Home  >  Article  >  Database  >  redis study notes-list principle

redis study notes-list principle

Golang菜鸟
Golang菜鸟forward
2023-08-07 16:36:141501browse

list Basic functions

##BRPOP key1 [key2 ] timeout##LREM key count valueRemove list element##LSET key index valueLTRIM key start stop

Chart from: https://www.cnblogs.com/amyzhu/p/13466311.html

Singly linked list

Before learning the list implementation of redis, let’s first take a look at how to implement a singly linked list:

Each one The nodes have a backward pointer (reference) pointing to the next node, the last node points to NULL to indicate the end, and there is a Head (head) pointer pointing to the first node to indicate the start.

redis study notes-list principle

##Similar to this, although new creation and deletion only require O(1) , but the search requires O(n) time; the reverse search is not possible, and you need to start from the beginning if you miss it.

Add a node:

redis study notes-list principle

Delete a node:

redis study notes-list principle

##Doubly linked list

Doubly linked list, also called doubly linked list, is a type of linked list. Each data node has two pointers, pointing to the direct successor and direct predecessor respectively. Therefore, starting from any node in the doubly linked list, you can easily access its predecessor nodes and successor nodes.

redis study notes-list principle

Features:

  1. Every time you insert or delete When selecting a certain node, four node references need to be processed instead of two. It is more difficult to implement

  2. Compared with a one-way linked list, it will inevitably occupy memory The space is larger.

  3. You can traverse from the beginning to the end, and you can traverse from the end to the beginning

This seems to solve the problem of redis being able to traverse both before and after.

Then let’s take a look at how redis’s linked list is processed:

redis study notes-list principle

Let’s take a look at its structure definition source code

ListNode:

typedef struct listNode
{
    // 前驱节点
    struct listNode *prev;
    // 后继节点
    struct listNode *next;
    // 节点的值
    void *value;
} listNode;

List:

typedef struct listNode
{
    // 表头节点
    listNode *head;
    // 表尾节点
    listNode *tail;
    // 链表所包含的节点数量
    unsigned long len;
    // 节点值复制函数
    void *(*dup)(void *ptr);
    // 节点值释放函数
    void *(*free)(void *ptr);
    // 节点值对比函数
    int (*match)(void *ptr,void *key)
} list;

Characteristics of redis linked list:

  • Bidirectional acyclic: linked list nodes have predecessor and successor pointers to obtain a node's The time complexity of the predecessor and successor nodes is O(1). The predecessor pointer of the head node and the successor pointer of the tail node both point to NULL, and access to the linked list ends with NULL.

  • Length counter: The time complexity of obtaining the number of nodes through the len attribute of the List structure is O(1).

Since list still has a problem of discontinuous memory allocation and memory fragmentation, is there a way to optimize their memory?

redis Compressed List

ZipList is not a basic data structure, but a data storage structure designed by Redis itself. It is somewhat similar to an array, storing data through a continuous memory space.

Different from an array, it allows the stored list elements to occupy different memory spaces. When it comes to the word compression, the first thing that everyone may think of is saving memory. The reason why this storage structure saves memory is that it is compared to arrays.

We all know that arrays require the storage space of each element to be the same size. If we want to store strings of different lengths, we must use the storage space occupied by the string with the maximum length. As the size of the storage space for each element of the string array (if it is 50 bytes).

Therefore, part of the storage space will be wasted when storing a string less than 50 bytes in character value.

The advantage of array is that it occupies a continuous space and can make good use of the CPU cache to quickly access data.

If you want to retain this advantage of the array and save storage space, then we can compress the array:

redis study notes-list principle

However, there is a problem. When traversing the compressed list, we do not know the memory size occupied by each element, so we cannot calculate the specific starting position of the next element.

