玩一把redis源码(一):为redis添加自己的列表类型
2019年第一篇文档,为2019年做个良好的开端,本文档通过step by step的方式向读者展示如何为redis添加一个数据类型,阅读本文档后读者对redis源码的执行逻辑会有比较清晰的认识,并且可以深入理解redis 源码中关于链表数据结构的使用,写这篇文档作者获益良多,阅读系统软件源码的兴趣也大大提高。
同时也再次感受到良好的基础是深入学习的前提。特别强调本文档仅用于学习,并非是要修改redis源码。 建议读者阅读本文档时实际动手敲一下代码,然后翻阅下redis源码,debug下redis
本文档分为三大部分:
环境介绍与效果演示
redis接收命令到返回数据的执行逻辑
代码实现
文档的重点和难点在第三部分,完全阅读本文档需要读者具备基本的c语言和数据结构知识。
环境介绍和效果演示
环境介绍
redis版本为5.0.3 64 bit
操作系统版本为Ubuntu 18.10 64bit
源码可以用gedit查看 gdb调试
ide 可以用eclipse+CDT
效果演示
本案例实现了一个链表,对应redis的list数据类型,对链表的操作实现了插入、设置某个节点的值、新建节点、获取一定范围内的节点、获取列表长度等.下面表格列出具体实现的命令与对应的redis原生命令
实现命令 redis原生命令 命令含义
myrpush rpush 从尾部插入链表
mylrange lrange 获取范围内值
myrpop rpop 从右侧弹出值
myllen llen 获取list长度
mylset lset 设置某个节点值
myinsert linsert 在指定位置插入值
mylindex lindex 获取指定位置值
实现命令演示:
127.0.0.1:6379> myrpush mygjw gjw0(integer)
1127.0.0.1:6379> myrpush mygjw gjw1
(integer) 2
127.0.0.1:6379> myrpush mygjw gjw2
(integer) 3
127.0.0.1:6379> mylrange 0 -1
(error) ERR wrong number of arguments for 'mylrange' command
127.0.0.1:6379> mylrange mygjw 0 -1
1) "gjw0"
2) "gjw1"
3) "gjw2"
127.0.0.1:6379> myrpop mygjw
"gjw2"
127.0.0.1:6379> mylrange mygjw 0 -1
1) "gjw0"
2) "gjw1"
127.0.0.1:6379> myllen mygjw
(integer) 2
127.0.0.1:6379> mylset mygjw 0 gjw00
OK127.0.0.1:6379> mylset mygjw 1 gjw01
OK127.0.0.1:6379> mylrange mygjw 0 -1
1) "gjw00"
2) "gjw01"
127.0.0.1:6379> mylinsert mygjw 0 gjw0
(integer) 3
127.0.0.1:6379> mylinsert mygjw 1 gjw1
(integer) 4
127.0.0.1:6379> mylrange mygjw 0 -1
1) "gjw0"
2) "gjw1"
3) "gjw00"
4) "gjw01"
127.0.0.1:6379> mylindex mygjw 0
"gjw0"
127.0.0.1:6379> mylindex mygjw 1
"gjw1"
redis接收命令到返回数据的执行逻辑
此部分只选择与本文档相关部分做简单介绍,实际命令执行涉及到保存文件、主备同步、集群分发等等会复杂很多。下面从server.c文件的processCommand函数开始。
640?wx_fmt=png
下面以myrpush命令执行的时序图来说明具体调用的函数及其所在的文件,其他命令执行方式都类似.图中方框中字符串是文件名,所有关于list类型的命令处理函数都在文件t_list.c中,其他类型的命令处理函数需要找找对应的源文件,redis命令非常规范,找起来不难。
640?wx_fmt=png
代码实现与redis源码阅读
代码实现
1.添加自己的源文件和头文件
mylinkedlist.h、mylinkedlist.h
mylist类型采用双向链表数据结构,数据域采用了哑型指针,方便插入任何类型数据,为了方便代码仅仅实现了char型,node 结构体有个size属性用来存储数据域占用的字节数。
此部分比redis源码实现简单很多,但是如果能看懂这部分代码,也应该可以比较容易的看懂redis源码。redis中list类型数据域是ziplist,默认情况下数据域最大8k字节,ziplist有自己的数据结构,这里列出redis 列表类型处理的相关源文件:quicklist.h、quicklist.c、ziplist.h、ziplist.c
头文件内容如下:
/*
* mylineklist.h
*
* Created on: Dec 29, 2018
* Author: gjw
*/typedef void * ADT;
typedef const void * CADT;
typedef int (*LIDESTROY) (ADT e);
typedef void (*LITRVAVEL) (ADT e);
struct NODE;
typedef struct NODE * PNODE;
typedef struct MylinkedListS MylinkedList;
typedef MylinkedList * MYLINKEDLIST;
MYLINKEDLIST MyCreateList();
void MyAppend(MYLINKEDLIST list,ADT data,int len);
int MyInsert(MYLINKEDLIST list,ADT data,unsigned int pos,int len);
void MyDelete(MYLINKEDLIST list,unsigned int pos,LIDESTROY destroy);
ADT Myrpop(MYLINKEDLIST list,int * sz);
int Mylset(MYLINKEDLIST list, ADT data, unsigned int pos,int len);
void MyClear(MYLINKEDLIST list,LIDESTROY destroy);
