基于Golang如何实现Redis协议解析器-Redis-PHP中文网

首页

数据库

Redis

基于Golang如何实现Redis协议解析器

WBOYWBOYWBOYWBOYWBOYWBOYWBOYWBOYWBOYWBOYWBOYWBOYWB

May 28, 2023 pm 07:13 PM

redisgolang

RESP协议

RESP是客户端与服务端通信的协议，格式有五种：

正常回复：以“+”开头，以“\r\n”结尾的字符串形式

错误回复：以“-”开头，以“\r\n”结尾的字符串形式

整数：以“:”开头，以“\r\n”结尾的字符串形式

多行字符串：以“$”开头，后跟实际发送字节数，再以“\r\n”开头和结尾

$3\r\nabc\r\n

数组：以“*”开头，后跟成员个数

SET key value
*3\r\n$3\r\nSET\r\n$3\r\nkey\r\n$5\r\nvalue\r\n

客户端和服务器发送的命令或数据一律以 \r\n （CRLF）作为换行符。

当我们输入*3\r\n$3\r\nSET\r\n$3\r\nkey\r\n$5\r\nvalue\r\n这样一串命令，服务端接收到的是如下的命令：
*3\r\n
$3\r\n
SET\r\n
$3\r\n
key\r\n
$5\r\n
value\r\n

interface/resp/conn.go

type Connection interface {
   Write([]byte) error
   GetDBIndex() int
   SelectDB(int)
}

interface/resp/reply.go
type Reply interface {
	ToBytes() []byte
}

Connection接口：Redis客户端的一个连接
Write：给客户端回复消息
GetDBIndex：Redis有16个DB
Reply接口：响应接口

resp/reply/consts.go

type PongReply struct{}

var pongBytes = []byte("+PONG\r\n")

func (r *PongReply) ToBytes() []byte {
    return pongBytes
}

var thePongReply = new(PongReply)

func MakePongReply() *PongReply {
    return thePongReply
}

type OkReply struct{}

var okBytes = []byte("+OK\r\n")

func (r *OkReply) ToBytes() []byte {
    return okBytes
}

var theOkReply = new(OkReply)

func MakeOkReply() *OkReply {
    return theOkReply
}

var nullBulkBytes = []byte("$-1\r\n")

type NullBulkReply struct{}

func (r *NullBulkReply) ToBytes() []byte {
    return nullBulkBytes
}

func MakeNullBulkReply() *NullBulkReply {
    return &NullBulkReply{}
}

var emptyMultiBulkBytes = []byte("*0\r\n")

type EmptyMultiBulkReply struct{}

func (r *EmptyMultiBulkReply) ToBytes() []byte {
    return emptyMultiBulkBytes
}

type NoReply struct{}

var noBytes = []byte("")

func (r *NoReply) ToBytes() []byte {
    return noBytes
}

定义五种回复：回复pong，ok，null，空数组，空

resp/reply/reply.go

type ErrorReply interface {
   Error() string
   ToBytes() []byte
}

ErrorReply：定义错误接口

resp/reply/errors.go

type UnknownErrReply struct{}

var unknownErrBytes = []byte("-Err unknown\r\n")

func (r *UnknownErrReply) ToBytes() []byte {
   return unknownErrBytes
}

func (r *UnknownErrReply) Error() string {
   return "Err unknown"
}

type ArgNumErrReply struct {
   Cmd string
}

func (r *ArgNumErrReply) ToBytes() []byte {
   return []byte("-ERR wrong number of arguments for &#39;" + r.Cmd + "&#39; command\r\n")
}

func (r *ArgNumErrReply) Error() string {
   return "ERR wrong number of arguments for &#39;" + r.Cmd + "&#39; command"
}

func MakeArgNumErrReply(cmd string) *ArgNumErrReply {
   return &ArgNumErrReply{
      Cmd: cmd,
   }
}

type SyntaxErrReply struct{}

var syntaxErrBytes = []byte("-Err syntax error\r\n")
var theSyntaxErrReply = &SyntaxErrReply{}

