2

我正在创建一个数据库视图，将在线订单与网站访问之前的用户关联起来。这适用于电子商务网站，因此一个用户可以访问和订购多次。

我已经在 user_id 上加入了访问表和订单表，并将最近的小于会话时间与订单时间相关联。现在，我希望在订单#1 之前的每次访问都是“1”，然后在访问之后直到订单#2 是“2”。此外，如果该特定用户没有 order_id，我想返回“0”。请参阅下面链接的屏幕截图以供参考。

我已经尝试使用dense_rank，但它只对存在order_id 的行进行排名。我想发扬这些行列。

SELECT v.id AS visit_id,
    v.user_id,
    v.started_at AS visit_date,
    dense_rank() OVER (PARTITION BY v.user_id ORDER BY v.started_at) AS visit_number,
    dense_rank() OVER (PARTITION BY v.user_id ORDER BY o.id) AS order_number,
    o.id AS order_id,
    o.created_at AS order_date
   FROM visits v
     FULL JOIN orders o ON v.user_id = o.user_id AND v.started_at < o.created_at AND o.created_at < (( SELECT min(visits.started_at) AS min
           FROM visits
          WHERE visits.user_id = v.user_id AND visits.started_at > v.started_at)) AND (v.started_at + '24:00:00'::interval) > o.created_at
  GROUP BY v.id, v.user_id, v.started_at, o.id, o.created_at
  ORDER BY v.started_at;

当前结果 预期成绩

4

2 回答 2

0

用于lag检查前一行是否为非空，以便将其标记为新组开始。设置标志后，您可以使用运行总和来定义组。

SELECT T.*,
       1+SUM(FLAG) OVER(PARTITION BY user_id ORDER BY visit_date) AS order_number
FROM (
SELECT v.id AS visit_id,
    v.user_id,
    v.started_at AS visit_date,
    dense_rank() OVER (PARTITION BY v.user_id ORDER BY v.started_at) AS visit_number,
    o.id AS order_id,
    o.created_at AS order_date,
    --conditioncheck with lag
    case when lag(o.id) over(partition by v.user_id order by v.started_at) is not null then 1 else 0 end as flag
   FROM visits v
     FULL JOIN orders o ON v.user_id = o.user_id AND v.started_at < o.created_at AND o.created_at < (( SELECT min(visits.started_at) AS min
           FROM visits
          WHERE visits.user_id = v.user_id AND visits.started_at > v.started_at)) AND (v.started_at + '24:00:00'::interval) > o.created_at
    ) T
于 2019-04-12T21:29:56.733 回答
0

似乎没有必要，GROUP BY但我会留下它。你基本上需要一个累积的总和。

我会在特定订单之前为所有访问分配订单号：

SELECT v.id AS visit_id, v.user_id,
       v.started_at AS visit_date,
       dense_rank() OVER (PARTITION BY v.user_id ORDER BY v.started_at) AS visit_number,
       dense_rank() OVER (PARTITION BY v.user_id ORDER BY o.id) AS order_number,
       o.id AS order_id,
       o.created_at AS order_date,
       count(o.id) over (partition by v.user_id order by v.started_at) as order_number
FROM visits v FULL JOIN
     orders o
     ON v.user_id = o.user_id AND
        v.started_at < o.created_at AND
        o.created_at < (SELECT min(visits.started_at)
                        FROM visits v2 
                        WHERE v2.user_id = v.user_id AND 
                              v2.started_at > v.started_at) AND
        (v.started_at + '24:00:00'::interval) > o.created_at
GROUP BY v.id, v.user_id, v.started_at, o.id, o.created_at
ORDER BY v.started_at;

我认为这是您想要的逻辑：

SELECT v.id AS visit_id, v.user_id,
       v.started_at AS visit_date,
       dense_rank() OVER (PARTITION BY v.user_id ORDER BY v.started_at) AS visit_number,
       dense_rank() OVER (PARTITION BY v.user_id ORDER BY o.id) AS order_number,
       o.id AS order_id,
       o.created_at AS order_date,
       MIN(o.order_number) OVER (PARTITION BY v.user_id ORDER BY v.started_at DESC) as order_number
FROM visits v FULL JOIN
     (SELECT o.*,
             ROW_NUMBER() OVER (PARTITION BY o.user_id ORDER BY o.id) as order_number
      FROM orders o
     ) o
     ON v.user_id = o.user_id AND
        v.started_at < o.created_at AND
        o.created_at < (SELECT min(visits.started_at)
                        FROM visits v2 
                        WHERE v2.user_id = v.user_id AND 
                              v2.started_at > v.started_at) AND
        (v.started_at + '24:00:00'::interval) > o.created_at
GROUP BY v.id, v.user_id, v.started_at, o.id, o.created_at
ORDER BY v.started_at;

但是，它可能会NULL在您想要的地方产生 s 0。

于 2019-04-12T22:09:44.387 回答