如何使用 SQL 逐行更新使用事务日志的表

Question

我有一个表格，其中显示了工单 ID 及其状态以及在 BI 数据库上分配给谁。这里没有重复和粒度是每张票。

票务表

Ticket_ID   Status  Assigned_To   Date
001         Open    Team A        01/01/2020 10:01
002         Working Team B        01/01/2020 10:01
003         Open    TeamC         02/01/2020 18:50
004         ResolvedTeam B        02/01/2020 19:22

现在，有一个包含每个更改的每日交易日志。但是，字段更改显示为如下行。

每日交易日志

Ticket_id       Field_Name      New Value      Date
001           Assigned_to       Team B        02/01/2020 10:23
002           Assigned_to       Team A        02/01/2020 10:23
002           Status            Resolved      02/01/2020 10:24
003           status            Resolved      03/01/2020 10:24
004           Assigned_to       Team C        03/01/2020 13:50
004           Assigned_to       Team A        03/01/2020 13:51
004           Status            Resolved      03/01/2020 13:51

现在我想编写一个更新脚本来根据日志上的新值更新 Ticket 表。考虑到每天都会创建事务日志表，我认为设置操作将不起作用。我需要 SQL 脚本逐行处理事务日志表并使用新值更新票证表。我知道这是低效的，但是，每天只有数百行。不是数千。

我希望看到的结果如下，其中票证表已更新为最新值..

Ticket_ID       Status     Assigned_To  Date
001           Open       Team B       02/01/2020 10:23
002           Resolved   Team A       02/01/2020 10:24
003           Resolved   TeamC        03/01/2020 10:24
004           Resolved   Team A       03/01/2020 13:51

谁能帮帮我。谢谢

Answer 1

您可以使用distinct on 获取每列的最新更改：

select distinct on (ticket_id, field_name) dtl.*
from daily_transaction_logs dtl
order by ticket_id, date desc;

然后您可以将其汇总以获得每张票的一行：

select ticket_id,
       max(new_value) filter (where field_name = 'Assigned_to') as assigned_to,
       max(new_value) filter (where field_name = 'Status') as status,
       max(date) as date
from (select distinct on (ticket_id, field_name) dtl.*
      from daily_transaction_logs dtl
      order by ticket_id, date desc
     ) tf
group by ticket_id;

最后，您可以将其合并到更新中：

update tickets t
    set assigned_to = tt.assigned_to,
        status = tt.status,
        date = tt.date
    from (select ticket_id,
                 max(new_value) filter (where field_name = 'Assigned_to') as assigned_to,
                 max(new_value) filter (where field_name = 'Status') as status,
                 max(date) as date
          from (select distinct on (ticket_id, field_name) dtl.*
                from daily_transaction_logs dtl
                order by ticket_id, date desc
               ) tf
          group by ticket_id
         ) tt
    where tt.ticket_id = t.ticket_id;