比较和更新更改 - SQL

Question

我有 2 个表：最新表和历史表。历史记录将包含之前加载的所有行，而最新行将包含最新数据。我想提出一个基于 load_date 的综合表，该表应指示更改的时间。

最新表：

 id   col1 col2 load_date
1001    a   g   1/3/2024
1003    q   r   1/3/202

历史表：

id    col1 col2 load_date
1001    a   b   1/1/2024
1002    d   e   1/1/2024
1001    a   g   1/2/2024

我希望我的综合表是这样的：

id     col1 col2 load_date  change
1001    a   b    1/1/2024   new entry
1001    a   g    1/2/2024   col2 changed
1001    a   g    1/3/2024   
1002    d   e    1/1/2024   new entry
1002    d   e    1/2/2024   no show
1003    q   r    1/3/2024   new entry

我在跟踪特定日期的更改和更新时遇到困难。特别是 ID 出现的日期以及 ID 最后出现的日期。

以下是我尝试过的： https://sqlfiddle.com/sql-server/online-compiler?id=5d406760-006a-4acb-9c96-e3a5236b1209

代码：

create table latest(id int, col1 varchar, col2 varchar, load_date date);

        insert into latest 
        values(1001,'a','g','1/3/2024'),
              (1003,'q','r','1/3/2024');--newly showed up on 1/3

        --select * from latest;
        
        create table history(id int, col1 varchar, col2 varchar, load_date date);

        insert into history 
        values
        (1001,'a','b','1/1/2024'),
        (1002,'d','e','1/1/2024'),
        (1001,'a','g','1/2/2024');--colb changed on 1/2
        --(1002,'d','e','1/2/2024')--did not show up
        --select * from history;
        
        with combined as
     (
     select *,'latest' as source from latest l
     union all
     select *, 'history' as source from history h
     ),
     
    changes AS (
        SELECT
            ct1.id,
            ct1.col1,
            ct1.col2,
            ct1.load_date,
            CASE
                WHEN ct1.col1 <> ct2.col1 THEN 'col1 changed'
                WHEN ct1.col2 <> ct2.col2 THEN 'col2 changed'
                ELSE NULL
            END AS changes
        FROM
            combined ct1
            LEFT JOIN combined ct2
            ON ct1.id = ct2.id
    )

select * from changes

我可以寻求帮助来解决这个问题吗？

Answer 1

首先您

union

两个表，然后使用窗口函数

lag()

获取上一行值。要获得最终的

change

，请使用

case

表达式将当前行值与上一行值进行比较。

要处理“未显示”，请使用

history

中的最新行并使用

latest

检查。这个逻辑在联合查询的最后一部分处理

with cte as
(
  select id, col1, col2, load_date, change = null
  from   latest
  union all
  select id, col1, col2, load_date, change = null
  from   history
  union all
  select id, col1, col2, load_date = dateadd(day, 1, load_date),
         change = 'no show'
  from   (
             select id, col1, col2, load_date,
                    rn = row_number() over (partition by id order by load_date desc)
             from history 
         ) h
  where  h.rn = 1
  and    not exists
         (
             select *
             from   latest x
             where  x.id = h.id
         )
),
cte2 as
(  
  select id, col1, col2, load_date, change, 
         prev_col1 = lag(col1) over (partition by id order by load_date),
         prev_col2 = lag(col2) over (partition by id order by load_date)  
  from   cte
)
select *,
       change = isnull(change, '')
              + case when prev_col1 is null and prev_col2 is null
                     then 'new entry'
                     else ''
                     end
              + case when prev_col1 <> col1 
                     then 'col1 changed '
                     else ''
                     end
              + case when prev_col2 <> col2 
                     then 'col2 changed '
                     else ''
                     end  
from   cte2
order by id, load_date

db<>小提琴演示