Golang实现snowflake算法（分布式唯一id生成器）

package main

import (
   "errors"
   "fmt"
   "sync"
   "time"
)

const (
   twepoch        = int64(1417937700000) // 默认起始的时间戳 1449473700000 。计算时，减去这个值
   DistrictIdBits = uint(5)              //区域 所占用位置
   NodeIdBits     = uint(9)              //节点 所占位置
   sequenceBits   = uint(10)             //自增ID 所占用位置

   /*
    * 1 符号位  |  39 时间戳                                     | 5 区域  |  9 节点       | 10 （毫秒内）自增ID
    * 0        |  0000000 00000000 00000000 00000000 00000000  | 00000  | 000000 000   |  000000 0000
    *
    */
   maxNodeId     = -1 ^ (-1 << NodeIdBits)     //节点 ID 最大范围
   maxDistrictId = -1 ^ (-1 << DistrictIdBits) //最大区域范围

   nodeIdShift        = sequenceBits //左移次数
   DistrictIdShift    = sequenceBits + NodeIdBits
   timestampLeftShift = sequenceBits + NodeIdBits + DistrictIdBits
   sequenceMask       = -1 ^ (-1 << sequenceBits)
   maxNextIdsNum      = 100 //单次获取ID的最大数量
)

type IdWorker struct {
   sequence      int64 //序号
   lastTimestamp int64 //最后时间戳
   nodeId        int64 //节点ID
   twepoch       int64
   districtId    int64
   mutex         sync.Mutex
}

// NewIdWorker new a snowflake id generator object.
func NewIdWorker(NodeId int64) (*IdWorker, error) {
   var districtId int64
   districtId = 1 //暂时默认给1 ，方便以后扩展
   idWorker := &IdWorker{}
   if NodeId > maxNodeId || NodeId < 0 {
      fmt.Sprintf("NodeId Id can't be greater than %d or less than 0", maxNodeId)
      return nil, errors.New(fmt.Sprintf("NodeId Id: %d error", NodeId))
   }
   if districtId > maxDistrictId || districtId < 0 {
      fmt.Sprintf("District Id can't be greater than %d or less than 0", maxDistrictId)
      return nil, errors.New(fmt.Sprintf("District Id: %d error", districtId))
   }
   idWorker.nodeId = NodeId
   idWorker.districtId = districtId
   idWorker.lastTimestamp = -1
   idWorker.sequence = 0
   idWorker.twepoch = twepoch
   idWorker.mutex = sync.Mutex{}
   fmt.Sprintf("worker starting. timestamp left shift %d, District id bits %d, worker id bits %d, sequence bits %d, workerid %d", timestampLeftShift, DistrictIdBits, NodeIdBits, sequenceBits, NodeId)
   return idWorker, nil
}

// timeGen generate a unix millisecond.
func timeGen() int64 {
   return time.Now().UnixNano() / int64(time.Millisecond)
}

// tilNextMillis spin wait till next millisecond.
func tilNextMillis(lastTimestamp int64) int64 {
   timestamp := timeGen()
   for timestamp <= lastTimestamp {
      timestamp = timeGen()
   }
   return timestamp
}

// NextId get a snowflake id.
func (id *IdWorker) NextId() (int64, error) {
   id.mutex.Lock()
   defer id.mutex.Unlock()
   return id.nextid()
}

// NextIds get snowflake ids.
func (id *IdWorker) NextIds(num int) ([]int64, error) {
   if num > maxNextIdsNum || num < 0 {
      fmt.Sprintf("NextIds num can't be greater than %d or less than 0", maxNextIdsNum)
      return nil, errors.New(fmt.Sprintf("NextIds num: %d error", num))
   }
   ids := make([]int64, num)
   id.mutex.Lock()
   defer id.mutex.Unlock()
   for i := 0; i < num; i++ {
      ids[i], _ = id.nextid()
   }
   return ids, nil
}

func (id *IdWorker) nextid() (int64, error) {
   timestamp := timeGen()
   if timestamp < id.lastTimestamp {
      //    fmt.Sprintf("clock is moving backwards.  Rejecting requests until %d.", id.lastTimestamp)
      return 0, errors.New(fmt.Sprintf("Clock moved backwards.  Refusing to generate id for %d milliseconds", id.lastTimestamp-timestamp))
   }
   if id.lastTimestamp == timestamp {
      id.sequence = (id.sequence + 1) & sequenceMask
      if id.sequence == 0 {
         timestamp = tilNextMillis(id.lastTimestamp)
      }
   } else {
      id.sequence = 0
   }
   id.lastTimestamp = timestamp
   return ((timestamp - id.twepoch) << timestampLeftShift) | (id.districtId << DistrictIdShift) | (id.nodeId << nodeIdShift) | id.sequence, nil
}

有疑问加站长微信联系（非本文作者）