评估移动由每个 ID 创建的最后 5 个订单

Posted 2023-03-21

【中文标题】评估移动由每个 ID 创建的最后 5 个订单

【英文标题】：Evaluate moving last 5 orders created by each ID

【发布时间】：2021-09-06 02:55:13

【问题描述】：

我有下表：

create table DemoTable
(
   buyer_id INT,
   order_id INT,
   variable_a INT,
   created_date datetime
);

INSERT INTO DemoTable(buyer_id, order_id, variable_a, created_date) VALUES (54, 75, 609, '2019-06-02 16:46:53');
INSERT INTO DemoTable(buyer_id, order_id, variable_a, created_date) VALUES (54, 74, 363, '2019-06-01 22:00:00');
INSERT INTO DemoTable(buyer_id, order_id, variable_a, created_date) VALUES (54, 73, 300, '2019-06-01 20:00:02');
INSERT INTO DemoTable(buyer_id, order_id, variable_a, created_date) VALUES (54, 72, 225, '2019-06-01 18:12:03');
INSERT INTO DemoTable(buyer_id, order_id, variable_a, created_date) VALUES (54, 71, 110, '2019-06-01 16:10:03');
INSERT INTO DemoTable(buyer_id, order_id, variable_a, created_date) VALUES (54, 55, 213, '2019-06-01 14:22:03');
INSERT INTO DemoTable(buyer_id, order_id, variable_a, created_date) VALUES (54, 51, 225, '2018-12-01 16:12:03');
INSERT INTO DemoTable(buyer_id, order_id, variable_a, created_date) VALUES (54, 32, 110, '2018-12-01 15:10:03');
INSERT INTO DemoTable(buyer_id, order_id, variable_a, created_date) VALUES (54, 12, 213, '2018-12-01 12:22:03');

buyer_id	order_id	variable_a	created_date
54	75	609	2019-06-02 16:46:53
54	74	363	2019-06-01 22:00:00
54	75	300	2019-06-01 20:00:02
54	73	225	2019-06-01 18:12:03
54	72	110	2019-06-01 16:10:03
54	55	213	2019-06-01 14:22:03
54	51	225	2018-12-01 16:12:03
54	32	110	2018-12-01 15:10:03
54	12	213	2018-12-01 12:22:03

我需要做的是评估由 buy_id 创建的最后 5 个订单 (l5o)，它是每个 order_id 的移动 l5o。下面的例子是对最后5个订单的variable_a进行平均：

buyer_id	variable_a_avg_l5o	created_date
54	321.4	2019-06-02
54	242.2	2019-06-01

2019-06-02：(609+363+300+225+110)/5 = 321.4

2019-06-01：(363+300+225+110+213)/5 = 242.2

请指导我如何应用此逻辑。请注意，我使用的是 presto，但尚不支持相关子查询。

谢谢！

【参考方案1】：

如果我理解正确，您只需要一个带有窗口框架规范的窗口函数：

select t.*,
       avg(variable_a) over (partition by buyer_id
                             order by created_date
                             rows between 4 preceding and current row
                            ) as avg_5
from t;