工业领域的Hadoop架构学习~系列文章13：数据湖架构 - 工业大数据的统一存储底座

第13期：数据湖架构 - 工业大数据的统一存储底座

导言：数据湖是工业大数据平台的核心基础设施，它解决了多源异构数据的统一存储与分析难题。本期深入对比Delta Lake、Apache Iceberg、Apache Hudi三大开源数据湖方案，从架构原理出发，详细讲解表格式事务、时间旅行、增量处理等核心能力，并给出工业场景的选型建议与实战代码。

13.1 数据湖核心概念与架构演进

13.1.1 工业数据湖需求分析

复制代码

工业大数据特点与数据湖需求映射：

┌─────────────────────────────────────────────────────────────────┐
│                    工业数据类型与存储需求                         │
├─────────────────┬─────────────────┬─────────────────────────────┤
│     数据类型     │     数据特征      │       数据湖需求           │
├─────────────────┼─────────────────┼─────────────────────────────┤
│ 时序传感器数据   │ TB/天，高频写入   │ 高吞吐写入，压缩存储         │
│ 生产日志文件     │ 文本/JSON，追加   │ 对象存储原生支持            │
│ 历史工艺参数     │ 定点采集，结构化   │ ACID事务，更新支持          │
│ 图像/视频数据   │ 非结构化，海量    │ S3兼容，元数据分离          │
│ 实时告警流      │ 高吞吐，流式     │ 流批一体，Schema演化        │
│ 外部ERP/MES    │ 定期同步，更新   │ UPSERT，CDC支持            │
└─────────────────┴─────────────────┴─────────────────────────────┘

数据湖核心能力要求：
1. Schema管理 - 统一元数据，类型安全
2. 事务支持 - ACID语义，隔离级别
3. 时间旅行 - 版本回溯，快照查询
4. 增量处理 - CDC同步，变更捕获
5. 优化治理 - 压缩/排序/清理

#mermaid-svg-cQ1bedwxzmeAbF0Z{font-family:"trebuchet ms",verdana,arial,sans-serif;font-size:16px;fill:#333;}@keyframes edge-animation-frame{from{stroke-dashoffset:0;}}@keyframes dash{to{stroke-dashoffset:0;}}#mermaid-svg-cQ1bedwxzmeAbF0Z .edge-animation-slow{stroke-dasharray:9,5!important;stroke-dashoffset:900;animation:dash 50s linear infinite;stroke-linecap:round;}#mermaid-svg-cQ1bedwxzmeAbF0Z .edge-animation-fast{stroke-dasharray:9,5!important;stroke-dashoffset:900;animation:dash 20s linear infinite;stroke-linecap:round;}#mermaid-svg-cQ1bedwxzmeAbF0Z .error-icon{fill:#552222;}#mermaid-svg-cQ1bedwxzmeAbF0Z .error-text{fill:#552222;stroke:#552222;}#mermaid-svg-cQ1bedwxzmeAbF0Z .edge-thickness-normal{stroke-width:1px;}#mermaid-svg-cQ1bedwxzmeAbF0Z .edge-thickness-thick{stroke-width:3.5px;}#mermaid-svg-cQ1bedwxzmeAbF0Z .edge-pattern-solid{stroke-dasharray:0;}#mermaid-svg-cQ1bedwxzmeAbF0Z .edge-thickness-invisible{stroke-width:0;fill:none;}#mermaid-svg-cQ1bedwxzmeAbF0Z .edge-pattern-dashed{stroke-dasharray:3;}#mermaid-svg-cQ1bedwxzmeAbF0Z .edge-pattern-dotted{stroke-dasharray:2;}#mermaid-svg-cQ1bedwxzmeAbF0Z .marker{fill:#333333;stroke:#333333;}#mermaid-svg-cQ1bedwxzmeAbF0Z .marker.cross{stroke:#333333;}#mermaid-svg-cQ1bedwxzmeAbF0Z svg{font-family:"trebuchet ms",verdana,arial,sans-serif;font-size:16px;}#mermaid-svg-cQ1bedwxzmeAbF0Z p{margin:0;}#mermaid-svg-cQ1bedwxzmeAbF0Z .label{font-family:"trebuchet ms",verdana,arial,sans-serif;color:#333;}#mermaid-svg-cQ1bedwxzmeAbF0Z .cluster-label text{fill:#333;}#mermaid-svg-cQ1bedwxzmeAbF0Z .cluster-label span{color:#333;}#mermaid-svg-cQ1bedwxzmeAbF0Z .cluster-label span p{background-color:transparent;}#mermaid-svg-cQ1bedwxzmeAbF0Z .label text,#mermaid-svg-cQ1bedwxzmeAbF0Z span{fill:#333;color:#333;}#mermaid-svg-cQ1bedwxzmeAbF0Z .node rect,#mermaid-svg-cQ1bedwxzmeAbF0Z .node circle,#mermaid-svg-cQ1bedwxzmeAbF0Z .node ellipse,#mermaid-svg-cQ1bedwxzmeAbF0Z .node polygon,#mermaid-svg-cQ1bedwxzmeAbF0Z .node path{fill:#ECECFF;stroke:#9370DB;stroke-width:1px;}#mermaid-svg-cQ1bedwxzmeAbF0Z .rough-node .label text,#mermaid-svg-cQ1bedwxzmeAbF0Z .node .label text,#mermaid-svg-cQ1bedwxzmeAbF0Z .image-shape .label,#mermaid-svg-cQ1bedwxzmeAbF0Z .icon-shape .label{text-anchor:middle;}#mermaid-svg-cQ1bedwxzmeAbF0Z .node .katex path{fill:#000;stroke:#000;stroke-width:1px;}#mermaid-svg-cQ1bedwxzmeAbF0Z .rough-node .label,#mermaid-svg-cQ1bedwxzmeAbF0Z .node .label,#mermaid-svg-cQ1bedwxzmeAbF0Z .image-shape .label,#mermaid-svg-cQ1bedwxzmeAbF0Z .icon-shape .label{text-align:center;}#mermaid-svg-cQ1bedwxzmeAbF0Z .node.clickable{cursor:pointer;}#mermaid-svg-cQ1bedwxzmeAbF0Z .root .anchor path{fill:#333333!important;stroke-width:0;stroke:#333333;}#mermaid-svg-cQ1bedwxzmeAbF0Z .arrowheadPath{fill:#333333;}#mermaid-svg-cQ1bedwxzmeAbF0Z .edgePath .path{stroke:#333333;stroke-width:2.0px;}#mermaid-svg-cQ1bedwxzmeAbF0Z .flowchart-link{stroke:#333333;fill:none;}#mermaid-svg-cQ1bedwxzmeAbF0Z .edgeLabel{background-color:rgba(232,232,232, 0.8);text-align:center;}#mermaid-svg-cQ1bedwxzmeAbF0Z .edgeLabel p{background-color:rgba(232,232,232, 0.8);}#mermaid-svg-cQ1bedwxzmeAbF0Z .edgeLabel rect{opacity:0.5;background-color:rgba(232,232,232, 0.8);fill:rgba(232,232,232, 0.8);}#mermaid-svg-cQ1bedwxzmeAbF0Z .labelBkg{background-color:rgba(232, 232, 232, 0.5);}#mermaid-svg-cQ1bedwxzmeAbF0Z .cluster rect{fill:#ffffde;stroke:#aaaa33;stroke-width:1px;}#mermaid-svg-cQ1bedwxzmeAbF0Z .cluster text{fill:#333;}#mermaid-svg-cQ1bedwxzmeAbF0Z .cluster span{color:#333;}#mermaid-svg-cQ1bedwxzmeAbF0Z div.mermaidTooltip{position:absolute;text-align:center;max-width:200px;padding:2px;font-family:"trebuchet ms",verdana,arial,sans-serif;font-size:12px;background:hsl(80, 100%, 96.2745098039%);border:1px solid #aaaa33;border-radius:2px;pointer-events:none;z-index:100;}#mermaid-svg-cQ1bedwxzmeAbF0Z .flowchartTitleText{text-anchor:middle;font-size:18px;fill:#333;}#mermaid-svg-cQ1bedwxzmeAbF0Z rect.text{fill:none;stroke-width:0;}#mermaid-svg-cQ1bedwxzmeAbF0Z .icon-shape,#mermaid-svg-cQ1bedwxzmeAbF0Z .image-shape{background-color:rgba(232,232,232, 0.8);text-align:center;}#mermaid-svg-cQ1bedwxzmeAbF0Z .icon-shape p,#mermaid-svg-cQ1bedwxzmeAbF0Z .image-shape p{background-color:rgba(232,232,232, 0.8);padding:2px;}#mermaid-svg-cQ1bedwxzmeAbF0Z .icon-shape .label rect,#mermaid-svg-cQ1bedwxzmeAbF0Z .image-shape .label rect{opacity:0.5;background-color:rgba(232,232,232, 0.8);fill:rgba(232,232,232, 0.8);}#mermaid-svg-cQ1bedwxzmeAbF0Z .label-icon{display:inline-block;height:1em;overflow:visible;vertical-align:-0.125em;}#mermaid-svg-cQ1bedwxzmeAbF0Z .node .label-icon path{fill:currentColor;stroke:revert;stroke-width:revert;}#mermaid-svg-cQ1bedwxzmeAbF0Z :root{--mermaid-font-family:"trebuchet ms",verdana,arial,sans-serif;} 计算引擎
Spark SQL
Presto
Flink SQL
数据湖
表格式层
Delta Lake
Iceberg
Hudi
统一元数据
对象存储
摄入层
批量导入
流式写入
CDC同步
数据源
SCADA系统
MES系统
ERP系统
IoT传感器
应用日志

