SQL如何根据另一个表中的日期优化查询一个表中的两个日期窗口？答案

【问题标题】：SQL how to query two date windows in one table based on a date in another table optimally?SQL如何根据另一个表中的日期优化查询一个表中的两个日期窗口？
【发布时间】：2022-01-20 09:03:13
【问题描述】：

例如，获取两者：

table_1 中日期前 5 天与 table_2 中每日值的最大值
table_1 中的日期后 5 天与 table_2 中的每日值的最大值

table1 (uid = id) - 日期窗口基于此表中的日期

id	date
a	2021-12-10
b	2021-12-21

表 2 (uid = id + date) - 每日表

id	date	value
a	2021-12-01	0
a	2021-12-02	3
a	...	...
a	2021-12-31	1
b	2021-12-01	4
b	...	...
b	2021-12-31	0

输出表（uid = id）

id	date	start_dt_before	end_dt_before	max_value_before	start_dt_after	end_dt_after	max_value_after
a	2021-12-10	2021-12-05	2021-12-09	3	2021-12-10	2021-12-14	4
b	2021-12-21	2021-12-16	2021-12-20	2	2021-12-21	2021-12-25	5

在两个单独的连接中两次查询其中一个表似乎非常耗费资源。有没有办法在单个查询中更优化地查询这个？

编辑——为每个请求添加示例代码

-- Create table_1
CREATE TABLE table_1 (id CHAR(1), event_dt DATE);
INSERT INTO table_1 (id,event_dt) VALUES ('a','2021-12-10');
INSERT INTO table_1 (id,event_dt) VALUES ('b','2021-12-21');


-- Create table_2

CREATE TABLE table_2 (id CHAR(1), day_dt DATE, value INT);

INSERT INTO table_2 (id,day_dt,value) VALUES ('a','2021-12-01',0);
INSERT INTO table_2 (id,day_dt,value) VALUES ('a','2021-12-02',3);
INSERT INTO table_2 (id,day_dt,value) VALUES ('a','2021-12-03',7);
INSERT INTO table_2 (id,day_dt,value) VALUES ('a','2021-12-04',0);
INSERT INTO table_2 (id,day_dt,value) VALUES ('a','2021-12-05',0);
INSERT INTO table_2 (id,day_dt,value) VALUES ('a','2021-12-06',0);
INSERT INTO table_2 (id,day_dt,value) VALUES ('a','2021-12-07',0);
INSERT INTO table_2 (id,day_dt,value) VALUES ('a','2021-12-08',3);
INSERT INTO table_2 (id,day_dt,value) VALUES ('a','2021-12-09',0);
INSERT INTO table_2 (id,day_dt,value) VALUES ('a','2021-12-10',0);
INSERT INTO table_2 (id,day_dt,value) VALUES ('a','2021-12-11',4);
INSERT INTO table_2 (id,day_dt,value) VALUES ('a','2021-12-12',0);
INSERT INTO table_2 (id,day_dt,value) VALUES ('a','2021-12-13',0);
INSERT INTO table_2 (id,day_dt,value) VALUES ('a','2021-12-14',0);
INSERT INTO table_2 (id,day_dt,value) VALUES ('a','2021-12-15',0);
INSERT INTO table_2 (id,day_dt,value) VALUES ('a','2021-12-16',0);
INSERT INTO table_2 (id,day_dt,value) VALUES ('a','2021-12-17',0);
INSERT INTO table_2 (id,day_dt,value) VALUES ('a','2021-12-18',9);
INSERT INTO table_2 (id,day_dt,value) VALUES ('a','2021-12-19',0);
INSERT INTO table_2 (id,day_dt,value) VALUES ('a','2021-12-20',0);
INSERT INTO table_2 (id,day_dt,value) VALUES ('a','2021-12-21',0);
INSERT INTO table_2 (id,day_dt,value) VALUES ('a','2021-12-22',0);
INSERT INTO table_2 (id,day_dt,value) VALUES ('a','2021-12-23',0);
INSERT INTO table_2 (id,day_dt,value) VALUES ('a','2021-12-24',9);
INSERT INTO table_2 (id,day_dt,value) VALUES ('a','2021-12-25',0);
INSERT INTO table_2 (id,day_dt,value) VALUES ('a','2021-12-26',0);
INSERT INTO table_2 (id,day_dt,value) VALUES ('a','2021-12-27',0);
INSERT INTO table_2 (id,day_dt,value) VALUES ('a','2021-12-28',0);
INSERT INTO table_2 (id,day_dt,value) VALUES ('a','2021-12-29',0);
INSERT INTO table_2 (id,day_dt,value) VALUES ('a','2021-12-30',0);
INSERT INTO table_2 (id,day_dt,value) VALUES ('a','2021-12-31',1);