int MyLength(MYLINKEDLIST list);
PNODE MyIndex(MYLINKEDLIST list,unsigned int index);
void MyTraverse(MYLINKEDLIST list,LITRVAVEL litravel);
int MyIsEmpty(MYLINKEDLIST list);
PNODE MyNext(PNODE node);
void getData(PNODE n,char ** data,int * sz);
在本例中并非所有的声明函数都已实现,仅仅实现了上面演示的几个命令,源文件内容如下:
/*
* mylinkedlist.c
*
* Created on: Dec 29, 2018
* Author: gjw
*/
#include <stdlib.h>
#include <stdio.h>
#include <string.h>
#include "mylinkedlist.h"
struct NODE {
ADT data;
int size;
PNODE next;
PNODE previous;
} ;
typedef struct NODE NODE;
struct MylinkedListS {
unsigned int len;
PNODE head, tail;
};
MylinkedList* MyCreateList() {
MylinkedList* list = (MylinkedList*) malloc(sizeof(MylinkedList));
PNODE head = (NODE *) malloc(sizeof(NODE));
PNODE tail = (NODE *) malloc(sizeof(NODE));
list->head = head;
list->tail = tail;
list->tail->next=NULL;
list->tail->previous=NULL;
list->head->next=NULL;
list->head->previous=NULL;
list->len=0;
printf("init a list");
return list;
}
void MyAppend(MYLINKEDLIST list, ADT data,int len) {
if (list->head->next == NULL) {
MyInsert(list, data, 0,len);
return;
}
printf("->");
PNODE ctail = (NODE *) malloc(sizeof(NODE));
ctail->data=(char *)malloc(len);
memcpy(ctail->data,data,len);
ctail->size=len;
ctail->next=NULL;
ctail->previous=list->tail->next;
list->tail->next->next = ctail;
list->tail->next = ctail;
list->len = list->len + 1;
}
int MyInsert(MYLINKEDLIST list, ADT data, unsigned int pos,int len) {
if ( pos > list->len) return 0;
printf("insert success!\n");
PNODE node = list->head;
for (unsigned int i = 0; i < pos; i++) {
node = node->next;
}
PNODE newNode = (NODE *) malloc(sizeof(NODE));
newNode->data=(char *)malloc(len);
memcpy(newNode->data,data,len);
newNode->size=len;
newNode->next = node->next;
newNode->previous=node; if(node->next){
newNode->next->previous=newNode;
}
node->next = newNode;
list->tail->next=newNode;
list->len = list->len + 1;
return 1;
}
ADT Myrpop(MYLINKEDLIST list,int * sz){
PNODE tail=list->tail->next;
if(list->len==0){
return NULL;
}
ADT data=tail->data;
*sz=tail->size; list->tail->next=tail->previous;
tail->previous->next=NULL;
list->len=list->len-1;
free(tail);
return data;
}int MyLength(MYLINKEDLIST list) {
return list->len;
}
int MyIsEmpty(MYLINKEDLIST list) {
return !list->len;
}
void getData(PNODE n,char ** data,int * sz) {
*data=n->data;
*sz=n->size;
}
PNODE MyNext(PNODE node) {
return node->next;
}
PNODE MyIndex(MYLINKEDLIST list,unsigned int index){
PNODE n=list->head->next;
if(index>=list->len){
return NULL;
}
for(int i=0;i<index;i++){
n=n->next;
}
return n;
}
int Mylset(MYLINKEDLIST list, ADT data, unsigned int pos,int len) {
PNODE node=list->head->next;
if(list->len==0){
MyInsert(list,data,pos,len);
return 1;
}
if(pos>=list->len){
return 0;
} for(int i=0;i<pos;i++){
node=node->next;
}
realloc(node->data,len);