13.1.2 数据湖架构模式对比

复制代码

┌────────────────────────────────────────────────────────────────────┐
│                    三种数据湖架构模式对比                            │
├──────────────────┬──────────────┬──────────────┬──────────────────┤
│      特性        │   纯数据湖     │   数据仓库    │   Lakehouse     │
├──────────────────┼──────────────┼──────────────┼──────────────────┤
│ 数据存储         │ S3/HDFS      │ 专用存储      │ S3/HDFS         │
│ Schema管理       │ 弱           │ 强           │ 强              │
│ ACID事务         │ ❌           │ ✅           │ ✅              │
│ 时间旅行         │ ❌           │ ✅           │ ✅              │
│ 增量处理         │ ❌           │ ✅           │ ✅              │
│ 流批一体         │ 流独立       │ 批优先       │ 流批一体        │
│ 数据类型         │ 全类型       │ 结构化       │ 全类型          │
│ 查询性能         │ 低           │ 高           │ 中高            │
│ 生态集成         │ 灵活         │ 绑定厂商     │ 开放            │
└──────────────────┴──────────────┴──────────────┴──────────────────┘

工业场景推荐：Lakehouse架构（Delta Lake / Iceberg / Hudi）

13.2 Apache Iceberg架构与实战

13.2.1 Iceberg核心概念与架构

Apache Iceberg是Netflix开源的表格式规范，其核心设计理念：

复制代码

Iceberg核心架构原理：

1. 快照模型 (Snapshot Model)
   Iceberg表中每个时刻的状态是一个Snapshot
   Snapshot = Manifest List + Data Files
   支持时间旅行查询与分支操作

2. 分区隐式转换 (Hidden Partition)
   用户按业务日期查询，Iceberg自动转换分区
   避免数据倾斜与热点问题

3. 元数据层架构
   Table Metadata (表级别)
       ↓
   Snapshot (快照版本)
       ↓
   Manifest List (文件清单)
       ↓
   Manifest (文件元数据)
       ↓
   Data Files (实际数据)