func MakeSyntaxErrReply() *SyntaxErrReply {
   return theSyntaxErrReply
}

func (r *SyntaxErrReply) ToBytes() []byte {
   return syntaxErrBytes
}

func (r *SyntaxErrReply) Error() string {
   return "Err syntax error"
}

type WrongTypeErrReply struct{}

var wrongTypeErrBytes = []byte("-WRONGTYPE Operation against a key holding the wrong kind of value\r\n")

func (r *WrongTypeErrReply) ToBytes() []byte {
   return wrongTypeErrBytes
}

func (r *WrongTypeErrReply) Error() string {
   return "WRONGTYPE Operation against a key holding the wrong kind of value"
}

type ProtocolErrReply struct {
   Msg string
}

func (r *ProtocolErrReply) ToBytes() []byte {
   return []byte("-ERR Protocol error: &#39;" + r.Msg + "&#39;\r\n")
}

func (r *ProtocolErrReply) Error() string {
   return "ERR Protocol error: &#39;" + r.Msg
}

errors定义5种错误：UnknownErrReply 未知错误，ArgNumErrReply 参数个数错误，SyntaxErrReply 语法错误，WrongTypeErrReply 数据类型错误，ProtocolErrReply 协议错误

resp/reply/reply.go

var (
   nullBulkReplyBytes = []byte("$-1")
   // 协议的结尾
   CRLF = "\r\n"
)

type BulkReply struct {
   Arg []byte
}

func MakeBulkReply(arg []byte) *BulkReply {
   return &BulkReply{
      Arg: arg,
   }
}

func (r *BulkReply) ToBytes() []byte {
   if len(r.Arg) == 0 {
      return nullBulkReplyBytes
   }
   return []byte("$" + strconv.Itoa(len(r.Arg)) + CRLF + string(r.Arg) + CRLF)
}

type MultiBulkReply struct {
   Args [][]byte
}

func (r *MultiBulkReply) ToBytes() []byte {
   argLen := len(r.Args)
   var buf bytes.Buffer
   buf.WriteString("*" + strconv.Itoa(argLen) + CRLF)
   for _, arg := range r.Args {
      if arg == nil {
         buf.WriteString("$-1" + CRLF)
      } else {
         buf.WriteString("$" + strconv.Itoa(len(arg)) + CRLF + string(arg) + CRLF)
      }
   }
   return buf.Bytes()
}

func MakeMultiBulkReply(args [][]byte) *MultiBulkReply {
   return &MultiBulkReply{
      Args: args,
   }
}

type StatusReply struct {
   Status string
}

func MakeStatusReply(status string) *StatusReply {
   return &StatusReply{
      Status: status,
   }
}

func (r *StatusReply) ToBytes() []byte {
   return []byte("+" + r.Status + CRLF)
}

type IntReply struct {
   Code int64
}

func MakeIntReply(code int64) *IntReply {
   return &IntReply{
      Code: code,
   }
}

func (r *IntReply) ToBytes() []byte {
   return []byte(":" + strconv.FormatInt(r.Code, 10) + CRLF)
}

type StandardErrReply struct {
   Status string
}

func (r *StandardErrReply) ToBytes() []byte {
   return []byte("-" + r.Status + CRLF)
}

func (r *StandardErrReply) Error() string {
   return r.Status
}

func MakeErrReply(status string) *StandardErrReply {
   return &StandardErrReply{
      Status: status,
   }
}

func IsErrorReply(reply resp.Reply) bool {
   return reply.ToBytes()[0] == &#39;-&#39;
}

