | 4 +-------------------+ +--------------+ | +--------------+ 5 |listNode *tail |----+ |listNode *next|----+ | +--|listNode *next| 6 +-------------------+ | +--------------+ | | | +--------------+ 7 |void *(*dup)(...) | | |void *value | | | | |int direction | 8 +-------------------+ | +--------------+ | | | +--------------+ 9 |void (*free)(...) | | | | | 10 +-------------------+ | | | | 11 |int (*match)(...) | | | | | 12 +-------------------+ +---> +--------------+ <--+----|----+ 13 |unsigned long len | |listNode *prev|---------+ 14 +-------------------+ +--------------+ 15 |listNode *next|-->NULL 16 +--------------+ 17 |void *value | 18 +--------------+ 19 */
再次调用:
1 /* 2 +-------------------+ +---> +--------------+ <-------+ 3 |listNode *head |----+ |listNode *prev|-->NULL | 4 +-------------------+ +--------------+ | +--------------+ 5 |listNode *tail |----+ |listNode *next|----+ | +--|listNode *next| 6 +-------------------+ | +--------------+ | | | +--------------+ 7 |void *(*dup)(...) | | |void *value | | | | |int direction | 8 +-------------------+ | +--------------+ | | | +--------------+ 9 |void (*free)(...) | | | | | 10 +-------------------+ | | | | 11 |int (*match)(...) | | | | | 12 +-------------------+ +---> +--------------+ <--+ | +-->NULL 13 |unsigned long len | |listNode *prev|---------+ 14 +-------------------+ +--------------+ 15 |listNode *next|-->NULL 16 +--------------+ 17 |void *value | 18 +--------------+ 19 */
调用next函数的返回值为调用之前的listNode首地址
五、其它操作
redis的双向链表还提供了其它操作。其中,查找指定的key与复制整个list依赖于迭代器的使用,并使用到自定义的比较/复制方法。
除此之外,还提供了类似随机读取的方式,其内部实现为遍历,且“越界”时返回NULL。同时,它支持index为负数,表示从尾开始。类似旋转的操作,把尾节点移至原头节点之前,成为新的头节点。当然,还有拼接两个链表的操作。
redis 5.0.7 下载链接
http://download.redis.io/releases/redis-5.0.7.tar.gz
源码阅读顺序参考:
https://github.com/huangz1990/blog/blob/master/diary/2014/how-to-read-redis-source-code.rst