在 spark/java 上按 desc 限制按顺序加入数据帧

Question

我正在使用以下代码：

Dataset <Row> dataframee =  df1.as("a").join(df2.as("b"),
            df2.col("id_device").equalTo(df1.col("ID_device_previous")).
                    and(df2.col("id_vehicule").equalTo(df1.col("ID_vehicule_previous"))).
                and(df2.col("tracking_time").lt(df1.col("date_track_previous")))
            ,"left").selectExpr("a.*", "b.ID_tracking as ID_pprevious", "b.km as KM_pprevious","b.tracking_time as tracking_time_pprevious","b.speed as speed_pprevious");

我从 df2 数据帧中获得多行的 df1 数据帧连接。

但我想要的是在相同的条件下将df1数据帧与df2数据帧连接起来， df2.col("tracking_time") desc limit(0,1)排序

编辑

我尝试了以下代码，但它不起作用。

df1.registerTempTable("data");
df2.createOrReplaceTempView("tdays");
Dataset<Row> d_f = sparkSession.sql("select a.*  from data as a  LEFT JOIN (select  b.tracking_time from tdays as b where  b.id_device = a.ID_device_previous and  b.id_vehicule = a.ID_vehicule_previous  and b.tracking_time < a.date_track_previous order by b.tracking_time desc limit 1 )");

 
  

我需要你的帮助

Answer 1

你可以用我知道的多种方式来做到这一点

1. 您可以在加入的数据框 DF 上执行 dropDuplicates。

   val finalDF = dataframee.dropDuplicates("") // 您希望在最终输出中不同/唯一的指定列

（或者）

2. spark-sql

    import spark.sql.implicits._ df1.createOrReplaceTempViews("table1") df2.createOrReplaceTempViews("table2") spark.sql("join query with groupBy distinct columns").select(df("*"))