BulkReply：回复一个字符串
MultiBulkReply：回复字符串数组
StatusReply：状态回复
IntReply：数字回复
StandardErrReply：标准错误回复
IsErrorReply：判断是否为错误回复
ToBytes：将字符串转成RESP协议规定的格式

resp/parser/parser.go

type Payload struct {
   Data resp.Reply
   Err  error
}

type readState struct {
   readingMultiLine  bool     
   expectedArgsCount int     
   msgType           byte    
   args              [][]byte 
   bulkLen           int64    
}

func (s *readState) finished() bool {
   return s.expectedArgsCount > 0 && len(s.args) == s.expectedArgsCount
}

func ParseStream(reader io.Reader) <-chan *Payload {
   ch := make(chan *Payload)
   go parse0(reader, ch)
   return ch
}

func parse0(reader io.Reader, ch chan<- *Payload) {
	 ......
}

Payload结构体：客服端给我们发的数据

Reply：客户端与服务端互相发的数据都称为Reply

readState结构体：

readingMultiLine：解析单行还是多行数据
expectedArgsCount：应该读取的参数个数
msgType：消息类型
args：消息内容
bulkLen：数据长度

finished方法：判断解析是否完成

ParseStream方法：异步解析数据后放入管道，返回管道数据

func readLine(bufReader *bufio.Reader, state *readState) ([]byte, bool, error) {
   var msg []byte
   var err error
   if state.bulkLen == 0 {
      msg, err = bufReader.ReadBytes(&#39;\n&#39;)
      if err != nil {
         return nil, true, err
      }
      if len(msg) == 0 || msg[len(msg)-2] != &#39;\r&#39; {
         return nil, false, errors.New("protocol error: " + string(msg))
      }
   } else {
      msg = make([]byte, state.bulkLen+2)
      _, err = io.ReadFull(bufReader, msg)
      if err != nil {
         return nil, true, err
      }
      if len(msg) == 0 || msg[len(msg)-2] != &#39;\r&#39; || msg[len(msg)-1] != &#39;\n&#39; {
         return nil, false, errors.New("protocol error: " + string(msg))
      }
      state.bulkLen = 0
   }
   return msg, false, nil
}

readLine：一行一行的读取。读正常的行，以\n分隔。读正文中包含\r\n字符的行时，state.bulkLen加上换行符\r\n（state.bulkLen+2）

func parseMultiBulkHeader(msg []byte, state *readState) error {
   var err error
   var expectedLine uint64
   expectedLine, err = strconv.ParseUint(string(msg[1:len(msg)-2]), 10, 32)
   if err != nil {
      return errors.New("protocol error: " + string(msg))
   }
   if expectedLine == 0 {
      state.expectedArgsCount = 0
      return nil
   } else if expectedLine > 0 {
      state.msgType = msg[0]
      state.readingMultiLine = true
      state.expectedArgsCount = int(expectedLine)
      state.args = make([][]byte, 0, expectedLine)
      return nil
   } else {
      return errors.New("protocol error: " + string(msg))
   }
}

func parseBulkHeader(msg []byte, state *readState) error {
   var err error
   state.bulkLen, err = strconv.ParseInt(string(msg[1:len(msg)-2]), 10, 64)
   if err != nil {
      return errors.New("protocol error: " + string(msg))
   }
   if state.bulkLen == -1 { // null bulk
      return nil
   } else if state.bulkLen > 0 {
      state.msgType = msg[0]
      state.readingMultiLine = true
      state.expectedArgsCount = 1
      state.args = make([][]byte, 0, 1)
      return nil
   } else {
      return errors.New("protocol error: " + string(msg))
   }
}

parseMultiBulkHeader：解析数组的头部，设置期望的行数和相关参数。

parseBulkHeader：解析多行字符串的头部。

func parseSingleLineReply(msg []byte) (resp.Reply, error) {
   str := strings.TrimSuffix(string(msg), "\r\n")
   var result resp.Reply
   switch msg[0] {
   case &#39;+&#39;: // status reply
      result = reply.MakeStatusReply(str[1:])
   case &#39;-&#39;: // err reply
      result = reply.MakeErrReply(str[1:])
   case &#39;:&#39;: // int reply
      val, err := strconv.ParseInt(str[1:], 10, 64)
      if err != nil {
         return nil, errors.New("protocol error: " + string(msg))
      }
      result = reply.MakeIntReply(val)
   }
   return result, nil
}