node->size=len;
memcpy(node->data,data,len);
return 1;
}
2.修改server.h主要是声明自己的类型,添加自定义类型的命令处理函数,加入自定义头文件,添加以下内容 :
server.h
#include "mylinkedlist.h" /* mylist ,guojiagnwei add */
#define MYOBJ_LIST 11 /* MYList object. */
void myrpushCommand(client *c);
void mylrangeCommand(client *c);
void myrpopCommand(client *c);
void myllenCommand(client *c);
void mylsetCommand(client *c);
void mylinsertCommand(client *c);
void mylindexCommand(client *c);
robj *createMylistObject(void);
3.修改object.c 此文件主要用来操作redis object,添加以下内容
robj *createMylistObject(void) {
MylinkedList* l = MyCreateList(); //quicklist *l = quicklistCreate();
robj *o = createObject(MYOBJ_LIST,l);
o->encoding = OBJ_ENCODING_QUICKLIST;
return o;
}
4.修改 t_list.c文件,实现server.h中声明的命令处理函数
//added by guojiangweivoid myrpushCommand(client *c){
int j, pushed = 0;
robj *lobj = lookupKeyWrite(c->db,c->argv[1]);
for (j = 2; j < c->argc; j++) {
if (!lobj) {
lobj = createMylistObject();
dbAdd(c->db,c->argv[1],lobj);
}
MyAppend(lobj->ptr,c->argv[j]->ptr,sdslen(c->argv[j]->ptr));
pushed++;
}
addReplyLongLong(c, MyLength(lobj->ptr)); if (pushed) {
char *event = "rpush";
signalModifiedKey(c->db,c->argv[1]);
notifyKeyspaceEvent(1,event,c->argv[1],c->db->id);
}
server.dirty += pushed;
}
void mylrangeCommand(client *c){
robj *o;
long start, end, llen, rangelen;
char * data;
int sz;
if ((getLongFromObjectOrReply(c, c->argv[2], &start, NULL) != C_OK) ||
(getLongFromObjectOrReply(c, c->argv[3], &end, NULL) != C_OK)) return;
if ((o = lookupKeyReadOrReply(c,c->argv[1],shared.emptymultibulk)) == NULL
|| checkType(c,o,MYOBJ_LIST)) return;
llen = MyLength(o->ptr); /* convert negative indexes */
if (start < 0) start = llen+start;
if (end < 0) end = llen+end;
if (start < 0) start = 0;
/* Invariant: start >= 0, so this test will be true when end < 0.
* The range is empty when start > end or start >= length. */
if (start > end || start >= llen) {
addReply(c,shared.emptymultibulk);
return;
} if (end >= llen) end = llen-1;
rangelen = (end-start)+1; /* Return the result in form of a multi-bulk reply */
addReplyMultiBulkLen(c,rangelen);
if (o->encoding == OBJ_ENCODING_QUICKLIST) {
PNODE n = MyIndex(o->ptr, start);
while(rangelen--) {
getData( n,&data,&sz);
addReplyBulkCBuffer(c,data,sz);
n=MyNext(n);
}
} else {
serverPanic("List encoding is not QUICKLIST!");
}
}
void myrpopCommand(client *c){
robj *o = lookupKeyWriteOrReply(c,c->argv[1],shared.nullbulk);
int sz;
if (o == NULL || checkType(c,o,MYOBJ_LIST)) return;
char * data=Myrpop(o->ptr,&sz);
robj *value = createStringObject(data,sz);
free(data); if (value == NULL) {
addReply(c,shared.nullbulk);
} else {
char *event = "rpop";
addReplyBulk(c,value);
decrRefCount(value);
notifyKeyspaceEvent(NOTIFY_LIST,event,c->argv[1],c->db->id);
if (listTypeLength(o) == 0) {
notifyKeyspaceEvent(NOTIFY_GENERIC,"del", c->argv[1],c->db->id);
