Amazon redshift 中的每月保留

Posted 2023-03-31

【中文标题】Amazon redshift 中的每月保留

【英文标题】：Monthly retention in Amazon redshift

【发布时间】：2016-04-04 23:07:03

【问题描述】：

我正在尝试计算 Amazon Redshift 中的每月保留率并提出以下查询：

Query 1

SELECT EXTRACT(year FROM activity.created_at) AS Year,
       EXTRACT(month FROM activity.created_at) AS Month,
       COUNT(DISTINCT activity.member_id) AS active_users,
       COUNT(DISTINCT future_activity.member_id) AS retained_users,
       COUNT(DISTINCT future_activity.member_id) / COUNT(DISTINCT activity.member_id)::float AS retention
FROM ads.fbs_page_view_staging activity
  LEFT JOIN ads.fbs_page_view_staging AS future_activity
         ON activity.mongo_id = future_activity.mongo_id
AND datediff ('month',activity.created_at,future_activity.created_at) = 1
GROUP BY Year,
         Month
ORDER BY Year,
         Month

由于某种原因，此查询返回 zero retained_users 和 zero retention。对于为什么会发生这种情况，或者完全不同的每月保留查询可能会起作用，我将不胜感激。

我根据另一篇 SO 帖子修改了查询，结果如下：

Query 2

WITH t AS (
   SELECT member_id
         ,date_trunc('month', created_at) AS month
         ,count(*) AS item_transactions
         ,lag(date_trunc('month', created_at)) OVER (PARTITION BY  member_id
                                           ORDER BY date_trunc('month', created_at)) 
          = date_trunc('month', created_at) - interval '1 month'
            OR NULL AS repeat_transaction
   FROM   ads.fbs_page_view_staging
   WHERE  created_at >= '2016-01-01'::date
   AND    created_at <  '2016-04-01'::date -- time range of interest.
   GROUP  BY 1, 2
   )
SELECT month
      ,sum(item_transactions) AS num_trans
      ,count(*) AS num_buyers
      ,count(repeat_transaction) AS repeat_buyers
      ,round(
          CASE WHEN sum(item_transactions) > 0
             THEN count(repeat_transaction) / sum(item_transactions) * 100
             ELSE 0
          END, 2) AS buyer_retention
FROM   t
GROUP  BY 1
ORDER  BY 1;

这个查询给了我以下错误：

An error occurred when executing the SQL command:
WITH t AS (
   SELECT member_id
         ,date_trunc('month', created_at) AS month
         ,count(*) AS item_transactions
         ,lag(date_trunc('m...

[Amazon](500310) Invalid operation: Interval values with month or year parts are not supported
Details: 
 -----------------------------------------------
  error:  Interval values with month or year parts are not supported
  code:      8001
  context:   interval months: "1"
  query:     616822
  location:  cg_constmanager.cpp:145
  process:   padbmaster [pid=15116]
  -----------------------------------------------;

我感觉Query 2 会比Query 1 表现更好，所以我更愿意修复这个错误。

任何帮助将不胜感激。

【问题讨论】：

我认为您对查询 1 的问题是您已将间隔条件 datediff ('month',activity.created_at,future_activity.created_at) = 1 放入 JOIN 中。不要认为那会奏效。因此连接失败，因此您在连接的右侧得到 NULL，导致计数为零。将条件移到“WHERE”时会发生什么？

【参考方案1】：

查询 1 看起来不错。我试过类似的。见下文。您在表 (ads.fbs_page_view_staging) 和同一列 (created_at) 上使用自联接。假设 mongo_id 是唯一的，datediff('month'....) 将始终返回 0，datediff ('month',activity.created_at,future_activity.created_at) = 1 将始终为 false。

-- Count distinct events of join_col_id that have lapsed for one month.
SELECT count(distinct E.join_col_id) dist_ct
FROM public.fact_events E
JOIN public.dim_table Z
  ON E.join_col_id = Z.join_col_id
WHERE datediff('month', event_time, sysdate) = 1;

-- 2771654 -- dist_ct

【讨论】：

完全正确，我加入了错误的列，应该是member_id，而不是mongo_id。我的愚蠢错误。感谢您指出这一点。