But then I thought about it, if we could have the length of each element before accessing it, wouldn't this problem be solved?

redis study notes-list principle

Next let’s look at how Redis combines them by implementing ZipList to retain the advantages of arrays and save memory.

redis study notes-list principle

  • zlbytes:压缩列表的字节长度,是uint32_t类型,占4字节,因此压缩列表最多有232-1字节,可以通过该值计算zlend的位置,也可以在内存重新分配的时候使用。

  • zltail:压缩列表尾元素相对于压缩列表起始地址的偏移量,是uint32_t类型,占4字节,可以通过该值快速定位到列表尾元素的位置。

  • zllen:压缩列表元素的个数,是uint16_t类型,占2字节,表示最多存储的元素个数为216-1=65 535,如果需要计算总数量,则需要遍历整个压缩列表。

  • entryx:压缩列表存储的元素,既可以是字节数组,也可以是整数,长度不限。

  • zlend:压缩列表的结尾,是uint8_t类型,占1字节,恒为0xFF。

  • previous_entry_length:表示前一个元素的字节长度,占1字节或者5字节,当前元素的长度小于254时用1字节,大于等于254时用5字节,previous_entry_length 字段的第一个字节固定是0xFE,后面的4字节才是真正的前一个元素的长度。

  • encoding:表示元素当前的编码,有整数或者字节数。为了节省内存,encoding字段长度可变。

  • content:表示当前元素的内容。

ZipList变量的读取和赋值都是通过宏来实现的,代码片段如下:

//返回整个压缩列表的总字节
#define ZIPLIST_BYTES(zl)       (*((uint32_t*)(zl)))

//返回压缩列表的tail_offset变量,方便获取最后一个节点的位置
#define ZIPLIST_TAIL_OFFSET(zl) (*((uint32_t*)((zl)+sizeof(uint32_t))))

//返回压缩列表的节点数量
#define ZIPLIST_LENGTH(zl)      (*((uint16_t*)((zl)+sizeof(uint32_t)*2)))

//返回压缩列表的表头的字节数
//(内存字节数zlbytes,最后一个节点地址ztail_offset,节点总数量zllength)
#define ZIPLIST_HEADER_SIZE     (sizeof(uint32_t)*2+sizeof(uint16_t))

//返回压缩列表最后结尾的字节数
#define ZIPLIST_END_SIZE        (sizeof(uint8_t))

//返回压缩列表首节点地址
#define ZIPLIST_ENTRY_HEAD(zl)  ((zl)+ZIPLIST_HEADER_SIZE)

//返回压缩列表尾节点地址
#define ZIPLIST_ENTRY_TAIL(zl)  ((zl)+intrev32ifbe(ZIPLIST_TAIL_OFFSET(zl)))

//返回压缩列表最后结尾的地址
#define ZIPLIST_ENTRY_END(zl)   ((zl)+intrev32ifbe(ZIPLIST_BYTES(zl))-1)


对此,可以通过定义的结构体和对应的操作定义知道大概 redis 设计 压缩list 的思路;

这种方式虽然节约了空间,但是会存在每次插入和创建存在内存重新分配的问题。

quicklist

quicklist是Redis 3.2中新引入的数据结构,能够在时间效率和空间效率间实现较好的折中。Redis中对quciklist的注释为A doubly linked list of ziplists。顾名思义,quicklist是一个双向链表,链表中的每个节点是一个ziplist结构。quicklist可以看成是用双向链表将若干小型的ziplist连接到一起组成的一种数据结构。

当ziplist节点个数过多,quicklist退化为双向链表,一个极端的情况就是每个ziplist节点只包含一个entry,即只有一个元素。当ziplist元素个数过少时,quicklist可退化为ziplist,一种极端的情况就是quicklist中只有一个ziplist节点。

redis study notes-list principle

redis study notes-list principle


quicklist 结构定义:

typedef struct quicklist {
    // 指向quicklist的首节点
    quicklistNode *head;
    // 指向quicklist的尾节点
    quicklistNode *tail;
    // quicklist中元素总数
    unsigned long count;        /* total count of all entries in all ziplists */
    // quicklistNode节点个数
    unsigned long len;          /* number of quicklistNodes */
    // ziplist大小设置,存放list-max-ziplist-size参数的值
    int fill : 16;              /* fill factor for individual nodes */
    // 节点压缩深度设置,存放list-compress-depth参数的值
    unsigned int compress : 16; /* depth of end nodes not to compress;0=off */
    unsigned int bookmark_count: 4;
    quicklistBookmark bookmarks[];
} quicklist;

typedef struct quicklistBookmark {
    quicklistNode *node;
    char *name;
} quicklistBookmark;

quicklistNode定义如下:

