从查询

Question

我尝试从查询中更新Redshift中的表：

update mr_usage_au au
inner join(select mr.UserId,
                  date(mr.ActionDate) as ActionDate,
                  count(case when mr.EventId in (32) then mr.UserId end) as Moods,
                  count(case when mr.EventId in (33) then mr.UserId end) as Activities,
                  sum(case when mr.EventId in (10) then mr.Duration end) as Duration
           from   mr_session_log mr
           where  mr.EventTime >= current_date - interval '1 days' and mr.EventTime < current_date
           Group By mr.UserId,
                    date(mr.ActionDate)) slog on slog.UserId=au.UserId
                                             and slog.ActionDate=au.Date
set au.Moods = slog.Moods,
    au.Activities=slog.Activities,
    au.Durarion=slog.Duration

但是我收到以下错误：

ERROR: syntax error at or near "au".

Answer 1

对于Redshift（或Postgres），这是完全无效的语法。让我想起了SQL Server ......

应该这样工作（至少在目前的Postgres上）：

UPDATE mr_usage_au
SET    Moods = slog.Moods
     , Activities = slog.Activities
     , Durarion = slog.Duration       
FROM (
   select UserId
        , ActionDate::date
        , count(CASE WHEN EventId = 32 THEN UserId END) AS Moods
        , count(CASE WHEN EventId = 33 THEN UserId END) AS Activities
        , sum(CASE WHEN EventId = 10 THEN Duration END) AS Duration
   FROM   mr_session_log
   WHERE  EventTime >= current_date - 1  -- just subtract integer from a date
   AND    EventTime <  current_date
   GROUP  BY UserId, ActionDate::date
   ) slog
WHERE slog.UserId = mr_usage_au.UserId
AND   slog.ActionDate = mr_usage_au.Date;

Postgres和Redshift的情况通常如此：

• 使用FROM子句加入其他表格。
• 您无法在SET子句中对目标列进行表格限定。

此外，Redshift was forked from PostgreSQL 8.0.2，很久以前。之后只对Postgres进行了一些更新。

• 例如，Postgres 8.0 did not allow a table alias in an UPDATE statement, yet - 这就是您看到错误的原因。

我简化了其他一些细节。