如何只保留 KSQL 中的最新窗口？

Question

我想监控过去 30 天内发生的事件计数。所以我在KSQL中使用了HOPPING WINDOW（大小30天，提前1天），但它会为每个组项生成总共30个窗口，我如何编写ksql脚本只保留最新的窗口？谢谢！

CREATE TABLE FEATURE1_INDIVID_APPLY_REJECTED_COUNT_HOPPING_30_DAYS_TBL AS \
SELECT INDIVID_REGION, WindowStart() AS WINSTART, COUNT(*) AS APPLY_REJECTED_COUNT \
FROM FEATURE1_INDIVID_APPLY_REJECTED_HACKTIME_STM  \
     WINDOW HOPPING (SIZE 30 DAYS, ADVANCE BY 1 DAYS) \
WHERE INDIVID_REGION IS NOT NULL \
GROUP BY INDIVID_REGION;

Answer 1

您需要一个 HAVING 过滤器来过滤掉任何具有历史性 WINDOWEND 的窗口，例如

CREATE TABLE FEATURE1_INDIVID_APPLY_REJECTED_COUNT_HOPPING_30_DAYS_TBL AS \
SELECT INDIVID_REGION, WindowStart() AS WINSTART, COUNT(*) AS APPLY_REJECTED_COUNT \
FROM FEATURE1_INDIVID_APPLY_REJECTED_HACKTIME_STM  \
     WINDOW HOPPING (SIZE 30 DAYS, ADVANCE BY 1 DAYS) \
WHERE INDIVID_REGION IS NOT NULL \
GROUP BY INDIVID_REGION
HAVING WINDOWEND < UNIX_TIMESTAMP();

（ksqlDB 0.11 版本的正确语法）

不幸的是，这样的HAVING 子句需要修复Github issue 4397。一旦 ksqlDB 在HAVING 子句中支持WINDOWEND，你就可以开始使用了！