First, the windowing function
1 Overview
Spark 1.4 after .x version for Spark SQL and DataFrame introduces windowing function, such as the most classic, most popular, row_number (),
allows us to achieve a logical grouping of taking topn.
Case: statistics for each kind of product sales in the top 3
Second, the group taking topN Case
1, java achieve
package cn.spark.study.sql; import org.apache.spark.SparkConf; import org.apache.spark.api.java.JavaSparkContext; import org.apache.spark.sql.DataFrame; import org.apache.spark.sql.hive.HiveContext; /** * row_number()开窗函数实战 * @author Administrator * */ public class RowNumberWindowFunction { @SuppressWarnings("deprecation") public static void main(String[] args) { SparkConf conf = new SparkConf() .setAppName("RowNumberWindowFunction"); JavaSparkContext sc = new JavaSparkContext(conf); HiveContext hiveContext = new HiveContext(sc.sc()); // 创建销售额表,sales表 hiveContext.sql("DROP TABLE IF EXISTS sales"); hiveContext.sql("CREATE TABLE IF NOT EXISTS sales (" + "product STRING," + "category STRING," + "revenue BIGINT)"); hiveContext.sql("LOAD DATA " + "LOCAL INPATH '/usr/local/spark-study/resources/sales.txt' " + "INTO TABLE sales"); // start writing our statistical logic, use row_number () windowing function // first explain the role row_number () windowing function // in fact, is to give each packet of data, according to their sort order, marked a line number within the packet // for example, a packet date = 20151001, there are three data, 1122,1121,1124, // so use, each line ROW_NUMBER () is a windowing function on this packet, three lines sequentially get the line number within a group // row number incremented from one, for example. 3 2,1124 1,1121 1122 @ from sub-queries: query result as an inner layer of a temporary table for the outer query; DataFrame top3SalesDF = hiveContext.sql ( "" + "the SELECT Product, category, Revenue" + "the FROM (" + "the SELECT" + "Product," + "category," + "revenue," // row_number () syntax description windowing function // First of all you can, when the SELECT query, use row_number () function // Second, row_number () function to keep up behind OVER keyword // and then in parentheses, is the PARTITION BY , that are grouped according to which field // followed by the group can be sorted bY with the ORDER // then row_number () can give the line within each group, a group of experts number // tmp_sales: alias sub-queries + "ROW_NUMBER () the OVER (the PARTITION BY the ORDER BY category Revenue DESC) Rank" + "the FROM Sales" + ") tmp_sales" + "the WHERE Rank <=. 3" ); // the data before each rank 3, saved to a table hiveContext.sql("DROP TABLE IF EXISTS top3_sales"); top3SalesDF.saveAsTable("top3_sales"); sc.close(); } }