func readBody(msg []byte, state *readState) error {
   line := msg[0 : len(msg)-2]
   var err error
   if line[0] == &#39;$&#39; {
      // bulk reply
      state.bulkLen, err = strconv.ParseInt(string(line[1:]), 10, 64)
      if err != nil {
         return errors.New("protocol error: " + string(msg))
      }
      if state.bulkLen <= 0 { // null bulk in multi bulks
         state.args = append(state.args, []byte{})
         state.bulkLen = 0
      }
   } else {
      state.args = append(state.args, line)
   }
   return nil
}

parseSingleLineReply：解析单行命令

readBody：读取多行的命令，如果是$开头，设置bulkLen，读取下一行时根据这个+2，不是$开头则直接添加到args

func parse0(reader io.Reader, ch chan<- *Payload) {
    defer func() {
       if err := recover(); err != nil {
          logger.Error(string(debug.Stack()))
      }
   }()
    bufReader := bufio.NewReader(reader)
    var state readState
    var err error
    var msg []byte
    for {
       var ioErr bool
       msg, ioErr, err = readLine(bufReader, &state)
       if err != nil {
          if ioErr {
             ch <- &Payload{
                Err: err,
            }
             close(ch)
             return
         }
          ch <- &Payload{
             Err: err,
         }
          state = readState{}
          continue
      }

       if !state.readingMultiLine {
          if msg[0] == &#39;*&#39; {
             // multi bulk reply
             err = parseMultiBulkHeader(msg, &state)
             if err != nil {
                ch <- &Payload{
                   Err: errors.New("protocol error: " + string(msg)),
               }
                state = readState{}
                continue
            }
             if state.expectedArgsCount == 0 {
                ch <- &Payload{
                   Data: &reply.EmptyMultiBulkReply{},
               }
                state = readState{}
                continue
            }
         } else if msg[0] == &#39;$&#39; { // bulk reply
             err = parseBulkHeader(msg, &state)
             if err != nil {
                ch <- &Payload{
                   Err: errors.New("protocol error: " + string(msg)),
               }
                state = readState{} // reset state
                continue
            }
             if state.bulkLen == -1 { // null bulk reply
                ch <- &Payload{
                   Data: &reply.NullBulkReply{},
               }
                state = readState{} // reset state
                continue
            }
         } else {
             // single line reply
             result, err := parseSingleLineReply(msg)
             ch <- &Payload{
                Data: result,
                Err:  err,
            }
             state = readState{} // reset state
             continue
         }
      } else {
          // read bulk reply
          err = readBody(msg, &state)
          if err != nil {
             ch <- &Payload{
                Err: errors.New("protocol error: " + string(msg)),
            }
             state = readState{} // reset state
             continue
         }
          // if sending finished
          if state.finished() {
             var result resp.Reply
             if state.msgType == &#39;*&#39; {
                result = reply.MakeMultiBulkReply(state.args)
            } else if state.msgType == &#39;$&#39; {
                result = reply.MakeBulkReply(state.args[0])
            }
             ch <- &Payload{
                Data: result,
                Err:  err,
            }
             state = readState{}
         }
      }
   }
}

parse0：解析指令，解析完成后通过channel发出去

resp/connection/conn.go

type Connection struct {
    conn net.Conn
    waitingReply wait.Wait
    mu sync.Mutex // 避免多个协程往客户端中写
    selectedDB int
}

func NewConn(conn net.Conn) *Connection {
    return &Connection{
        conn: conn,
    }
}

func (c *Connection) RemoteAddr() net.Addr {
    return c.conn.RemoteAddr()
}

func (c *Connection) Close() error {
    c.waitingReply.WaitWithTimeout(10 * time.Second)
    _ = c.conn.Close()
    return nil
}

func (c *Connection) Write(b []byte) error {
    if len(b) == 0 {
        return nil
    }
    c.mu.Lock()
    c.waitingReply.Add(1)
    defer func() {
        c.waitingReply.Done()
        c.mu.Unlock()
    }()

