如何在同一个表上多次加入每行的最大值

Question

我有一个包含 5 个字段的表格活动：

id - numeric
user_id - numeric 
type_id - numeric
content_id - numeric
date_of_activity date

我需要做一个查询，其结果将是 7 列：

id(type1 activity),content_id(type 1) at max(date_of_activity for type 1),max(date_of_activity for type 1)content_id(type 2) at max(date_of_activity for type 2),max(date_of_activity for type 2)content_id(type 3) at max(date_of_activity for type 3),max(date_of_activity for type 3)

如果用户的 type_id 没有对应的值，则对应列中应显示 NULL 值。

例子：

users
    user_id: 1
    user_id: 2
    user_id: 3

activities

    id: 1
    user_id: 1
    type_id: 2
    date_of_activity: 2021/05/01
    content_id: 12

    id: 2
    user_id: 1
    type_id: 3
    date_of_activity: 2021/05/01
    content_id: 102

    id: 3
    user_id: 2
    type_id: 3
    date_of_activity: 2021/05/01
    content_id: 213

    id: 4
    user_id: 1
    type_id: 2
    date_of_activity: 2021/05/02
    content_id: 13

    id: 5
    user_id: 1
    type_id: 3
    date_of_activity: 2021/05/02
    content_id: 103

    id: 6
    user_id: 1
    type_id: 1
    date_of_activity: 2021/05/03
    content_id: 1

    id: 7
    user_id: 2
    type_id: 1
    date_of_activity: 2021/05/03
    content_id: 212
    
    id: 8
    user_id: 1
    type_id: 3
    date_of_activity: 2021/05/03
    content_id: 104

    id: 8
    user_id: 1
    type_id: 3
    date_of_activity: 2021/05/04
    content_id: 105

    id: 10
    user_id: 1
    type_id: 1
    date_of_activity: 2021/05/05
    content_id: 2

    id: 11
    user_id: 1
    type_id: 3
    date_of_activity: 2021/05/05
    content_id: 106

结果应该是：

id:10,
user_id:1,
type_id:1,
date_of_activity: 2021/05/05,
activity2.date_of_activity:2021/05/02,
activity2.content_id:13,
activity3.date_of_activity: 2021/05/05,
activity3.content_id:106

id:7,
user_id: 2,
type_id: 1,
date_of_activity: 2021/05/03,
content_id: 212,
activity2.date_of_activity: NULL,
activity2.content_id: NULL,
activity3.date_of_activity:2021/05/01
activity3.content_id:213

如您所见，我检索了一个活动，每个用户的每个活动类型，每个活动都是最新的，等等所有用户...

我有一个查询，可以检索每个用户的 ONE 活动类型的最新活动：

SELECT
activities.*,
dDates.max
FROM activities
JOIN users on users.id = activities.user_id
INNER JOIN (SELECT user_id,max(activities.date_of_activity) FROM activities WHERE activities.type_id = 6 AND activities.deleted_at IS NULL group by user_id) as dDates on activities.user_id = maxDates.user_id
WHERE activities.type_id = 6 AND activities.deleted_at IS NULL AND activities.date_of_activity = dDates.max;

但有时我需要将多个活动作为同一个查询的一部分加入，所以我试图一次获取所有内容，但我找不到一种不是很慢的方法。

我尝试为每种类型创建一个视图/使用“With”：

CREATE VIEW MostRecentType1 AS
SELECT
activities.*,
dDates.max
FROM activities
JOIN users on users.id = activities.user_id
INNER JOIN (SELECT user_id,max(activities.date_of_activity) FROM activities WHERE activities.type_id = 1 AND activities.deleted_at IS NULL group by user_id) as dDates on activities.user_id = maxDates.user_id
WHERE activities.type_id = 1 AND activities.deleted_at IS NULL AND activities.date_of_activity = dDates.max;

CREATE VIEW MostRecentType2 AS
SELECT
activities.*,
dDates.max
FROM activities
JOIN users on users.id = activities.user_id
INNER JOIN (SELECT user_id,max(activities.date_of_activity) FROM activities WHERE activities.type_id = 1 AND activities.deleted_at IS NULL group by user_id) as dDates on activities.user_id = maxDates.user_id
WHERE activities.type_id = 1 AND activities.deleted_at IS NULL AND activities.date_of_activity = dDates.max;

最后是视图的左连接

SELECT
activities.*,
dDates.max,
type2.date_of_activity as "type2.date_of_activity as",
type2.content_id as "type2.content_id",
type3.date_of_activity as "type3.date_of_activity" ,
type3.content_id as "type3.content_id" ,
FROM activities
JOIN users on users.id = activities.user_id
INNER JOIN (SELECT user_id,max(activities.date_of_activity) FROM activities WHERE activities.type_id = 1 AND activities.deleted_at IS NULL group by user_id) as dDates on activities.user_id = maxDates.user_id
LEFT JOIN MostRecentType2 as type2 on activities.user_id = type2.user_id
LEFT JOIN MostRecentType3 as type3 on activities.user_id = type3.user_id
WHERE activities.type_id = 1 AND activities.deleted_at IS NULL AND activities.date_of_activity = dDates.max;

但这根本不能很好地扩展。

在我大部分为空的活动表上，如果我只执行第一个查询，它将需要 4 毫秒，但是一旦我添加第一个“MostRecentType2”加入它就会跳转 到 40 毫秒，然后 MostRecentType3 到 80 毫秒并继续以这种方式递增，在某些情况下我需要大量的连接。

关于如何（更好地）进行查询的任何想法？

Answer 1

嗯。 . .您可以使用DISTINCT ON 获取每个用户的每种类型的最新行。然后只需使用条件聚合：

SELECT user_id,
       MAX(date_of_activity) FILTER (WHERE type_id = 1) as date_1,
       MAX(content_id) FILTER (WHERE type_id = 1) as content_1,
       MAX(date_of_activity) FILTER (WHERE type_id = 2) as date_2,
       MAX(content_id) FILTER (WHERE type_id = 2) as content_2,
       MAX(date_of_activity) FILTER (WHERE type_id = 3) as date_3,
       MAX(content_id) FILTER (WHERE type_id = 3) as content_3
FROM (SELECT DISTINCT ON (user_id, type_id) a.*
      FROM activities a
      WHERE a.deleted_at IS NULL
      ORDER BY user_id, type_id, date_of_activity DESC
     ) a
GROUP BY user_id