4. 并发控制
   MVCC + 乐观锁
   optimistic concurrent write

#mermaid-svg-exw9AQdtDWQPfZvy{font-family:"trebuchet ms",verdana,arial,sans-serif;font-size:16px;fill:#333;}@keyframes edge-animation-frame{from{stroke-dashoffset:0;}}@keyframes dash{to{stroke-dashoffset:0;}}#mermaid-svg-exw9AQdtDWQPfZvy .edge-animation-slow{stroke-dasharray:9,5!important;stroke-dashoffset:900;animation:dash 50s linear infinite;stroke-linecap:round;}#mermaid-svg-exw9AQdtDWQPfZvy .edge-animation-fast{stroke-dasharray:9,5!important;stroke-dashoffset:900;animation:dash 20s linear infinite;stroke-linecap:round;}#mermaid-svg-exw9AQdtDWQPfZvy .error-icon{fill:#552222;}#mermaid-svg-exw9AQdtDWQPfZvy .error-text{fill:#552222;stroke:#552222;}#mermaid-svg-exw9AQdtDWQPfZvy .edge-thickness-normal{stroke-width:1px;}#mermaid-svg-exw9AQdtDWQPfZvy .edge-thickness-thick{stroke-width:3.5px;}#mermaid-svg-exw9AQdtDWQPfZvy .edge-pattern-solid{stroke-dasharray:0;}#mermaid-svg-exw9AQdtDWQPfZvy .edge-thickness-invisible{stroke-width:0;fill:none;}#mermaid-svg-exw9AQdtDWQPfZvy .edge-pattern-dashed{stroke-dasharray:3;}#mermaid-svg-exw9AQdtDWQPfZvy .edge-pattern-dotted{stroke-dasharray:2;}#mermaid-svg-exw9AQdtDWQPfZvy .marker{fill:#333333;stroke:#333333;}#mermaid-svg-exw9AQdtDWQPfZvy .marker.cross{stroke:#333333;}#mermaid-svg-exw9AQdtDWQPfZvy svg{font-family:"trebuchet ms",verdana,arial,sans-serif;font-size:16px;}#mermaid-svg-exw9AQdtDWQPfZvy p{margin:0;}#mermaid-svg-exw9AQdtDWQPfZvy .label{font-family:"trebuchet ms",verdana,arial,sans-serif;color:#333;}#mermaid-svg-exw9AQdtDWQPfZvy .cluster-label text{fill:#333;}#mermaid-svg-exw9AQdtDWQPfZvy .cluster-label span{color:#333;}#mermaid-svg-exw9AQdtDWQPfZvy .cluster-label span p{background-color:transparent;}#mermaid-svg-exw9AQdtDWQPfZvy .label text,#mermaid-svg-exw9AQdtDWQPfZvy span{fill:#333;color:#333;}#mermaid-svg-exw9AQdtDWQPfZvy .node rect,#mermaid-svg-exw9AQdtDWQPfZvy .node circle,#mermaid-svg-exw9AQdtDWQPfZvy .node ellipse,#mermaid-svg-exw9AQdtDWQPfZvy .node polygon,#mermaid-svg-exw9AQdtDWQPfZvy .node path{fill:#ECECFF;stroke:#9370DB;stroke-width:1px;}#mermaid-svg-exw9AQdtDWQPfZvy .rough-node .label text,#mermaid-svg-exw9AQdtDWQPfZvy .node .label text,#mermaid-svg-exw9AQdtDWQPfZvy .image-shape .label,#mermaid-svg-exw9AQdtDWQPfZvy .icon-shape .label{text-anchor:middle;}#mermaid-svg-exw9AQdtDWQPfZvy .node .katex path{fill:#000;stroke:#000;stroke-width:1px;}#mermaid-svg-exw9AQdtDWQPfZvy .rough-node .label,#mermaid-svg-exw9AQdtDWQPfZvy .node .label,#mermaid-svg-exw9AQdtDWQPfZvy .image-shape .label,#mermaid-svg-exw9AQdtDWQPfZvy .icon-shape .label{text-align:center;}#mermaid-svg-exw9AQdtDWQPfZvy .node.clickable{cursor:pointer;}#mermaid-svg-exw9AQdtDWQPfZvy .root .anchor path{fill:#333333!important;stroke-width:0;stroke:#333333;}#mermaid-svg-exw9AQdtDWQPfZvy .arrowheadPath{fill:#333333;}#mermaid-svg-exw9AQdtDWQPfZvy .edgePath .path{stroke:#333333;stroke-width:2.0px;}#mermaid-svg-exw9AQdtDWQPfZvy .flowchart-link{stroke:#333333;fill:none;}#mermaid-svg-exw9AQdtDWQPfZvy .edgeLabel{background-color:rgba(232,232,232, 0.8);text-align:center;}#mermaid-svg-exw9AQdtDWQPfZvy .edgeLabel p{background-color:rgba(232,232,232, 0.8);}#mermaid-svg-exw9AQdtDWQPfZvy .edgeLabel rect{opacity:0.5;background-color:rgba(232,232,232, 0.8);fill:rgba(232,232,232, 0.8);}#mermaid-svg-exw9AQdtDWQPfZvy .labelBkg{background-color:rgba(232, 232, 232, 0.5);}#mermaid-svg-exw9AQdtDWQPfZvy .cluster rect{fill:#ffffde;stroke:#aaaa33;stroke-width:1px;}#mermaid-svg-exw9AQdtDWQPfZvy .cluster text{fill:#333;}#mermaid-svg-exw9AQdtDWQPfZvy .cluster span{color:#333;}#mermaid-svg-exw9AQdtDWQPfZvy div.mermaidTooltip{position:absolute;text-align:center;max-width:200px;padding:2px;font-family:"trebuchet ms",verdana,arial,sans-serif;font-size:12px;background:hsl(80, 100%, 96.2745098039%);border:1px solid #aaaa33;border-radius:2px;pointer-events:none;z-index:100;}#mermaid-svg-exw9AQdtDWQPfZvy .flowchartTitleText{text-anchor:middle;font-size:18px;fill:#333;}#mermaid-svg-exw9AQdtDWQPfZvy rect.text{fill:none;stroke-width:0;}#mermaid-svg-exw9AQdtDWQPfZvy .icon-shape,#mermaid-svg-exw9AQdtDWQPfZvy .image-shape{background-color:rgba(232,232,232, 0.8);text-align:center;}#mermaid-svg-exw9AQdtDWQPfZvy .icon-shape p,#mermaid-svg-exw9AQdtDWQPfZvy .image-shape p{background-color:rgba(232,232,232, 0.8);padding:2px;}#mermaid-svg-exw9AQdtDWQPfZvy .icon-shape .label rect,#mermaid-svg-exw9AQdtDWQPfZvy .image-shape .label rect{opacity:0.5;background-color:rgba(232,232,232, 0.8);fill:rgba(232,232,232, 0.8);}#mermaid-svg-exw9AQdtDWQPfZvy .label-icon{display:inline-block;height:1em;overflow:visible;vertical-align:-0.125em;}#mermaid-svg-exw9AQdtDWQPfZvy .node .label-icon path{fill:currentColor;stroke:revert;stroke-width:revert;}#mermaid-svg-exw9AQdtDWQPfZvy :root{--mermaid-font-family:"trebuchet ms",verdana,arial,sans-serif;} Data Files
Manifests
Manifest List
Snapshots
Table Metadata
metadata.json
current_metadata_location
snapshot-1 (v1)
snapshot-2 (v2)
snapshot-3 (v3)
manifest-list-2
manifest-list-3
manifest-1.avro
manifest-2.avro
manifest-3.avro
parquet-1.parquet
parquet-2.parquet
parquet-3.parquet
parquet-4.parquet