    _, err := c.conn.Write(b)
    return err
}

func (c *Connection) GetDBIndex() int {
    return c.selectedDB
}

func (c *Connection) SelectDB(dbNum int) {
    c.selectedDB = dbNum
}

我们之前编写的EchoHandler用于接收用户的输入并将其原封不动地返回。现在需要编写一个RespHandler来替换EchoHandler，由解析器来进行解析。一个管理客户端连接的结构体Connection需要存在于RespHandler中。

Connection：客户端连接，在协议层的handler中会用到

resp/handler/handler.go

var (
   unknownErrReplyBytes = []byte("-ERR unknown\r\n")
)

type RespHandler struct {
   activeConn sync.Map
   db         databaseface.Database
   closing    atomic.Boolean
}

func MakeHandler() *RespHandler {
   var db databaseface.Database
   db = database.NewEchoDatabase()
   return &RespHandler{
      db: db,
   }
}

func (h *RespHandler) closeClient(client *connection.Connection) {
   _ = client.Close()
   h.db.AfterClientClose(client)
   h.activeConn.Delete(client)
}

func (h *RespHandler) Handle(ctx context.Context, conn net.Conn) {
   if h.closing.Get() {
      // closing handler refuse new connection
      _ = conn.Close()
   }

   client := connection.NewConn(conn)
   h.activeConn.Store(client, 1)

   ch := parser.ParseStream(conn)
   for payload := range ch {
      if payload.Err != nil {
         if payload.Err == io.EOF ||
            payload.Err == io.ErrUnexpectedEOF ||
            strings.Contains(payload.Err.Error(), "use of closed network connection") {
            // connection closed
            h.closeClient(client)
            logger.Info("connection closed: " + client.RemoteAddr().String())
            return
         }
         // protocol err
         errReply := reply.MakeErrReply(payload.Err.Error())
         err := client.Write(errReply.ToBytes())
         if err != nil {
            h.closeClient(client)
            logger.Info("connection closed: " + client.RemoteAddr().String())
            return
         }
         continue
      }
      if payload.Data == nil {
         logger.Error("empty payload")
         continue
      }
      r, ok := payload.Data.(*reply.MultiBulkReply)
      if !ok {
         logger.Error("require multi bulk reply")
         continue
      }
      result := h.db.Exec(client, r.Args)
      if result != nil {
         _ = client.Write(result.ToBytes())
      } else {
         _ = client.Write(unknownErrReplyBytes)
      }
   }
}

func (h *RespHandler) Close() error {
   logger.Info("handler shutting down...")
   h.closing.Set(true)
   // TODO: concurrent wait
   h.activeConn.Range(func(key interface{}, val interface{}) bool {
      client := key.(*connection.Connection)
      _ = client.Close()
      return true
   })
   h.db.Close()
   return nil
}

RespHandler：和之前的echo类似，加了核心层的db.exec执行解析的指令

interface/database/database.go

type CmdLine = [][]byte

type Database interface {
	Exec(client resp.Connection, args [][]byte) resp.Reply
	AfterClientClose(c resp.Connection)
	Close()
}

type DataEntity struct {
	Data interface{}
}

Exec：核心层的执行

AfterClientClose：关闭之后的善后方法

CmdLine：二维字节数组的指令别名

DataEntity：表示Redis的数据，包括string, list, set等等

database/echo_database.go

type EchoDatabase struct {
}

func NewEchoDatabase() *EchoDatabase {
   return &EchoDatabase{}
}

func (e EchoDatabase) Exec(client resp.Connection, args [][]byte) resp.Reply {
   return reply.MakeMultiBulkReply(args)
}

func (e EchoDatabase) AfterClientClose(c resp.Connection) {
   logger.Info("EchoDatabase AfterClientClose")
}

func (e EchoDatabase) Close() {
   logger.Info("EchoDatabase Close")
}

echo_database：测试协议层

Exec：指令解析后，再使用MakeMultiBulkReply包装一下返回去

main.go

err := tcp.ListenAndServeWithSignal(
   &tcp.Config{
      Address: fmt.Sprintf("%s:%d",
         config.Properties.Bind,
         config.Properties.Port),
   },
   handler.MakeHandler())
if err != nil {
   logger.Error(err)
}

main改成刚才写的：handler.MakeHandler()

