根据两个条件条件为每个唯一 ID 返回不同值的多个 COUNT 字段

Question

我正在尝试使用 MySQL 中的一些日志记录数据，我希望能够告诉每个人，不同的成功请求、不同的不成功请求、非不同的成功请求和非不同的不成功请求的数量都符合额外的全球标准。

以下是代表原始表的组成数据示例：

+--------------+--------+---------------+----------+------+
| request_date | client |    request    | response | site |
+--------------+--------+---------------+----------+------+
| 10/30/2018   | Amy    | RE_food       |      200 | MD   |
| 10/30/2018   | Amy    | RE_food       |      200 | MD   |
| 10/30/2018   | Amy    | RE_transport  |      200 | MD   |
| 10/30/2018   | Amy    | RE_assistance |      404 | MD   |
| 10/30/2018   | Amy    | OA_assistance |      404 | MD   |
| 10/30/2018   | Bob    | RE_food       |      200 | MD   |
| 10/30/2018   | Dan    | RE_food       |      404 | MD   |
| 10/30/2018   | Dan    | RE_food       |      404 | SE   |
| 10/30/2018   | Mike   | RE_transport  |      200 | SE   |
| 10/30/2018   | Sally  | RE_food       |      404 | SE   |
| 10/30/2018   | Sally  | OA_food       |      404 | MD   |
| 10/30/2018   | Zeb    | RE_assistance |      404 | MD   |
| 10/30/2018   | Zeb    | OA_assistance |      404 | MD   |
+--------------+--------+---------------+----------+------+

我想写一个返回的查询

+--------+-----------+------------+----------+-----------+
| client | unique_OK | unique_404 | total_OK | total_404 |
+--------+-----------+------------+----------+-----------+

其中唯一意味着只计算不同的客户端、请求、响应分组。 它会返回结果

+--------+-----------+------------+----------+-----------+
| client | unique_OK | unique_404 | total_OK | total_404 |
+--------+-----------+------------+----------+-----------+
| Amy    |         2 |          1 |        3 |         1 |
| Bob    |         1 |          0 |        1 |         0 |
| Dan    |         0 |          1 |        0 |         1 |
| Mike   |         0 |          0 |        0 |         0 |
| Sally  |         0 |          0 |        0 |         0 |
| Zeb    |         0 |          1 |        0 |         1 |
+--------+-----------+------------+----------+-----------+

（请注意，为了清楚地了解该过程，上面显示了所有四列中计数为 0 的客户端；如果它们不包含在响应中也可以。）

我尝试了这样不起作用的查询，但希望它对我正在尝试做的事情有意义：

SELECT client AS person, 
COUNT(DISTINCT request WHERE response = '200') AS unique_OK, 
COUNT(DISTINCT request WHERE response = '404') AS unique_404 
COUNT(WHERE response = '200') as total_OK,
COUNT(WHERE response = '404') as total_404
FROM transactions 
WHERE request like '%RE%' and site = MD and DATE_FORMAT(request_date, '%Y%m%d')='20181030';

Answer 1

使用条件聚合：

select
    client,
    count(distinct case when response = 200 then request end) unique_ok,
    count(distinct case when response = 404 then request end) unique_404,
    sum(response = 200) total_ok,
    sum(response = 404) total_404
from mytable
where 
    request like '%RE%'
    and site = 'MD' 
    and request_date = '2018-10-30'
group by client

DB Fiddle 上的演示：

| client | unique_ok | unique_404 | total_ok | total_404 |
| ------ | --------- | ---------- | -------- | --------- |
| Amy    | 2         | 1          | 3        | 1         |
| Bob    | 1         | 0          | 1        | 0         |
| Dan    | 0         | 1          | 0        | 1         |
| Zeb    | 0         | 1          | 0        | 1         |