13.2.2 Iceberg Java API实战

java 复制代码

// Iceberg表创建与操作
package com.industrial.datalake;

import org.apache.iceberg.*;
import org.apache.iceberg.catalog.*;
import org.apache.iceberg.catalog.TableIdentifier;
import org.apache.iceberg.types.Types.*;

public class IndustrialDataLake {
    
    private final Catalog catalog;
    
    public IndustrialDataLake(Catalog catalog) {
        this.catalog = catalog;
    }
    
    /**
     * 创建工业传感器数据表
     */
    public Table createSensorDataTable(String namespace, String tableName) {
        // 定义表Schema
        Schema schema = new Schema(
            StructField.of("sensor_id",     Types.StringType.get(),     false, "传感器ID"),
            StructField.of("timestamp",      Types.TimestampType.withZone(), false, "采集时间"),
            StructField.of("location",       Types.StringType.get(),     false, "设备位置"),
            StructField.of("temperature",    Types.DoubleType.get(),     true,  "温度值"),
            StructField.of("pressure",       Types.DoubleType.get(),     true,  "压力值"),
            StructField.of("vibration",      Types.DoubleType.get(),     true,  "振动值"),
            StructField.of("process_id",     Types.StringType.get(),     true,  "工艺编号"),
            StructField.of("quality_flag",   Types.IntegerType.get(),    true,  "质量标识")
        );
        
        // 定义分区策略 - 按小时分区
        PartitionSpec partitionSpec = PartitionSpec.builderFor(schema)
            .hour("timestamp")          // 按小时分区
            .identity("location")        // 按位置分区
            .build();
        
        // 定义表属性
        TableProperties properties = TableProperties.builder()
            .put("write.format.default", "parquet")
            .put("write.parquet.compression-codec", "zstd")
            .put("read.split.target-size", "134217728")  // 128MB
            .put("write.target-file-size-bytes", "134217728")
            .build();
        
        // 创建表
        Table table = catalog.buildTable(
            TableIdentifier.of(namespace, tableName),
            schema
        ).withPartitionSpec(partitionSpec)
         .withProperties(properties)
         .create();
         
        return table;
    }
    
    /**
     * 批量写入传感器数据
     */
    public void batchWriteSensorData(Table table, List<SensorRecord> records) {
        // 构建DataWriter
        DataWriter<Record> writer = Avro.createDataWriter(
            table, 
            FlinkWriteBuilder.builderFor(writer)
                .withSpec(table.spec())
                .withLocation("")
                .build(),
            schema,
            outputFile
        );
        
        // 写入数据
        try (DataWriter<Record> dataWriter = writer) {
            for (SensorRecord record : records) {
                Record icebergRecord = convertToIcebergRecord(record);
                dataWriter.add(icebergRecord);
            }
        }
        
        // 提交事务
        table.newAppend()
            .appendManifest(dataWriter.toAvro())
            .commit();
    }
    
    /**
     * 时间旅行查询 - 查询历史版本数据
     */
    public Dataset<Row> timeTravelQuery(Table table, String versionOrTimestamp) {
        // 方式1: 通过snapshot ID查询
        // Dataset<Row> df = spark.read()
        //     .format("iceberg")
        //     .option("snapshot-id", "10963874102873L")
        //     .load("industrial.sensor_data");
        
        // 方式2: 通过时间戳查询
        Dataset<Row> df = spark.read()
            .format("iceberg")
            .option("as-of-timestamp", versionOrTimestamp)
            .load("industrial.sensor_data");
            
        // 方式3: 通过版本号查询
        // Dataset<Row> df = spark.read()
        //     .format("iceberg")
        //     .option("version", "3")
        //     .load("industrial.sensor_data");
        
        return df;
    }
    
    /**
     * 增量读取 - CDC场景
     */
    public Dataset<Row> incrementalRead(Table table, long fromSnapshotId) {
        // 获取增量数据（用于CDC同步）
        SparkScan scan = (SparkScan) table.newScan()
            .filter(Expressions.equal("quality_flag", 1))  // 只读取异常数据
            .useSnapshot(fromSnapshotId);
            
        return spark.read().format("iceberg").load(table.location());
    }
}

13.2.3 Iceberg Spark SQL操作

sql 复制代码

-- Iceberg SQL DDL与DML

-- 1. 创建数据库与表
CREATE DATABASE IF NOT EXISTS industrial;
USE industrial;

-- 创建传感器数据表（带分区）
CREATE TABLE sensor_data (
    sensor_id    STRING,
    timestamp    TIMESTAMP,
    location     STRING,
    temperature  DOUBLE,
    pressure     DOUBLE,
    vibration    DOUBLE,
    process_id   STRING,
    quality_flag INT
)
USING iceberg
PARTITIONED BY (hours(timestamp), location)
TBLPROPERTIES (
    'write.format.default' = 'parquet',
    'write.parquet.compression-codec' = 'zstd',
    'history.expire.max-snapshot-age-ms' = '604800000'  -- 7天
);

-- 2. 插入数据
INSERT INTO sensor_data VALUES
    ('S001', '2024-01-15 10:30:00', 'Line-A-01', 85.5, 1.2, 0.05, 'P001', 0),
    ('S002', '2024-01-15 10:30:00', 'Line-A-02', 86.1, 1.3, 0.06, 'P001', 0),
    ('S003', '2024-01-15 10:30:00', 'Line-B-01', 84.2, 1.1, 0.04, 'P002', 1);

-- 批量插入
INSERT INTO sensor_data
SELECT * FROM staging_sensor_data WHERE processing_date = '2024-01-15';

-- 3. 时间旅行查询
-- 查询指定时间点的数据
SELECT * FROM sensor_data TIMESTAMP AS OF '2024-01-15 12:00:00';
SELECT * FROM sensor_data VERSION AS OF 3;

-- 4. 查询历史变更
SELECT * FROM sensor_data HISTORY;  -- 查看所有版本

-- 5. 更新数据（ACID事务支持）
UPDATE sensor_data 
SET quality_flag = 2 
WHERE sensor_id = 'S001' AND timestamp = '2024-01-15 10:30:00';

