源文件内容示例:



http://bigdata.beiwang.cn/laoli
http://bigdata.beiwang.cn/laoli
http://bigdata.beiwang.cn/haiyuan
http://bigdata.beiwang.cn/haiyuan



 

实现代码:



object SparkSqlDemo11 {
  /**
    * 使用开窗函数,计算TopN
    * @param args
    */
  def main(args: Array[String]): Unit = {

    val session = SparkSession.builder()
      .appName(this.getClass.getSimpleName)
      .master("local")
      .getOrCreate()

    import session.implicits._

    //原数据:http://bigdata.beiwang.cn/laoli
    val sourceData = session.read.textFile("E:\\北网学习\\K_第十一个月_Spark 2(2019.8)\\8.5\\teacher.log")

    val df = sourceData.map(line => {
      val index = line.lastIndexOf("/")
      val t_name = line.substring(index + 1)

      val url = new URL(line.substring(0, index))
      val subject = url.getHost.split("\\.")(0)

      (subject, t_name)
    }).toDF("subject", "t_name")



操作01:得到所有专业下所有老师的访问数:



df.createTempView("temp")

    //获得所有学科下老师的访问量:
    val middleData: DataFrame = session.sql("select subject,t_name,count(*) cnts from temp group by subject,t_name")

    //middleData.show()


+-------+--------+----+
|subject|  t_name|cnts|
+-------+--------+----+
|bigdata|   laoli|   2|
|bigdata| haiyuan|  15|
| javaee|chenchan|   6|
|    php|  laoliu|   1|
|    php|   laoli|   3|
| javaee|  laoshi|   9|
|bigdata|  lichen|   6|
+-------+--------+----+


操作02:row_number() over()【按照老师的访问数,降序开窗】



//再将中间值middleData注册成一张表
middleData.createTempView("middleTemp")

//执行第二部查询,使用row_number()开窗函数,对所有的老师的访问数进行排序并添加编号
//开窗后生成的编号列 rn 是一个伪列,只能用于展示,不能用于查询
//row_number() over() 函数是按照某种规则对数据进行编号,需要我们在over()中指定一个排序规则,无规则将会报错
//此处是按照cnts列降序开窗
session.sql(
  """
    |select subject,t_name,cnts,row_number() over(order by cnts desc) rn from middleTemp
  """.stripMargin).show()



+-------+--------+----+---+
|subject|  t_name|cnts| rn|
+-------+--------+----+---+
|bigdata| haiyuan|  15|  1|
| javaee|  laoshi|   9|  2|
| javaee|chenchan|   6|  3|
|bigdata|  lichen|   6|  4|
|    php|   laoli|   3|  5|
|bigdata|   laoli|   2|  6|
|    php|  laoliu|   1|  7|
+-------+--------+----+---+



♈ 注意:over()内必须指定开窗规则,否则会抛出解析异常:



session.sql(
  """
    |select subject,t_name,cnts,row_number() over() rn from middleTemp
  """.stripMargin).show()



Exception in thread "main" org.apache.spark.sql.AnalysisException: Window function row_number() requires window to be ordered, please add ORDER BY clause. For example SELECT row_number()(value_expr) OVER (PARTITION BY window_partition ORDER BY window_ordering) from table;
	at org.apache.spark.sql.catalyst.analysis.CheckAnalysis$class.failAnalysis(CheckAnalysis.scala:39)
	at org.apache.spark.sql.catalyst.analysis.Analyzer.failAnalysis(Analyzer.scala:91)
	at org.apache.spark.sql.catalyst.analysis.Analyzer$ResolveWindowOrder$$anonfun$apply$31$$anonfun$applyOrElse$12.applyOrElse(Analyzer.scala:2173)
	at org.apache.spark.sql.catalyst.analysis.Analyzer$ResolveWindowOrder$$anonfun$apply$31$$anonfun$applyOrElse$12.applyOrElse(Analyzer.scala:2171)



操作03:row_number() over(partition by.. 【根据学科进行分区后为每个分区开窗】



//根据学科进行分区后为每个分区开窗
session.sql(
  """
    |select subject,t_name,cnts,row_number() over(partition by subject order by cnts desc) rn from middleTemp
  """.stripMargin).show()


+-------+--------+----+---+
|subject|  t_name|cnts| rn|
+-------+--------+----+---+
| javaee|  laoshi|   9|  1|
| javaee|chenchan|   6|  2|
|bigdata| haiyuan|  15|  1|
|bigdata|  lichen|   6|  2|
|bigdata|   laoli|   2|  3|
|    php|   laoli|   3|  1|
|    php|  laoliu|   1|  2|
+-------+--------+----+---+


♎ 注意:开窗生成的列是伪列,不能用于实际操作:



