如何组合2个SQL查询并检索累积计数？

Question

我正在尝试检索特定A/B测试组合每天的访问次数和转化次数.每种组合代表A/B测试的不同变化.在这里,我只使用'1'并'2'代表变化,但从技术上讲,可能会有更多的变化.

我写了以下2个查询,它们独立工作.是否可以组合这些或写一个检索我想要的数据的查询？

访问查询:

SELECT DATE(visit.created), visit.combination, COUNT(visit.id) 
FROM visit 
WHERE visit.user_id = 6
AND visit.experiment_id = 1
GROUP BY DATE(visit.created), visit.combination

访问结果:

转化查询:

SELECT DATE(conversion.created), conversion.combination, COUNT(conversion.id) 
FROM conversion 
WHERE conversion.user_id = 6
AND conversion.experiment_id = 1
AND conversion.goal_id = 1
GROUP BY DATE(conversion.created), conversion.combination

转化结果:

如果我可以检索一个正在运行的总计(累计)计数,如下所示,请查看最后两列.我已经按照组合对下表进行了分组,因此累积计数更容易理解:

+---------------+-------------+----------------------+-----------------+--------------+--------------+
| DATE(created) | combination | COUNT(conversion.id) | COUNT(visit.id) | cumulative_c | cumulative_v |
+---------------+-------------+----------------------+-----------------+--------------+--------------+
| 2015-11-17    | 1           |                    1 |               3 |            1 |            3 |
| 2015-11-18    | 1           |                    7 |               4 |            8 |            7 |
| 2015-11-19    | 1           |                    3 |               8 |           11 |           15 |
| 2015-11-17    | 2           |                    4 |               1 |            4 |            1 |
| 2015-11-18    | 2           |                    2 |               6 |            6 |            7 |
| 2015-11-19    | 2           |                    9 |               6 |           15 |           13 |
+---------------+-------------+----------------------+-----------------+--------------+--------------+

数据库架构:

Answer 1

组合非常简单：添加 0 值列，执行UNION_ALL，然后再次分组并求和。

SELECT dt, combination, SUM(v_count) as v_count, SUM(c_count) as c_count
FROM 
(
SELECT DATE(visit.created) as dt, visit.combination as combination, COUNT(visit.id) as v_count, 0 as c_count
FROM visit 
WHERE visit.user_id = 6
AND visit.experiment_id = 1
GROUP BY DATE(visit.created), visit.combination

UNION ALL

SELECT DATE(conversion.created) as dt, conversion.combination as combination, 0 as v_count, COUNT(conversion.id) as c_count
FROM conversion 
WHERE conversion.user_id = 6
AND conversion.experiment_id = 1
AND conversion.goal_id = 1
GROUP BY DATE(conversion.created), conversion.combination
) as t
GROUP BY dt, combination

现在，运行总计。在更高级的 DBMS 中，这称为“窗口”或“分析”函数。例如，在 Oracle 中您可以这样做：

SELECT dt, combination, SUM(v_count) OVER (PARTITION BY combination ORDER BY dt) as v_cumulative

对于上面的查询，它会给你你想要的。然而MySQL没有这样的功能。有很多方法，例如这里和这里描述的，但它们非常棘手。