-- 6. MERGE INTO（Upsert场景）
MERGE INTO sensor_data AS target
USING sensor_data_incremental AS source
ON target.sensor_id = source.sensor_id 
   AND target.timestamp = source.timestamp
WHEN MATCHED THEN UPDATE SET 
    temperature = source.temperature,
    quality_flag = source.quality_flag
WHEN NOT MATCHED THEN INSERT *;

-- 7. 删除数据
DELETE FROM sensor_data 
WHERE timestamp < '2024-01-01' AND quality_flag = 0;

-- 8. 快照过期与文件清理
CALL system.expire_snapshots('industrial.sensor_data', TIMESTAMP '2024-01-15 00:00:00', 100);
CALL system.rewrite_data_files('industrial.sensor_data');

13.3 Delta Lake架构与实战

13.3.1 Delta Lake核心特性

复制代码

Delta Lake核心能力矩阵：

┌─────────────────────────────────────────────────────────────────┐
│                     Delta Lake特性体系                           │
├─────────────────┬───────────────────────────────────────────────┤
│     特性        │                    说明                       │
├─────────────────┼───────────────────────────────────────────────┤
│ ACID事务        │ 写入即提交，失败回滚，支持并发写入              │
│ 可扩展元数据    │ 元数据也存储为Parquet，支持PB级表              │
│ 时间旅行        │ 访问历史版本，数据回滚与审计                    │
│ UPSERT/MERGE   │ 支持CDC同步，全量/增量更新                     │
│ 流批一体        │ 既是数据源也是数据汇，支持Streaming Sink        │
│ Schema执行     │ 自动验证，阻止脏数据写入                        │
│ 开放格式        │ Parquet格式，兼容各类引擎                       │
│ 缓存加速        │ Delta Lake Cache (Databricks)                 │
└─────────────────┴───────────────────────────────────────────────┘

13.3.2 Delta Lake实战代码

python 复制代码

# Delta Lake工业数据湖实战
from delta.tables import DeltaTable
from pyspark.sql import SparkSession
from pyspark.sql.functions import *
from pyspark.sql.types import *

# 初始化Spark
spark = SparkSession.builder \
    .appName("IndustrialDataLake") \
    .config("spark.sql.extensions", "io.delta.sql.DeltaSparkSessionExtension") \
    .config("spark.sql.catalog.spark_catalog", "org.apache.spark.sql.delta.catalog.DeltaCatalog") \
    .getOrCreate()

def create_sensor_data_table():
    """创建传感器数据表"""
    
    schema = StructType([
        StructField("sensor_id", StringType(), False, "传感器ID"),
        StructField("timestamp", TimestampType(), False, "采集时间"),
        StructField("location", StringType(), False, "设备位置"),
        StructField("temperature", DoubleType(), True, "温度"),
        StructField("pressure", DoubleType(), True, "压力"),
        StructField("vibration", DoubleType(), True, "振动"),
        StructField("process_id", StringType(), True, "工艺编号"),
        StructField("quality_flag", IntegerType(), True, "质量标识"),
        StructField("_rescued_data", StringType(), True, "未匹配列数据")
    ])
    
    # 创建表
    (spark.createDataFrame([], schema)
        .write
        .format("delta")
        .mode("ignore")
        .option("mergeSchema", "true")
        .partitionBy("timestamp")  # 按时间分区
        .saveAsTable("industrial.sensor_data"))

def upsert_sensor_data(batch_df, batch_id):
    """增量Upsert - CDC同步场景"""
    
    delta_table = DeltaTable.forName(spark, "industrial.sensor_data")
    
    # UPSERT逻辑：匹配则更新，不匹配则插入
    (delta_table.alias("target")
        .merge(
            batch_df.alias("source"),
            "target.sensor_id = source.sensor_id AND " +
            "target.timestamp = source.timestamp"
        )
        .whenMatchedUpdateAll()    # 匹配则更新所有列
        .whenNotMatchedInsertAll() # 不匹配则插入
        .execute())
        
    print(f"Batch {batch_id} upsert completed")

def streaming_to_delta():
    """流式写入Delta Lake"""
    
    # 定义Kafka数据源
    kafka_df = (spark
        .readStream
        .format("kafka")
        .option("kafka.bootstrap.servers", "kafka:9092")
        .option("subscribe", "industrial-sensors")
        .option("startingOffsets", "latest")
        .load())
    
    # 解析JSON数据
    sensor_schema = StructType([
        StructField("sensor_id", StringType()),
        StructField("timestamp", StringType()),
        StructField("location", StringType()),
        StructField("temperature", DoubleType()),
        StructField("pressure", DoubleType()),
        StructField("vibration", DoubleType()),
        StructField("process_id", StringType())
    ])
    
    parsed_df = (kafka_df
        .select(from_json(col("value").cast("string"), sensor_schema).alias("data"))
        .select("data.*")
        .withColumn("timestamp", to_timestamp("timestamp"))
        .withColumn("quality_flag", lit(0)))
    
    # 流式写入Delta Lake
    query = (parsed_df
        .writeStream
        .format("delta")
        .outputMode("append")
        .option("checkpointLocation", "s3://warehouse/checkpoints/sensor_data")
        .trigger(processingTime="5 seconds")
        .start("s3://warehouse/sensor_data"))
    
    return query

def time_travel_queries():
    """时间旅行查询"""
    
    # 查询当前版本
    current_df = spark.read.format("delta").table("industrial.sensor_data")
    
    # 查询指定版本
    v3_df = spark.read \
        .format("delta") \
        .option("versionAsOf", 3) \
        .load("s3://warehouse/sensor_data")
    
    # 查询指定时间点
    yesterday_df = spark.read \
        .format("delta") \
        .option("timestampAsOf", "2024-01-14 00:00:00") \
        .load("s3://warehouse/sensor_data")
    
    # 获取数据历史
    history_df = DeltaTable.forName(spark, "industrial.sensor_data").history()
    history_df.show()
    
    # 数据回滚（撤销删除操作）
    delta_table = DeltaTable.forName(spark, "industrial.sensor_data")
    delta_table.restoreToVersion(10)  # 回滚到第10个版本