typedef struct quicklistNode {
    struct quicklistNode *prev;
    struct quicklistNode *next;
    unsigned char *zl;
    unsigned int sz;             /* ziplist size in bytes */
    unsigned int count : 16;     /* count of items in ziplist */
    unsigned int encoding : 2;   /* RAW==1 or LZF==2 */
    unsigned int container : 2;  /* NONE==1 or ZIPLIST==2 */
    unsigned int recompress : 1; /* was this node previous compressed? */
    unsigned int attempted_compress : 1; /* node can't compress; too small */
    unsigned int extra : 10; /* more bits to steal for future usage */
} quicklistNode;

redis为了节省内存空间,会将quicklist的节点用LZF压缩后存储,但这里不是全部压缩,可以配置compress的值,compress为0表示所有节点都不压缩,否则就表示从两端开始有多少个节点不压缩;compress如果是1,表示从两端开始,有1个节点不做LZF压缩。compress默认是0(不压缩),具体可以根据你们业务实际使用场景去配置。

redis 的配置文件可以配置该参数

list-compress-depth 0

##Commands Description
#BLPOP key1,key2,... timeout Remove and get the first element of the list, if the list has no elements it will block The list waits until it times out or the element is popped.
##Move out and Get the last element of the list. If there is no element in the list, the list will be blocked until the wait times out or a pop-up element is found.
BRPOPLPUSH source destination timeout pop from list A value that inserts the popped element into another list and returns it; if the list has no elements, the list will be blocked until the wait times out or a popupable element is found.
LIndex key index 通过索引获取列表中的元素
Linsert key before/after pivot value 在列表的元素前或者后插入元素
LLEN key 获取列表长度
LPOP key 移出并获取列表的第一个元素
##LPUSH key value1,value2,… will One or more values ​​are inserted into the head of the list
LPUSHX key value Insert a value into the head of an existing list
##LRANGE key srart stop Get the elements within the specified range of the list
Set the value of a list element by index
Pruning a list means that only the elements within the specified range are retained in the list, and the elements that are not within the specified range are deleted. The index starts from 0, and the range is inclusive.
RPOP key remove listThe last element, the return value is the removed element
RPOPPUSH source destination Remove the last element of the list and replace The element is added to another list and returns
RPUSH key value1 value2 …… Add one or more values ​​to the end of the list
##RPUSHX key value Add a value to an already existing list
##0##1
##Value ##Meaning
##Special value means no compression
##There is 1 on each end of the quicklist The nodes are not compressed, the middle nodes are compressed 2
There are 2 nodes at both ends of the quicklist that are not compressed, and the nodes in the middle are compressed n
There are n nodes at both ends of the quicklist that are not compressed, and the nodes in the middle are compressed


There is also a fill field, which means the maximum capacity of each quicknode node , different values ​​have different meanings, the default is -2, of course it can also be configured to other values;

##list-max-ziplist-size -2

    When the value is a positive number, it indicates the length of the ziplist on the quicklistNode node. For example, when this value is 5, the ziplist of each quicklistNode node contains at most 5 data items
    When the value is a negative number, Indicates that the length of the ziplist on the quicklistNode node is limited according to the number of bytes. The optional values ​​are -1 to -5.
##ziplist node maximum is 32kb
Value Meaning
-1 ziplist node maximum The maximum number of ziplist nodes is 4kb
##-2 8kb
-3 ziplist node maximum is 16kb
##-4
-5 ##The maximum ziplist node size is 64kb

Why is there configuration provided?

  • #The shorter the ziplist, the more memory fragments will occur, affecting storage efficiency. When a ziplist only stores one element, the quicklist degenerates into a doubly linked list.

  • The longer the ziplist, the more difficult it is to allocate a large continuous memory space for the ziplist. The larger the value, the more small blocks of memory space will be wasted. When the quicklist has only one node and all elements are stored in a ziplist, the quicklist degenerates into a ziplist.

Conclusion

Although we do not fully understand its source code, we can also familiarize ourselves with a design idea of ​​redis through this article. And know how it is optimized step by step. Let's get a general idea of ​​performance.

The above is the detailed content of redis study notes-list principle. For more information, please follow other related articles on the PHP Chinese website!

Statement:
This article is reproduced at:Golang菜鸟. If there is any infringement, please contact admin@php.cn delete