今天尝试一下将昨天的数据操作建立的表换成分区表，参考Delta Lake Up and Running做法用分区表的方式来更新数据。还要比较一下分区表的查询与非分区表的查询，结果显示分区表的查询速度要比非分区表要快。直接上代码：

import io.delta.tables.DeltaTable;
import org.apache.spark.sql.*;
import org.apache.spark.sql.types.DataTypes;

import java.text.SimpleDateFormat;
import java.util.ArrayList;
import java.util.Date;
import java.util.HashMap;
import java.util.List;

public class DeltaLakePartitionCURD {

    //将字符串转换成java.sql.Timestamp
    public static java.sql.Timestamp strToSqlDate(String strDate, String dateFormat) {
        SimpleDateFormat sf = new SimpleDateFormat(dateFormat);
        Date date = null;
        try {
            date = sf.parse(strDate);
        } catch (Exception e) {
            e.printStackTrace();
        }
        java.sql.Timestamp dateSQL = new java.sql.Timestamp(date.getTime());
        return dateSQL;
    }

    public static void main(String[] args) {

        SparkSession spark = SparkSession.builder()
                .master("local[*]")
                .appName("delta_lake")
                .config("spark.sql.extensions", "io.delta.sql.DeltaSparkSessionExtension")
                .config("spark.databricks.delta.autoCompact.enabled", "true")
                .config("spark.sql.catalog.spark_catalog", "org.apache.spark.sql.delta.catalog.DeltaCatalog")
                .getOrCreate();


        SimpleDateFormat sdf=new SimpleDateFormat("yyyy-MM-dd HH:mm:ss.SSS");
        String savePath="D:\\bigdata\\detla-lake-with-java\\YellowTaxiPartitioned";
        String csvPath="D:\\bookcode\\delta-lake-up-and-running-main\\data\\YellowTaxisLargeAppend.csv";
        String tableName = "taxidb.YellowTaxiPartitioned";


        String savePath2="D:\\bigdata\\detla-lake-with-java\\YellowTaxi";


        spark.sql("CREATE DATABASE IF NOT EXISTS taxidb");

        //定义表
        DeltaTable.createIfNotExists(spark)
                .tableName(tableName)
                .addColumn("RideId","INT")
                .addColumn("VendorId","INT")
                .addColumn("PickupTime","TIMESTAMP")
                .addColumn("DropTime","TIMESTAMP")
                .partitionedBy("VendorId")
                .location(savePath)
                .execute();


        //加载csv数据并导入delta表
        var df=spark.read().format("delta").table(tableName);
        var schema=df.schema();
        System.out.println(schema.simpleString());
        var df_for_append=spark.read().option("header","true").schema(schema).csv(csvPath);
        System.out.println("记录总行数："+df_for_append.count());
        System.out.println("导入数据,开始时间"+  sdf.format(new Date()));
        df_for_append.write().format("delta").mode(SaveMode.Overwrite).saveAsTable(tableName);
        System.out.println("导入数据，结束时间" + sdf.format(new Date()));


        DeltaTable deltaTable = DeltaTable.forName(spark,tableName);
        deltaTable.optimize().executeZOrderBy("RideId");

        //插入数据
        List<Row> list = new ArrayList<Row>();
        list.add(RowFactory.create(-1,-1,strToSqlDate("2023-01-01 10:00:00","yyyy-MM-dd HH:mm:ss"),strToSqlDate("2023-01-01 10:00:00","yyyy-MM-dd HH:mm:ss")));
        var yellowTaxipDF=spark.createDataFrame(list,schema);//建立需要新增数据并转换成dataframe
        System.out.println("插入数据,开始时间"+  sdf.format(new Date()));
        yellowTaxipDF.write().format("delta").mode(SaveMode.Append).saveAsTable(tableName);
        System.out.println("插入数据,结束时间"+  sdf.format(new Date()));
        System.out.println("插入后数据");
        deltaTable.toDF().select("*").where("RideId=-1").show(false);





        System.out.println("更新数据,开始时间"+  sdf.format(new Date()));
        spark.read().format("delta").load(savePath).where("VendorId==-1 and RideId==-1")
                .withColumn("PickupTime", functions.lit("2023-11-01 10:00:00").cast(DataTypes.TimestampType))
                .write().format("delta")
                .option("replaceWhere","VendorId==-1")
                .mode(SaveMode.Overwrite).saveAsTable(tableName);
        System.out.println("更新数据,结束时间"+  sdf.format(new Date()));
        System.out.println("更新后数据");
        deltaTable.toDF().select("*").where("RideId=-1").show(false);


        //更新数据
        System.out.println("更新前数据");
        deltaTable.toDF().select("*").where("RideId=-1").show(false);
        System.out.println("更新数据,开始时间"+  sdf.format(new Date()));
        deltaTable.update(
                functions.col("RideId").equalTo("-1"),
                new HashMap<String, Column>() {{
                    put("PickupTime", functions.lit("2023-01-01 10:00:00").cast(DataTypes.TimestampType));
                }}
        );
        System.out.println("更新数据,结束时间"+  sdf.format(new Date()));
        System.out.println("更新后数据");
        deltaTable.toDF().select("*").where("RideId=-1").show(false);


        //查询数据
        System.out.println("不分区表查询数据,开始时间"+  sdf.format(new Date()));
        spark.read().format("delta").load(savePath2).where("VendorId==4 and RideId==859744").show(false);
        System.out.println("不分区表查询数据，结束时间" + sdf.format(new Date()));


        System.out.println("分区表查询数据,开始时间"+  sdf.format(new Date()));
        spark.read().format("delta").load(savePath).where("VendorId==4 and RideId==859744").show(false);
        System.out.println("分区表查询数据，结束时间" + sdf.format(new Date()));




    }
}

最终运行结果：

从最后一张图红框可以看到比较结果，分区表查询效率要高一点。