13.4 数据湖方案选型与最佳实践

13.4.1 三大数据湖方案对比

复制代码

┌────────────────────────────────────────────────────────────────────────┐
│                   Delta Lake vs Iceberg vs Hudi 对比                   │
├─────────────────┬──────────────┬──────────────┬───────────────────────┤
│      特性        │ Delta Lake  │  Apache Iceberg │    Apache Hudi      │
├─────────────────┼──────────────┼──────────────┼───────────────────────┤
│ 社区活跃度       │ ⭐⭐⭐⭐⭐     │ ⭐⭐⭐⭐      │ ⭐⭐⭐⭐               │
│ 云厂商支持       │ Databricks   │ AWS/腾讯云   │ Uber内部             │
│                  │ Azure        │ Netflix      │ 阿里的Flink-Hudi     │
├─────────────────┼──────────────┼──────────────┼───────────────────────┤
│ 核心优势        │ 生态完善      │ 架构优秀     │ 增量处理             │
│                  │ Spark深度集成 │ 分区隐式转换 │ 快速Upsert           │
│                  │ 性能优化好    │ 开放标准     │ CDC支持好            │
├─────────────────┼──────────────┼──────────────┼───────────────────────┤
│ 适用场景        │ Databricks环境│ 通用场景     │ 数据CDC同步          │
│                  │ 已有Spark    │ 多引擎统一   │ 实时数仓             │
│                  │ 追求性能     │ 需要开放标准  │ 需要更新/删除历史数据 │
├─────────────────┼──────────────┼──────────────┼───────────────────────┤
│ 工业场景评分     │ ⭐⭐⭐⭐      │ ⭐⭐⭐⭐⭐     │ ⭐⭐⭐⭐               │
│ 推荐理由        │ 成熟稳定      │ 架构先进开放 │ 增量同步首选         │
└─────────────────┴──────────────┴──────────────┴───────────────────────┘

13.4.2 工业场景选型决策树

#mermaid-svg-WMgUHFaXxsrOe7cC{font-family:"trebuchet ms",verdana,arial,sans-serif;font-size:16px;fill:#333;}@keyframes edge-animation-frame{from{stroke-dashoffset:0;}}@keyframes dash{to{stroke-dashoffset:0;}}#mermaid-svg-WMgUHFaXxsrOe7cC .edge-animation-slow{stroke-dasharray:9,5!important;stroke-dashoffset:900;animation:dash 50s linear infinite;stroke-linecap:round;}#mermaid-svg-WMgUHFaXxsrOe7cC .edge-animation-fast{stroke-dasharray:9,5!important;stroke-dashoffset:900;animation:dash 20s linear infinite;stroke-linecap:round;}#mermaid-svg-WMgUHFaXxsrOe7cC .error-icon{fill:#552222;}#mermaid-svg-WMgUHFaXxsrOe7cC .error-text{fill:#552222;stroke:#552222;}#mermaid-svg-WMgUHFaXxsrOe7cC .edge-thickness-normal{stroke-width:1px;}#mermaid-svg-WMgUHFaXxsrOe7cC .edge-thickness-thick{stroke-width:3.5px;}#mermaid-svg-WMgUHFaXxsrOe7cC .edge-pattern-solid{stroke-dasharray:0;}#mermaid-svg-WMgUHFaXxsrOe7cC .edge-thickness-invisible{stroke-width:0;fill:none;}#mermaid-svg-WMgUHFaXxsrOe7cC .edge-pattern-dashed{stroke-dasharray:3;}#mermaid-svg-WMgUHFaXxsrOe7cC .edge-pattern-dotted{stroke-dasharray:2;}#mermaid-svg-WMgUHFaXxsrOe7cC .marker{fill:#333333;stroke:#333333;}#mermaid-svg-WMgUHFaXxsrOe7cC .marker.cross{stroke:#333333;}#mermaid-svg-WMgUHFaXxsrOe7cC svg{font-family:"trebuchet ms",verdana,arial,sans-serif;font-size:16px;}#mermaid-svg-WMgUHFaXxsrOe7cC p{margin:0;}#mermaid-svg-WMgUHFaXxsrOe7cC .label{font-family:"trebuchet ms",verdana,arial,sans-serif;color:#333;}#mermaid-svg-WMgUHFaXxsrOe7cC .cluster-label text{fill:#333;}#mermaid-svg-WMgUHFaXxsrOe7cC .cluster-label span{color:#333;}#mermaid-svg-WMgUHFaXxsrOe7cC .cluster-label span p{background-color:transparent;}#mermaid-svg-WMgUHFaXxsrOe7cC .label text,#mermaid-svg-WMgUHFaXxsrOe7cC span{fill:#333;color:#333;}#mermaid-svg-WMgUHFaXxsrOe7cC .node rect,#mermaid-svg-WMgUHFaXxsrOe7cC .node circle,#mermaid-svg-WMgUHFaXxsrOe7cC .node ellipse,#mermaid-svg-WMgUHFaXxsrOe7cC .node polygon,#mermaid-svg-WMgUHFaXxsrOe7cC .node path{fill:#ECECFF;stroke:#9370DB;stroke-width:1px;}#mermaid-svg-WMgUHFaXxsrOe7cC .rough-node .label text,#mermaid-svg-WMgUHFaXxsrOe7cC .node .label text,#mermaid-svg-WMgUHFaXxsrOe7cC .image-shape .label,#mermaid-svg-WMgUHFaXxsrOe7cC .icon-shape .label{text-anchor:middle;}#mermaid-svg-WMgUHFaXxsrOe7cC .node .katex path{fill:#000;stroke:#000;stroke-width:1px;}#mermaid-svg-WMgUHFaXxsrOe7cC .rough-node .label,#mermaid-svg-WMgUHFaXxsrOe7cC .node .label,#mermaid-svg-WMgUHFaXxsrOe7cC .image-shape .label,#mermaid-svg-WMgUHFaXxsrOe7cC .icon-shape .label{text-align:center;}#mermaid-svg-WMgUHFaXxsrOe7cC .node.clickable{cursor:pointer;}#mermaid-svg-WMgUHFaXxsrOe7cC .root .anchor path{fill:#333333!important;stroke-width:0;stroke:#333333;}#mermaid-svg-WMgUHFaXxsrOe7cC .arrowheadPath{fill:#333333;}#mermaid-svg-WMgUHFaXxsrOe7cC .edgePath .path{stroke:#333333;stroke-width:2.0px;}#mermaid-svg-WMgUHFaXxsrOe7cC .flowchart-link{stroke:#333333;fill:none;}#mermaid-svg-WMgUHFaXxsrOe7cC .edgeLabel{background-color:rgba(232,232,232, 0.8);text-align:center;}#mermaid-svg-WMgUHFaXxsrOe7cC .edgeLabel p{background-color:rgba(232,232,232, 0.8);}#mermaid-svg-WMgUHFaXxsrOe7cC .edgeLabel rect{opacity:0.5;background-color:rgba(232,232,232, 0.8);fill:rgba(232,232,232, 0.8);}#mermaid-svg-WMgUHFaXxsrOe7cC .labelBkg{background-color:rgba(232, 232, 232, 0.5);}#mermaid-svg-WMgUHFaXxsrOe7cC .cluster rect{fill:#ffffde;stroke:#aaaa33;stroke-width:1px;}#mermaid-svg-WMgUHFaXxsrOe7cC .cluster text{fill:#333;}#mermaid-svg-WMgUHFaXxsrOe7cC .cluster span{color:#333;}#mermaid-svg-WMgUHFaXxsrOe7cC div.mermaidTooltip{position:absolute;text-align:center;max-width:200px;padding:2px;font-family:"trebuchet ms",verdana,arial,sans-serif;font-size:12px;background:hsl(80, 100%, 96.2745098039%);border:1px solid #aaaa33;border-radius:2px;pointer-events:none;z-index:100;}#mermaid-svg-WMgUHFaXxsrOe7cC .flowchartTitleText{text-anchor:middle;font-size:18px;fill:#333;}#mermaid-svg-WMgUHFaXxsrOe7cC rect.text{fill:none;stroke-width:0;}#mermaid-svg-WMgUHFaXxsrOe7cC .icon-shape,#mermaid-svg-WMgUHFaXxsrOe7cC .image-shape{background-color:rgba(232,232,232, 0.8);text-align:center;}#mermaid-svg-WMgUHFaXxsrOe7cC .icon-shape p,#mermaid-svg-WMgUHFaXxsrOe7cC .image-shape p{background-color:rgba(232,232,232, 0.8);padding:2px;}#mermaid-svg-WMgUHFaXxsrOe7cC .icon-shape .label rect,#mermaid-svg-WMgUHFaXxsrOe7cC .image-shape .label rect{opacity:0.5;background-color:rgba(232,232,232, 0.8);fill:rgba(232,232,232, 0.8);}#mermaid-svg-WMgUHFaXxsrOe7cC .label-icon{display:inline-block;height:1em;overflow:visible;vertical-align:-0.125em;}#mermaid-svg-WMgUHFaXxsrOe7cC .node .label-icon path{fill:currentColor;stroke:revert;stroke-width:revert;}#mermaid-svg-WMgUHFaXxsrOe7cC :root{--mermaid-font-family:"trebuchet ms",verdana,arial,sans-serif;} 是
否
是
否
是
否
是
否
是
否
开始选型
已有Databricks?
选择Delta Lake
多引擎统一访问?
选择Apache Iceberg
需要高频CDC更新?
选择Apache Hudi
需要分区隐式转换?
追求成熟稳定?
部署验证

