怎么使用 Flink 向 Apache Doris 表中写 Bitmap 类型的数据

Bitmap是一种经典的数据结构，用于高效地对大量的二进制数据进行压缩存储和快速查询。Doris支持bitmap数据类型，在Flink计算场景中，可以结合Flink doris Connector对bitmap数据做计算。

社区里很多小伙伴在是Doris Flink Connector的时候，不知道怎么写Bitmap类型的数据，本文将介绍如何使用 Flink Doris Connector 如何将 bitmap 数据写入 Doris 中。

前置准备

Doris2.0.1的环境

Flink1.16，同时将 Doris Flink Connector的Jar包放在<FLINK_HOME>/lib 下面。

创建Doris表

CREATE TABLE `page_view_bitmap` (
`dt` int,
`page` varchar(256),
`user_id` bitmap bitmap_union
)
AGGREGATE KEY(`dt`, page)
DISTRIBUTED BY HASH(`dt`) BUCKETS 1
PROPERTIES (
"replication_num" = "1"
)

写入Bitmap数据

这里模拟Flink读取MySQL数据写入Doris，同时将user_id存储到bitmap中。

模拟数据

创建MySQL表

CREATE TABLE `page_view` (
 `id` int NOT NULL,
 `dt` int,
 `page` varchar(256),
 `user_id` int,
 PRIMARY KEY (`id`)
);

#模拟数据
INSERT INTO `test`.`page_view` (`id`, `dt`, `page`, `user_id`) VALUES (1, 20230921, 'home', 1001);
INSERT INTO `test`.`page_view` (`id`, `dt`, `page`, `user_id`) VALUES (2, 20230921, 'home', 1002);
INSERT INTO `test`.`page_view` (`id`, `dt`, `page`, `user_id`) VALUES (3, 20230921, 'search', 1003);
INSERT INTO `test`.`page_view` (`id`, `dt`, `page`, `user_id`) VALUES (4, 20230922, 'mine', 1001);
INSERT INTO `test`.`page_view` (`id`, `dt`, `page`, `user_id`) VALUES (5, 20230922, 'home', 1002);
FlinkSQL写入Bitmap
#使用JDBC读取mysql数据
CREATE TABLE page_view (
   `dt` int,
   `page` string,
   `user_id` int
) WITH (
   'connector' = 'jdbc',
   'url' = 'jdbc:mysql://127.0.0.1:3306/test',
   'table-name' = 'page_view',
   'username' = 'root',
   'password' = '123456'
);

doris connector写入数据

CREATE TABLE page_view_bitmap (
dt int,
page string,
user_id int
)
WITH (
 'connector' = 'doris',
 'fenodes' = '127.0.0.1:8030',
 'table.identifier' = 'test.page_view_bitmap',
 'username' = 'root',
 'password' = '',
 'sink.label-prefix' = 'doris_label1',
 'sink.properties.columns' = 'dt,page,user_id,user_id=to_bitmap(user_id)'
);

insert into page_view_bitmap select * from page_view

我们知道 Doris Flink Connector Sink 底层是基于 Doris Stream Load 来实现的，同样 Stream load 在 Connector 里也是一样适用，我们将这个参数封装在了 ：sink.properties 参数里，

这里我们可以看到上面这个例子里我们在是 With 属性里加入了我们 Columns 参数，这里我们配置了列的转换操作，将 user_id 通过 to_bitmap 函数进行转换，并导入到 Doris 表里。

查询结果

mysql> select dt,page,bitmap_to_string(user_id) from `test`.`page_view_bitmap`;
+----------+--------+---------------------------+
| dt       | page   | bitmap_to_string(user_id) |
+----------+--------+---------------------------+
| 20230921 | home   | 1001,1002                 |
| 20230921 | search | 1003                      |
| 20230922 | home   | 1002                      |
| 20230922 | mine   | 1001                      |
+----------+--------+---------------------------+
4 rows in set (0.00 sec)

Flink DataStream

使用 DataStream API 模拟数据写入刚才的表中。

DataStream API 对 Bitmap 的操作也是和上面 SQL 操作的方式一样。

public static void main(String[] args) throws Exception {
       StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();
       env.setParallelism(1);
       env.setRuntimeMode(RuntimeExecutionMode.BATCH);

       DorisSink.Builder<String> builder = DorisSink.builder();
       final DorisReadOptions.Builder readOptionBuilder = DorisReadOptions.builder();
       Properties properties = new Properties();
       properties.setProperty("column_separator", ",");
       properties.setProperty("format", "csv");
       properties.setProperty("columns", "dt,page,user_id,user_id=to_bitmap(user_id)");

       DorisOptions.Builder dorisBuilder = DorisOptions.builder();
       dorisBuilder.setFenodes("127.0.0.1:8030")
              .setTableIdentifier("test.page_view_bitmap")
              .setUsername("root")
              .setPassword("");
       DorisExecutionOptions.Builder executionBuilder = DorisExecutionOptions.builder();
       executionBuilder.setLabelPrefix("doris_label")
              .setStreamLoadProp(properties)
              .setDeletable(false);

       builder.setDorisReadOptions(readOptionBuilder.build())
              .setDorisExecutionOptions(executionBuilder.build())
              .setSerializer(new SimpleStringSerializer())
              .setDorisOptions(dorisBuilder.build());

       //mock data
       DataStreamSource<String> stringDataStreamSource = env.fromCollection(
               Arrays.asList("20230921,home,1003", "20230921,search,1001", "20230923,home,1001"));
       stringDataStreamSource.sinkTo(builder.build());
       env.execute("doris bitmap write");
  }

查询结果

mysql> select dt,page,bitmap_to_string(user_id) from `test`.`page_view_bitmap`;
+----------+--------+---------------------------+
| dt       | page   | bitmap_to_string(user_id) |
+----------+--------+---------------------------+
| 20230921 | home   | 1001,1002,1003            |
| 20230921 | search | 1001,1003                 |
| 20230922 | home   | 1002                      |
| 20230922 | mine   | 1001                      |
| 20230923 | home   | 1001                      |
+----------+--------+---------------------------+
5 rows in set (0.00 sec)