Flink SQL Client中的Chained (Event) Time Windows

向右看齐 2022-10-27 13:44 132阅读 0赞

Flink SQL Cookbook中有[Chained (Event) Time Windows][Chained _Event_ Time Windows]

该md文件举了两个例子，

前面一个例子是Chained Windows

后面一个例子是Non-Chained Windows

该md文件的前半部分图示如下:

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2FwcGxleXVjaGk_size_16_color_FFFFFF_t_70][]

上图中存在两个server\_logs\_window\_5m

①Chained Windows

②Non-Chained Windows

所以②比①更加有计算上的效率，但是耦合性也更强。

上图图示中，**三条红线就是所谓的Chained (Event) Time Windows**

这个概念出自Ververcica公司的flink sql cookbook

Chained (Event) Time Windows的SQL距离

**Chained (Event) Time Windows的SQL举例(对应上图中的3个红色箭头):**

<table> 
 <tbody> 
 <tr> 
 <td style="width:399px;">Chained (Event) Time Windows的数据流向</td> 
 <td style="width:306px;">具体对应的SQL</td> 
 </tr> 
 <tr> 
 <td style="width:399px;">server_logs_window_1m-&gt;server_logs_window_5m</td> 
 <td style="width:306px;">CREATE TEMPORARY VIEW server_logs_window_5m AS&nbsp; SELECT&nbsp; &nbsp; TUMBLE_START(window_end, INTERVAL '5' MINUTE) AS window_start, &nbsp; TUMBLE_ROWTIME(window_end, INTERVAL '5' MINUTE) AS window_end, &nbsp; SUM(total_size) AS total_size, &nbsp; SUM(num_requests) AS num_requests FROM server_logs_window_1m GROUP BY&nbsp; &nbsp; TUMBLE(window_end, INTERVAL '5' MINUTE);</td> 
 </tr> 
 <tr> 
 <td style="width:399px;">server_logs_window_1m-&gt;avg_request_size_1m</td> 
 <td style="width:306px;">INSERT INTO avg_request_size_1m SELECT &nbsp; window_start, &nbsp; window_end,&nbsp; &nbsp; total_size/num_requests AS avg_size FROM server_logs_window_1m;</td> 
 </tr> 
 <tr> 
 <td style="width:399px;">server_logs_window_5m-&gt;avg_request_size_5m</td> 
 <td style="width:306px;">INSERT INTO avg_request_size_5m SELECT &nbsp; window_start, &nbsp; window_end,&nbsp; &nbsp; total_size/num_requests AS avg_size FROM server_logs_window_5m; &nbsp;</td> 
 </tr> 
 </tbody> 
</table>

Chained Windows与Non-Chained Windows的SQL写法对比

<table> 
 <tbody> 
 <tr> 
 <td style="width:181px;">Windows类型</td> 
 <td style="width:546px;">具体对应的SQL</td> 
 </tr> 
 <tr> 
 <td style="width:181px;">Chained Windows</td> 
 <td style="width:546px;">CREATE TEMPORARY VIEW server_logs_window_5m AS&nbsp; SELECT&nbsp; &nbsp; TUMBLE_START(window_end, INTERVAL '5' MINUTE) AS window_start, &nbsp; TUMBLE_ROWTIME(window_end, INTERVAL '5' MINUTE) AS window_end, &nbsp; SUM(total_size) AS total_size, &nbsp; SUM(num_requests) AS num_requests FROM server_logs_window_1m GROUP BY&nbsp; &nbsp; TUMBLE(window_end, INTERVAL '5' MINUTE);</td> 
 </tr> 
 <tr> 
 <td style="width:181px;">Non-Chained Windows</td> 
 <td style="width:546px;">CREATE TEMPORARY VIEW server_logs_window_5m AS&nbsp; SELECT&nbsp; &nbsp; TUMBLE_START &nbsp;(log_time, INTERVAL '5' MINUTE) AS window_start, &nbsp; TUMBLE_ROWTIME(log_time, INTERVAL '5' MINUTE) AS window_end, &nbsp; SUM(size) AS total_size, &nbsp; COUNT(*) AS num_requests FROM server_logs GROUP BY&nbsp; &nbsp; TUMBLE(log_time, INTERVAL '5' MINUTE);</td> 
 </tr> 
 </tbody> 
</table>

[Chained _Event_ Time Windows]: https://github.com/ververica/flink-sql-cookbook/blob/master/aggregations-and-analytics/07/07_chained_windows.md
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2FwcGxleXVjaGk_size_16_color_FFFFFF_t_70]: /images/20221024/d4ad2b3edad74943976bb2857cfcaf72.png