13.4.3 工业数据湖最佳实践配置

yaml 复制代码

# docker-compose.yml - 数据湖组件部署
version: '3.8'
services:
  minio:
    image: minio/minio:latest
    ports:
      - "9000:9000"
      - "9001:9001"
    environment:
      MINIO_ROOT_USER: industrial
      MINIO_ROOT_PASSWORD: data-lake-secret
    command: server /data --console-address ":9001"
    volumes:
      - minio-data:/data
    healthcheck:
      test: ["CMD", "mc", "ready", "local"]
      interval: 10s
      timeout: 5s
      retries: 5

  spark-iceberg:
    image: apache/iceberg-spark:3.5_1.0
    depends_on:
      - minio
    environment:
      AWS_ACCESS_KEY_ID: industrial
      AWS_SECRET_ACCESS_KEY: data-lake-secret
      AWS_REGION: us-east-1
      SPARK_iceberg_warehouse: s3a://warehouse/
    volumes:
      - ./warehouse:/home/iceberg/warehouse

  nessie:
    image: projectnessie/nessie:latest
    ports:
      - "19120:19120"
    environment:
      QUARKUS_HTTP_PORT: 19120

volumes:
  minio-data:

13.5 知识体系总结

#mermaid-svg-RAQQ5wosPTGxuKm9{font-family:"trebuchet ms",verdana,arial,sans-serif;font-size:16px;fill:#333;}@keyframes edge-animation-frame{from{stroke-dashoffset:0;}}@keyframes dash{to{stroke-dashoffset:0;}}#mermaid-svg-RAQQ5wosPTGxuKm9 .edge-animation-slow{stroke-dasharray:9,5!important;stroke-dashoffset:900;animation:dash 50s linear infinite;stroke-linecap:round;}#mermaid-svg-RAQQ5wosPTGxuKm9 .edge-animation-fast{stroke-dasharray:9,5!important;stroke-dashoffset:900;animation:dash 20s linear infinite;stroke-linecap:round;}#mermaid-svg-RAQQ5wosPTGxuKm9 .error-icon{fill:#552222;}#mermaid-svg-RAQQ5wosPTGxuKm9 .error-text{fill:#552222;stroke:#552222;}#mermaid-svg-RAQQ5wosPTGxuKm9 .edge-thickness-normal{stroke-width:1px;}#mermaid-svg-RAQQ5wosPTGxuKm9 .edge-thickness-thick{stroke-width:3.5px;}#mermaid-svg-RAQQ5wosPTGxuKm9 .edge-pattern-solid{stroke-dasharray:0;}#mermaid-svg-RAQQ5wosPTGxuKm9 .edge-thickness-invisible{stroke-width:0;fill:none;}#mermaid-svg-RAQQ5wosPTGxuKm9 .edge-pattern-dashed{stroke-dasharray:3;}#mermaid-svg-RAQQ5wosPTGxuKm9 .edge-pattern-dotted{stroke-dasharray:2;}#mermaid-svg-RAQQ5wosPTGxuKm9 .marker{fill:#333333;stroke:#333333;}#mermaid-svg-RAQQ5wosPTGxuKm9 .marker.cross{stroke:#333333;}#mermaid-svg-RAQQ5wosPTGxuKm9 svg{font-family:"trebuchet ms",verdana,arial,sans-serif;font-size:16px;}#mermaid-svg-RAQQ5wosPTGxuKm9 p{margin:0;}#mermaid-svg-RAQQ5wosPTGxuKm9 .label{font-family:"trebuchet ms",verdana,arial,sans-serif;color:#333;}#mermaid-svg-RAQQ5wosPTGxuKm9 .cluster-label text{fill:#333;}#mermaid-svg-RAQQ5wosPTGxuKm9 .cluster-label span{color:#333;}#mermaid-svg-RAQQ5wosPTGxuKm9 .cluster-label span p{background-color:transparent;}#mermaid-svg-RAQQ5wosPTGxuKm9 .label text,#mermaid-svg-RAQQ5wosPTGxuKm9 span{fill:#333;color:#333;}#mermaid-svg-RAQQ5wosPTGxuKm9 .node rect,#mermaid-svg-RAQQ5wosPTGxuKm9 .node circle,#mermaid-svg-RAQQ5wosPTGxuKm9 .node ellipse,#mermaid-svg-RAQQ5wosPTGxuKm9 .node polygon,#mermaid-svg-RAQQ5wosPTGxuKm9 .node path{fill:#ECECFF;stroke:#9370DB;stroke-width:1px;}#mermaid-svg-RAQQ5wosPTGxuKm9 .rough-node .label text,#mermaid-svg-RAQQ5wosPTGxuKm9 .node .label