INSERT INTO table_2 (id,day_dt,value) VALUES ('b','2021-12-01',4);
INSERT INTO table_2 (id,day_dt,value) VALUES ('b','2021-12-02',0);
INSERT INTO table_2 (id,day_dt,value) VALUES ('b','2021-12-03',0);
INSERT INTO table_2 (id,day_dt,value) VALUES ('b','2021-12-04',0);
INSERT INTO table_2 (id,day_dt,value) VALUES ('b','2021-12-05',0);
INSERT INTO table_2 (id,day_dt,value) VALUES ('b','2021-12-06',0);
INSERT INTO table_2 (id,day_dt,value) VALUES ('b','2021-12-07',0);
INSERT INTO table_2 (id,day_dt,value) VALUES ('b','2021-12-08',9);
INSERT INTO table_2 (id,day_dt,value) VALUES ('b','2021-12-09',9);
INSERT INTO table_2 (id,day_dt,value) VALUES ('b','2021-12-10',9);
INSERT INTO table_2 (id,day_dt,value) VALUES ('b','2021-12-11',9);
INSERT INTO table_2 (id,day_dt,value) VALUES ('b','2021-12-12',0);
INSERT INTO table_2 (id,day_dt,value) VALUES ('b','2021-12-13',0);
INSERT INTO table_2 (id,day_dt,value) VALUES ('b','2021-12-14',0);
INSERT INTO table_2 (id,day_dt,value) VALUES ('b','2021-12-15',0);
INSERT INTO table_2 (id,day_dt,value) VALUES ('b','2021-12-16',0);
INSERT INTO table_2 (id,day_dt,value) VALUES ('b','2021-12-17',0);
INSERT INTO table_2 (id,day_dt,value) VALUES ('b','2021-12-18',2);
INSERT INTO table_2 (id,day_dt,value) VALUES ('b','2021-12-19',0);
INSERT INTO table_2 (id,day_dt,value) VALUES ('b','2021-12-20',0);
INSERT INTO table_2 (id,day_dt,value) VALUES ('b','2021-12-21',0);
INSERT INTO table_2 (id,day_dt,value) VALUES ('b','2021-12-22',0);
INSERT INTO table_2 (id,day_dt,value) VALUES ('b','2021-12-23',0);
INSERT INTO table_2 (id,day_dt,value) VALUES ('b','2021-12-24',5);
INSERT INTO table_2 (id,day_dt,value) VALUES ('b','2021-12-25',0);
INSERT INTO table_2 (id,day_dt,value) VALUES ('b','2021-12-26',0);
INSERT INTO table_2 (id,day_dt,value) VALUES ('b','2021-12-27',0);
INSERT INTO table_2 (id,day_dt,value) VALUES ('b','2021-12-28',0);
INSERT INTO table_2 (id,day_dt,value) VALUES ('b','2021-12-29',0);
INSERT INTO table_2 (id,day_dt,value) VALUES ('b','2021-12-30',0);
INSERT INTO table_2 (id,day_dt,value) VALUES ('b','2021-12-31',0);


-- Example queries that creates desired output:

SELECT
    x.id
    , x.event_dt
    , x.before_min_dt
    , x.before_max_dt
    , x.before_value
    , MIN(t2.day_dt) after_min_dt
    , MAX(t2.day_dt) after_max_dt
    , MAX(t2.value) after_value
FROM(
SELECT
    t1.id
    , t1.event_dt
    , MIN(t2.day_dt) before_min_dt
    , MAX(t2.day_dt) before_max_dt
    , MAX(t2.value) before_value
FROM table_1 t1
JOIN table_2 t2
ON (t1.id = t2.id 
    AND t2.day_dt >= DATE_ADD(t1.event_dt, INTERVAL -5 DAY) 
    AND t2.day_dt < t1.event_dt)
GROUP BY 
    t1.id
    , t1.event_dt
) x
JOIN table_2 t2
ON (x.id = t2.id 
    AND t2.day_dt >= x.event_dt 
    AND t2.day_dt < DATE_ADD(x.event_dt, INTERVAL 5 DAY))
GROUP BY     
    x.id
    , x.event_dt
    , x.before_min_dt
    , x.before_max_dt
    , x.before_value;

【问题讨论】：

一个连接，然后是 2 个具有单独窗口定义的窗口函数。
您使用的是哪个版本的 MySQL？那么确切的日期呢？ 2021-12-10...是包含在-5天还是+5天？
@Akina 听起来很理想，你能举个例子吗？谢谢
将样本数据提供为 CREATE TABLE + INSERT INTO 格式化代码，并严格提供此数据所需的表格格式输出。
@Akina 谢谢，我已经添加了示例数据的脚本和生成所需输出的示例查询

标签： mysql sql

【解决方案1】：

SELECT table_2.id,
       table_1.event_dt, 
       table_1.event_dt - INTERVAL 5 DAY before_min_dt,
       table_1.event_dt - INTERVAL 1 DAY before_max_dt,
       MAX(CASE WHEN table_2.day_dt BETWEEN table_1.event_dt - INTERVAL 5 DAY
                                        AND table_1.event_dt - INTERVAL 1 DAY
                THEN value
                END) max_value_before,
       table_1.event_dt after_min_dt,
       table_1.event_dt + INTERVAL 4 DAY after_max_dt,
       MAX(CASE WHEN table_2.day_dt BETWEEN table_1.event_dt
                                        AND table_1.event_dt + INTERVAL 4 DAY
                THEN value
                END) max_value_after
FROM table_1
JOIN table_2 ON table_2.id = table_1.id
            AND table_2.day_dt BETWEEN table_1.event_dt - INTERVAL 5 DAY
                                   AND table_1.event_dt + INTERVAL 4 DAY
GROUP BY table_1.event_dt, table_2.id

https://dbfiddle.uk/?rdbms=mysql_8.0&fiddle=ffda56d3e03e64b8891d438a68a5bbc2

【讨论】：