Flink-Table-&-SQL

简介

Apache Flink具有两个关系API - 表API和SQL - 用于统一流和批处理。Table API是Scala和Java的语言集成查询API，允许以非常直观的方式组合来自关系运算符的查询，Table API和SQL接口彼此紧密集成，以及Flink的DataStream和DataSet API。您可以轻松地在基于API构建的所有API和库之间切换。例如，您可以使用CEP库从DataStream中提取模式，然后使用Table API分析模式，或者可以在预处理上运行Gelly图算法之前使用SQL查询扫描，过滤和聚合批处理表数据。

Flink SQL的编程模型

创建一个TableEnvironment

TableEnvironment是Table API和SQL集成的核心概念，它主要负责:
1、在内部目录中注册一个Table
2、注册一个外部目录
3、执行SQL查询
4、注册一个用户自定义函数(标量、表及聚合)
5、将DataStream或者DataSet转换成Table
6、持有ExecutionEnvironment或者StreamExecutionEnvironment的引用
一个Table总是会绑定到一个指定的TableEnvironment中，相同的查询不同的TableEnvironment是无法通过join、union合并在一起。
TableEnvironment有一个在内部通过表名组织起来的表目录，Table API或者SQL查询可以访问注册在目录中的表，并通过名称来引用它们。

在目录中注册表

TableEnvironment允许通过各种源来注册一个表:

1、一个已存在的Table对象，通常是Table API或者SQL查询的结果

 Table projTable = tableEnv.scan("X").select(...); 

2、TableSource，可以访问外部数据如文件、数据库或者消息系统

 TableSource csvSource = new CsvTableSource("/path/to/file", ...); 

3、DataStream或者DataSet程序中的DataStream或者DataSet

 //将DataSet转换为Table Table table= tableEnv.fromDataSet(tableset); 

注册TableSink

注册TableSink可用于将 Table API或SQL查询的结果发送到外部存储系统，例如数据库，键值存储，消息队列或文件系统（在不同的编码中，例如，CSV，Apache [Parquet] ，Avro，ORC]，......）:

TableSink csvSink = new CsvTableSink("/path/to/file", ...); 

 2、 String[] fieldNames = {"a", "b", "c"}; TypeInformation[] fieldTypes = {Types.INT, Types.STRING, Types.LONG}; tableEnv.registerTableSink("CsvSinkTable", fieldNames, fieldTypes, csvSink);

实战案例一

基于Flink SQL的WordCount:

public class WordCountSQL { public static void main(String[] args) throws Exception{ ExecutionEnvironment env = ExecutionEnvironment.getExecutionEnvironment(); BatchTableEnvironment tEnv = TableEnvironment.getTableEnvironment(env); List list = new ArrayList(); String wordsStr = "Hello Flink Hello TOM"; String[] words = wordsStr.split("\\W+"); for(String word : words){ WC wc = new WC(word, 1); list.add(wc); } DataSet<WC> input = env.fromCollection(list); tEnv.registerDataSet("WordCount", input, "word, frequency"); Table table = tEnv.sqlQuery( "SELECT word, SUM(frequency) as frequency FROM WordCount GROUP BY word"); DataSet<WC> result = tEnv.toDataSet(table, WC.class); result.print(); }//main public static class WC { public String word;//hello public long frequency;//1 // public constructor to make it a Flink POJO public WC() {} public WC(String word, long frequency) { this.word = word; this.frequency = frequency; } @Override public String toString() { return "WC " + word + " " + frequency; } } }

输出如下：

WC TOM 1 WC Hello 2 WC Flink 1

实战案例二

本例稍微复杂，首先读取一个文件中的内容进行统计，并写入到另外一个文件中：

public class SQLTest { public static void main(String[] args) throws Exception{ ExecutionEnvironment env = ExecutionEnvironment.getExecutionEnvironment(); BatchTableEnvironment tableEnv = BatchTableEnvironment.getTableEnvironment(env); env.setParallelism(1); DataSource<String> input = env.readTextFile("test.txt"); input.print(); //转换成dataset DataSet<Orders> topInput = input.map(new MapFunction<String, Orders>() { @Override public Orders map(String s) throws Exception { String[] splits = s.split(" "); return new Orders(Integer.valueOf(splits[0]), String.valueOf(splits[1]),String.valueOf(splits[2]), Double.valueOf(splits[3])); } }); //将DataSet转换为Table Table order = tableEnv.fromDataSet(topInput); //orders表名 tableEnv.registerTable("Orders",order); Table tapiResult = tableEnv.scan("Orders").select("name"); tapiResult.printSchema(); Table sqlQuery = tableEnv.sqlQuery("select name, sum(price) as total from Orders group by name order by total desc"); //转换回dataset DataSet<Result> result = tableEnv.toDataSet(sqlQuery, Result.class); //将dataset map成tuple输出 /*result.map(new MapFunction<Result, Tuple2<String,Double>>() { @Override public Tuple2<String, Double> map(Result result) throws Exception { String name = result.name; Double total = result.total; return Tuple2.of(name,total); } }).print();*/ TableSink sink = new CsvTableSink("SQLTEST.txt", "|"); //writeToSink /*sqlQuery.writeToSink(sink); env.execute();*/ String[] fieldNames = {"name", "total"}; TypeInformation[] fieldTypes = {Types.STRING, Types.DOUBLE}; tableEnv.registerTableSink("SQLTEST", fieldNames, fieldTypes, sink); sqlQuery.insertInto("SQLTEST"); env.execute(); } /** * 源数据的映射类 */ public static class Orders { /** * 序号，姓名，书名，价格 */ public Integer id; public String name; public String book; public Double price; public Orders() { super(); } public Orders(Integer id, String name, String book, Double price) { this.id = id; this.name = name; this.book = book; this.price = price; } } /** * 统计结果对应的类 */ public static class Result { public String name; public Double total; public Result() {} } }//