雪花id生成算法

什么是雪花算法

雪花算法的本质为生成一个64位长度的具有自增性的分布式全局唯一id。在64bits中，会对不同段的位进行划分。可分为：

1. 符号段
2. 时间戳段
3. 机器码段(data center + worker)
4. 自增序列号段

位段详解

1. 第一位 : 符号位，正数为0。
2. [2, 42] : 41位时间戳位，表明id的生成时间点(完整时间戳: 起始时间戳 + 41位时间戳)。41位最多能表示的时间为： (2^41-1) / (1000 * 60 * 60 * 24 * 365) 约等为69.73年。
3. [43, 47] : 5位data center id。data center id + worker id 共10位，最多能表示1024个机器。不同机器保证机器码段的位值不同即可。
4. [48, 52] : 5位worker id。data center id + worker id 共10位，最多能表示1024个机器。不同机器保证机器码段的位值不同即可。
5. [53, 64] : 12位自增序列号，用于区分同一毫秒内生成的id。序列号范围: [0, 2^12-1]，最多有2^12个，即4096个。

优点

1. 算法简单，基于内存，生成效率高
2. 支持分布式环境下的多节点服务（机器码段），秒内可生成百万个唯一id
3. 基于时间戳 与 同时间戳下自增序列号，生成的id具有自增性
4. 具有业务定制性，根据业务的不同可以对不同段的位数进行变更。比如业务持续时长不会那么久，就可以将时间戳段减少位数，补充给自增序列段，使每一毫秒能生成更多的id。

问题

依赖服务器时间。若服务器时钟回拨，可能会导致生成的id重复。可在代码中新增lastTimeMillis字段，在获取nextId时根据系统当前时间进行判断解决。
但若不进行持久化处理，服务重启后发生时钟回拨依旧会出现重复问题。

实际应用

1. mybatis plus：使用雪花算法生成id：@TableId(value = “id”, type = IdType.ID_WORKER)。id字段若不指定类型，默认使用雪花算法生成id
2. Hutool工具包：IdUtil.createSnowflake(workerId, datacenterId);

具体实现

/**
 * Created by QQ.Cong on 2022-07-22 / 9:48
 *
 * @author: CongQingquan
 * @Description: Snowflake util
 */
public class SnowflakeUtils {

    // ============================== Basic field ==============================//

    // Datacenter id
    private long datacenterId;

    // Worker id
    private long workerId;

    // Increment sequence
    private long sequence;

    // ============================== Bits ==============================//

    // Bits of datacenter id
    private long datacenterIdBits;

    // Bits of worker id
    private long workerIdBits;

    // Bits of sequence
    private long sequenceBits;

    // ============================== Largest ==============================//

    // Largest datacenter id
    private long largestDatacenterId;

    // Largest worker id
    private long largestWorkerId;

    // Largest sequence
    private long largestSequence;

    // ============================== Shift ==============================//

    // Left shift num of worker id
    private long workerIdShift;

    // Left shift num of datacenter id
    private long datacenterIdShift;

    // Left shift num of timestamp
    private long timestampShift;

    // ============================== Other ==============================//

    // Epoch
    private long epoch;

    // The timestamp that last get snowflake id
    private long lastTimestamp;

    // ============================== End ==============================//

    public SnowflakeUtils(long dataCenterId, long workerId) {
        // Default epoch: 2022-07-22 00:00:00
        this(1658419200000L, -1L, dataCenterId, workerId, 5L, 5L, 5L);
    }

    public SnowflakeUtils(long epoch, long lastTimestamp, long datacenterId, long workerId,
        long datacenterIdBits, long workerIdBits, long sequenceBits) {
        this.epoch = epoch;
        this.lastTimestamp = lastTimestamp;
        this.datacenterId = datacenterId;
        this.workerId = workerId;
        this.sequence = 0L;
        this.datacenterIdBits = datacenterIdBits;
        this.workerIdBits = workerIdBits;
        this.sequenceBits = sequenceBits;
        this.largestDatacenterId = ~(-1L << datacenterIdBits);
        this.largestWorkerId = ~(-1L << workerIdBits);
        this.largestSequence = ~(-1L << sequenceBits);
        if (datacenterId > largestDatacenterId || datacenterId < 0) {
            throw new IllegalArgumentException(
                String.format("The datacenter id param can't be greater than %s or less than 0",
                    largestDatacenterId));
        }
        if (workerId > largestWorkerId || workerId < 0) {
            throw new IllegalArgumentException(
                String.format("The worker id param can't be greater than %s or less than 0",
                    largestWorkerId));
        }
        this.workerIdShift = sequenceBits;
        this.datacenterIdShift = workerIdShift + workerIdBits;
        this.timestampShift = datacenterIdShift + datacenterIdBits;
    }

    /**
     * Get snowflake id
     * @return
     */
    public synchronized long nextId() {
        long timestamp = System.currentTimeMillis();
        // 若时钟回退
        if (timestamp < lastTimestamp) {
            throw new RuntimeException(
                "System clock moved backward, cannot to generate snowflake id");
        }
        // 若当前毫秒内多次生成雪花id
        if (timestamp == lastTimestamp) {
            sequence = (sequence + 1) & largestSequence;
            // 序列溢出
            if (sequence == 0) {
                timestamp = waitUntilNextMilli(timestamp);
            }
        }
        // 若当前毫秒内首次生成雪花id
        else {
            sequence = 0L;
        }
        // 更新获取雪花id的时间戳
        lastTimestamp = timestamp;
        // 生成雪花id (通过位或运算符进行拼接)
        return ((timestamp - epoch) << timestampShift) // 时间戳段
            | (datacenterId << datacenterIdShift) // 机器码段
            | (workerId << workerIdShift) // 机器码段
            | sequence; // 自增序列段
    }

    /**
     * Wait until next millisecond
     * @param lastTimestamp
     * @return
     */
    private long waitUntilNextMilli(long lastTimestamp) {
        long currentTimeMillis;
        do {
            currentTimeMillis = System.currentTimeMillis();
        }
        while (currentTimeMillis <= lastTimestamp);
        return currentTimeMillis;
    }

    /**
     * Get util instance
     * @param dataCenterId
     * @param workerId
     * @return
     */
    public static SnowflakeUtils getInstance(long dataCenterId, long workerId) {
        return new SnowflakeUtils(dataCenterId, workerId);
    }
}