加入4个表with group by，2 having和where子句

Question

我的数据库由4个表组成：

1. 用户（id，“姓名”，姓氏，生日）
2. 友谊（userid1，userid2，“timestamp”）
3. 帖子（id，userid，“text”，“timestamp”）
4. 喜欢（postid，userid，“timestamp”）

我需要在2018年1月之间获得一组具有3个以上友谊的独特用户名，并且他们的“喜欢”平均每个“帖子”的范围为[10; 35）。

我在第一步写了这句话：

select  distinct u."name"
from users u
join friendships f on u.id = f.userid1
where f."timestamp" between '2018-01-01'::timestamp and '2018-01-31'::timestamp
group by u.id
having count(f.userid1) > 3;

它工作正常，返回3行。但是当我以这种方式添加第二部分时：

select  distinct u."name"
from users u
join friendships f on u.id = f.userid1
join posts p on p.userid = u.id
join likes l on p.id = l.postid
where f."timestamp" between '2018-01-01'::timestamp and '2018-01-31'::timestamp
group by u.id
having count(f.userid1) > 3 
    and ((count(l.postid) / count(distinct l.postid)) >= 10 
        and (count(l.postid) / count(distinct l.postid)) < 35);

我疯了94行。我不知道为什么。将感谢可能的帮助。

Answer 1

您在distinct中不需要u.name，因为聚合将删除副本。

select
   u."name"
from 
   users u
   inner join friendships f on u.id = f.userid1
   inner join posts p on u.id = p.userid
   inner join likes l on p.id = l.postid
where 
   f."timestamp" >= '2018-01-01'::timestamp 
   and f."timestamp" < '2018-02-01'::timestamp
group by 
    u."name"
having 
    count(distinct f.userid1) > 3 
    and ((count(l.postid) / count(distinct l.postid)) >= 10 
            and (count(l.postid) / count(distinct l.postid)) < 35);

正如评论所述。当你使用between为date做范围时，不是一个好主意。

f."timestamp" >= '2018-01-01'::timestamp 
and f."timestamp" < '2018-02-01'::timestamp

会给你一个月的整月。

Answer 2

试试下面！使用“count（f.userid1）> 3”的问题在于，如果用户具有例如2个朋友和6个帖子以及3个喜欢他们将得到2 x 6 = 12行，所以有12个非空f.userid1的记录。通过计算不同的f.userid2，您可以统计不同的朋友。对于用于过滤的其他计数也会出现类似问题。

select  u."name"
from users u
join friendships f on u.id = f.userid1
join posts p on p.userid = u.id
left join likes l on p.id = l.postid
where f."timestamp" > '2018-01-01'::timestamp and f."timestamp" < '2018-02-01'::timestamp
group by u.id, u."name"
having
 --at least three distinct friends
 count( distinct f.userid2) > 3 
  --distinct likes / distinct posts
  --we use l.* to count distinct likes since there's no primary key
  and ((count(distinct l.*) / count(distinct p.id)) >= 10 
        and ((count(distinct l.*) / count(distinct p.id)) < 35);