//开窗形成的列是伪列,不能用于实际操作
session.sql(
  """
    |select subject,t_name,cnts,row_number() over(partition by subject order by cnts desc) rn from middleTemp
    |where rn <=2
  """.stripMargin).show()



spark dataframe 开窗聚合计数 spark开窗函数原理_sql

 

操作04:伪列的使用:

由于开窗形成的伪列不能被直接用于查询,那么我们可以将整个开窗语句的操作作为一个子查询使用,那么开窗语句的结果集对于父查询来说就是一张完整的表,这时候伪列就是一个有效的列,可以用于查询:



//开窗生成的伪列不能用于直接查询,但是我们可以将开窗语句的结果集作为一张表或者说一个子查询,这时候伪列就是一个有效的列,可以进行再次嵌套查询,
session.sql(
  """
    |select * from (
    |select subject,t_name,cnts,row_number() over(partition by subject order by cnts desc) rn from middleTemp
    |) where rn <= 2
  """.stripMargin).show()



+-------+--------+----+---+
|subject|  t_name|cnts| rn|
+-------+--------+----+---+
| javaee|  laoshi|   9|  1|
| javaee|chenchan|   6|  2|
|bigdata| haiyuan|  15|  1|
|bigdata|  lichen|   6|  2|
|    php|   laoli|   3|  1|
|    php|  laoliu|   1|  2|
+-------+--------+----+---+



  

操作05:【开窗嵌套开窗】rank() over() 函数

在row_number() over() 分区+开窗的基础上,再次进行rank() over() 按照cnts进行全部数据的开窗



//开窗嵌套开窗:
//rank() over() 函数
session.sql(
  """
    |select t.*,rank() over(order by cnts desc) rn1 from (
    |select subject,t_name,cnts,row_number() over(partition by subject order by cnts desc) rn from middleTemp
    |) t 
    |where rn <= 2
  """.stripMargin).show()



+-------+--------+----+---+---+
|subject|  t_name|cnts| rn|rn1|
+-------+--------+----+---+---+
|bigdata| haiyuan|  15|  1|  1|
| javaee|  laoshi|   9|  1|  2|
| javaee|chenchan|   6|  2|  3|
|bigdata|  lichen|   6|  2|  3|
|    php|   laoli|   3|  1|  5|
|    php|  laoliu|   1|  2|  6|
+-------+--------+----+---+---+


  

操作06:dense_rank() over() 函数 【三个开窗函数的业务对比】:



//dense_rank() over() 函数
//三个开窗函数的业务对比:
session.sql(
  """
    |select t.*,rank() over(order by cnts desc) rank,
    |row_number() over(order by cnts desc) row_n,
    |dense_rank() over(order by cnts desc) dense_n
    |from (
    |select subject,t_name,cnts,row_number() over(partition by subject order by cnts desc) row_n_par from middleTemp
    |) t
    |where row_n_par <= 2
  """.stripMargin).show()


+-------+--------+----+---------+----+-----+-------+
|subject|  t_name|cnts|row_n_par|rank|row_n|dense_n|
+-------+--------+----+---------+----+-----+-------+
|bigdata| haiyuan|  15|        1|   1|    1|      1|
| javaee|  laoshi|   9|        1|   2|    2|      2|
| javaee|chenchan|   6|        2|   3|    3|      3|
|bigdata|  lichen|   6|        2|   3|    4|      3|
|    php|   laoli|   3|        1|   5|    5|      4|
|    php|  laoliu|   1|        2|   6|    6|      5|
+-------+--------+----+---------+----+-----+-------+


spark dataframe 开窗聚合计数 spark开窗函数原理_php_02

操作07:整合为一句SQL完成:



//合并两个SQL语句:
session.sql(
  """
    |select t.*,rank() over(order by cnts desc) rank,
    |row_number() over(order by cnts desc) row_n,
    |dense_rank() over(order by cnts desc) dense_n
    |from
    |(select subject,t_name,cnts,row_number() over(partition by subject order by cnts desc) row_n_par from
    |(select subject,t_name,count(*) cnts from temp group by subject,t_name)) t
    |where row_n_par <= 2
  """.stripMargin).show()



+-------+--------+----+---------+----+-----+-------+
|subject|  t_name|cnts|row_n_par|rank|row_n|dense_n|
+-------+--------+----+---------+----+-----+-------+
|bigdata| haiyuan|  15|        1|   1|    1|      1|
| javaee|  laoshi|   9|        1|   2|    2|      2|
| javaee|chenchan|   6|        2|   3|    3|      3|
|bigdata|  lichen|   6|        2|   3|    4|      3|
|    php|   laoli|   3|        1|   5|    5|      4|
|    php|  laoliu|   1|        2|   6|    6|      5|
+-------+--------+----+---------+----+-----+-------+