以上是基于Golang如何实现Redis协议解析器的详细内容。更多信息请关注PHP中文网其他相关文章！

声明

本文转载于：亿速云。如有侵权，请联系admin@php.cn删除

REDIS：NOSQL方法的优势Apr 27, 2025 am 12:09 AM

Redis是NoSQL数据库，提供高性能和灵活性。1)通过键值对存储数据，适合处理大规模数据和高并发。2)内存存储和单线程模型确保快速读写和原子性。3)使用RDB和AOF机制进行数据持久化，支持高可用性和横向扩展。

REDIS：了解其架构和目的Apr 26, 2025 am 12:11 AM

Redis是一种内存数据结构存储系统，主要用作数据库、缓存和消息代理。它的核心特点包括单线程模型、I/O多路复用、持久化机制、复制与集群功能。 Redis在实际应用中常用于缓存、会话存储和消息队列，通过选择合适的数据结构、使用管道和事务、以及进行监控和调优，可以显着提升其性能。

REDIS与SQL数据库：关键差异Apr 25, 2025 am 12:02 AM

Redis和SQL数据库的主要区别在于：Redis是内存数据库，适用于高性能和灵活性需求；SQL数据库是关系型数据库，适用于复杂查询和数据一致性需求。具体来说，1)Redis提供高速数据访问和缓存服务，支持多种数据类型，适用于缓存和实时数据处理；2)SQL数据库通过表格结构管理数据，支持复杂查询和事务处理，适用于电商和金融系统等需要数据一致性的场景。

REDIS：它如何充当数据存储和服务Apr 24, 2025 am 12:08 AM

REDISACTSASBOTHADATASTOREANDASERVICE.1）ASADATASTORE，ITUSESIN-MEMORYSTOOGATOFORFOFFASTESITION，支持VariousDatharptructuresLikeKey-valuepairsandsortedsetsetsetsetsetsetsets.2）asaservice，ItprovidespunctionslikeItionitionslikepunikeLikePublikePublikePlikePlikePlikeAndluikeAndluAascriptingiationsmpleplepleclexplectiations

REDIS与其他数据库：比较分析Apr 23, 2025 am 12:16 AM

Redis与其他数据库相比，具有以下独特优势：1)速度极快，读写操作通常在微秒级别；2)支持丰富的数据结构和操作；3)灵活的使用场景，如缓存、计数器和发布订阅。选择Redis还是其他数据库需根据具体需求和场景，Redis在高性能、低延迟应用中表现出色。

REDIS的角色：探索数据存储和管理功能Apr 22, 2025 am 12:10 AM

Redis在数据存储和管理中扮演着关键角色，通过其多种数据结构和持久化机制成为现代应用的核心。1）Redis支持字符串、列表、集合、有序集合和哈希表等数据结构，适用于缓存和复杂业务逻辑。2）通过RDB和AOF两种持久化方式，Redis确保数据的可靠存储和快速恢复。

REDIS：了解NOSQL概念Apr 21, 2025 am 12:04 AM

Redis是一种NoSQL数据库，适用于大规模数据的高效存储和访问。1.Redis是开源的内存数据结构存储系统，支持多种数据结构。2.它提供极快的读写速度，适合缓存、会话管理等。3.Redis支持持久化，通过RDB和AOF方式确保数据安全。4.使用示例包括基本的键值对操作和高级的集合去重功能。5.常见错误包括连接问题、数据类型不匹配和内存溢出，需注意调试。6.性能优化建议包括选择合适的数据结构和设置内存淘汰策略。