text,#mermaid-svg-RAQQ5wosPTGxuKm9 .image-shape .label,#mermaid-svg-RAQQ5wosPTGxuKm9 .icon-shape .label{text-anchor:middle;}#mermaid-svg-RAQQ5wosPTGxuKm9 .node .katex path{fill:#000;stroke:#000;stroke-width:1px;}#mermaid-svg-RAQQ5wosPTGxuKm9 .rough-node .label,#mermaid-svg-RAQQ5wosPTGxuKm9 .node .label,#mermaid-svg-RAQQ5wosPTGxuKm9 .image-shape .label,#mermaid-svg-RAQQ5wosPTGxuKm9 .icon-shape .label{text-align:center;}#mermaid-svg-RAQQ5wosPTGxuKm9 .node.clickable{cursor:pointer;}#mermaid-svg-RAQQ5wosPTGxuKm9 .root .anchor path{fill:#333333!important;stroke-width:0;stroke:#333333;}#mermaid-svg-RAQQ5wosPTGxuKm9 .arrowheadPath{fill:#333333;}#mermaid-svg-RAQQ5wosPTGxuKm9 .edgePath .path{stroke:#333333;stroke-width:2.0px;}#mermaid-svg-RAQQ5wosPTGxuKm9 .flowchart-link{stroke:#333333;fill:none;}#mermaid-svg-RAQQ5wosPTGxuKm9 .edgeLabel{background-color:rgba(232,232,232, 0.8);text-align:center;}#mermaid-svg-RAQQ5wosPTGxuKm9 .edgeLabel p{background-color:rgba(232,232,232, 0.8);}#mermaid-svg-RAQQ5wosPTGxuKm9 .edgeLabel rect{opacity:0.5;background-color:rgba(232,232,232, 0.8);fill:rgba(232,232,232, 0.8);}#mermaid-svg-RAQQ5wosPTGxuKm9 .labelBkg{background-color:rgba(232, 232, 232, 0.5);}#mermaid-svg-RAQQ5wosPTGxuKm9 .cluster rect{fill:#ffffde;stroke:#aaaa33;stroke-width:1px;}#mermaid-svg-RAQQ5wosPTGxuKm9 .cluster text{fill:#333;}#mermaid-svg-RAQQ5wosPTGxuKm9 .cluster span{color:#333;}#mermaid-svg-RAQQ5wosPTGxuKm9 div.mermaidTooltip{position:absolute;text-align:center;max-width:200px;padding:2px;font-family:"trebuchet ms",verdana,arial,sans-serif;font-size:12px;background:hsl(80, 100%, 96.2745098039%);border:1px solid #aaaa33;border-radius:2px;pointer-events:none;z-index:100;}#mermaid-svg-RAQQ5wosPTGxuKm9 .flowchartTitleText{text-anchor:middle;font-size:18px;fill:#333;}#mermaid-svg-RAQQ5wosPTGxuKm9 rect.text{fill:none;stroke-width:0;}#mermaid-svg-RAQQ5wosPTGxuKm9 .icon-shape,#mermaid-svg-RAQQ5wosPTGxuKm9 .image-shape{background-color:rgba(232,232,232, 0.8);text-align:center;}#mermaid-svg-RAQQ5wosPTGxuKm9 .icon-shape p,#mermaid-svg-RAQQ5wosPTGxuKm9 .image-shape p{background-color:rgba(232,232,232, 0.8);padding:2px;}#mermaid-svg-RAQQ5wosPTGxuKm9 .icon-shape .label rect,#mermaid-svg-RAQQ5wosPTGxuKm9 .image-shape .label rect{opacity:0.5;background-color:rgba(232,232,232, 0.8);fill:rgba(232,232,232, 0.8);}#mermaid-svg-RAQQ5wosPTGxuKm9 .label-icon{display:inline-block;height:1em;overflow:visible;vertical-align:-0.125em;}#mermaid-svg-RAQQ5wosPTGxuKm9 .node .label-icon path{fill:currentColor;stroke:revert;stroke-width:revert;}#mermaid-svg-RAQQ5wosPTGxuKm9 :root{--mermaid-font-family:"trebuchet ms",verdana,arial,sans-serif;} 数据湖架构
Iceberg
Delta Lake
Apache Hudi
快照模型
隐藏分区
开放规范
ACID事务
Spark集成
流批一体
增量处理
Upsert
CDC支持

数据湖组件	核心能力	适用场景	工业推荐度
Apache Iceberg	开放标准、隐藏分区、时间旅行	多引擎统一、开放架构	⭐⭐⭐⭐⭐
Delta Lake	ACID事务、Spark深度集成	Databricks环境、性能优先	⭐⭐⭐⭐
Apache Hudi	增量处理、CDC同步	实时数仓、变更数据捕获	⭐⭐⭐⭐

下期预告

第14期我们将深入探讨《Hadoop集群部署》，从规划到实践，详细讲解物理机、虚拟机、Kubernetes等多种部署方式的选型与配置。敬请期待！

作者：高炉炼铁智能化技术研究者，专注钢铁冶金与人工智能交叉领域。

👍 如果觉得有帮助，请点赞、收藏、转发！

版权归作者所有，未经许可请勿抄袭，套用，商用(或其它具有利益性行为) 。

🔔 关注专栏，不错过后续精彩内容！