dbDelete(c->db,c->argv[1]);
} signalModifiedKey(c->db,c->argv[1]);
server.dirty++;
}
}
void myllenCommand(client *c){
robj *o = lookupKeyReadOrReply(c,c->argv[1],shared.czero);
if (o == NULL || checkType(c,o,MYOBJ_LIST)) return;
addReplyLongLong(c,MyLength(o->ptr));
}
void mylsetCommand(client *c){
robj *o = lookupKeyWriteOrReply(c,c->argv[1],shared.nokeyerr);
if (o == NULL || checkType(c,o,MYOBJ_LIST)) return;
long index;
robj *value = c->argv[3];
if ((getLongFromObjectOrReply(c, c->argv[2], &index, NULL) != C_OK))
return;
if (o->encoding == OBJ_ENCODING_QUICKLIST) {
MYLINKEDLIST ql = o->ptr;
int replaced = Mylset(ql,value->ptr, index,sdslen(value->ptr));
if (!replaced) {
addReply(c,shared.outofrangeerr);
} else {
addReply(c,shared.ok);
signalModifiedKey(c->db,c->argv[1]);
notifyKeyspaceEvent(NOTIFY_LIST,"lset",c->argv[1],c->db->id);
server.dirty++;
}
} else {
serverPanic("Unknown list encoding");
}
}
void mylinsertCommand(client *c){
long index;
robj *subject;
int inserted = 0;
if ((getLongFromObjectOrReply(c, c->argv[2], &index, NULL) != C_OK)){
addReply(c,shared.syntaxerr);
return;
}
if ((subject = lookupKeyWriteOrReply(c,c->argv[1],shared.czero)) == NULL ||
checkType(c,subject,MYOBJ_LIST)) return;
inserted=MyInsert(subject->ptr,c->argv[3]->ptr,index,sdslen(c->argv[3]->ptr));
if (inserted) {
signalModifiedKey(c->db,c->argv[1]);
notifyKeyspaceEvent(NOTIFY_LIST,"linsert", c->argv[1],c->db->id);
server.dirty++;
} else { /* Notify client of a failed insert */
addReply(c,shared.cnegone);
return;
} addReplyLongLong(c,MyLength(subject->ptr));
}
void mylindexCommand(client *c){
long index;
int size;
robj *o = lookupKeyWriteOrReply(c,c->argv[1],shared.nokeyerr);
if (o == NULL || checkType(c,o,MYOBJ_LIST)) return;
if ((getLongFromObjectOrReply(c, c->argv[2], &index, NULL) != C_OK))
return;
PNODE n=MyIndex(o->ptr,index);
if(n==NULL){
addReply(c,shared.nullbulk);
return;
}
char *data;
getData( n,&data,&size);
if (o->encoding == OBJ_ENCODING_QUICKLIST) {
if (data) {
robj *value=createStringObject(data,size);
addReplyBulk(c,value);
decrRefCount(value);
} else {
addReply(c,shared.nullbulk);
}
} else {
serverPanic("Unknown list encoding");
}
}
5.修改server.c 文件,在变量struct redisCommand redisCommandTable[] 中添加自定义类型命令字符串与命令处理函数映射
{"myrpush",myrpushCommand,-3,"wmF",0,NULL,1,1,1,0,0},
{"mylrange",mylrangeCommand,4,"wmF",0,NULL,1,1,1,0,0},
{"myrpop",myrpopCommand,2,"wmF",0,NULL,1,1,1,0,0},
{"myllen",myllenCommand,2,"wmF",0,NULL,1,1,1,0,0},
{"mylset",mylsetCommand,4,"wmF",0,NULL,1,1,1,0,0},
{"mylinsert",mylinsertCommand,4,"wmF",0,NULL,1,1,1,0,0},
{"mylindex",mylindexCommand,3,"r",0,NULL,1,1,1,0,0},
另外还需要编辑下make文件,比较简单这里不单独列出。
redis源码阅读
代码实现部分列出了需要修改的文件和修改的内容,redis源码设计的非常好,命名非常规范,如果读者按照文档所列进行了操作,对redis 源码结构和程序逻辑应该有了一个大致的认识,作者本来打算详细写下redis list类型的源码实现,但是想了想,水平有限,很难做到严谨与通俗,推荐大家看看《redis 设计与实现》这本书吧。个人觉得看源码最主要的是自己修改和不断的debug。