我们有两张表：
车辆：
分配：
用最新分配更新 Vehicle 表中每一行的最有效（最简单）的方法是什么？在 SQL Server 中，我将使用 UPDATE FROM 并加入具有最新分配的每辆车。Oracle 没有 UPDATE FROM。你是如何在甲骨文中做到这一点的？
** 编辑 **
我要求更新的最佳 SQL 查询。我将使用触发器来更新主表中的数据。我知道如何编写触发器。我要问的是如何编写 SQL 查询来更新 Vehicle 表。例子会很好。谢谢你。
当前设置要求您在 ALLOCATIONS 表上使用触发器来维持糟糕的决策选择。也就是说，使用：
UPDATE VEHICLE
SET (LastAllocationUserName, LastAllocationDate, LastAllocationId) =
(SELECT a.username,
a.date,
a.id
FROM ALLOCATIONS a
JOIN (SELECT b.vehicleid,
MAX(b.date) AS max_date
FROM ALLOCATIONS b
GROUP BY b.vehicleid) x ON x.vehicleid = a.vehicleid
AND x.max_date = a.date
WHERE a.vehicleid = VEHICLE.id)
通过从 VEHICLE 表中删除有问题的列并使用视图提供最新的分配信息会更好地解决这个问题。
从设计的角度来看，我更愿意在 Vehicle 表上积极维护这三个字段，并将“Allocations”填充为历史表（可能由触发器）。将父表上的更新推送到子表上的插入比反过来要容易得多。
正如大多数其他人所指出的那样：由于您的数据模型，您遇到了一个大问题。为这个模型编写的大多数代码将比它需要的要困难得多。我已经通过上下投票以及在一些评论中表达了这一点，但这还不够。
如果您继续前进，那么下面的代码将演示需要做什么。希望它吓到你:-)
样本表：
SQL> create table vehicles (id,registrationnumber,lastallocationusername,lastallocationdate,lastallocationid)
2 as
3 select 1, 1, 'Me', sysdate-1, 2 from dual union all
4 select 2, 2, 'Me', sysdate, 3 from dual
5 /
Table created.
SQL> create table allocations (id,vehicleid,username,mydate)
2 as
3 select 1, 1, 'Me', sysdate-2 from dual union all
4 select 2, 1, 'Me', sysdate-1 from dual union all
5 select 3, 2, 'Me', sysdate-1 from dual
6 /
Table created.
触发器必须查看它自己的表以确定最后一次分配。Oracle 通过引发变异表错误来防止这种类型的脏读。为了避免这种情况，我创建了一个 SQL 类型和一个包：
SQL> create type t_vehicle_ids is table of number;
2 /
Type created.
SQL> create package allocations_mutating_table
2 as
3 procedure reset_vehicleids;
4 procedure store_vehicleid (p_vehicle_id in vehicles.id%type);
5 procedure adjust_vehicle_last_allocation;
6 end allocations_mutating_table;
7 /
Package created.
SQL> create package body allocations_mutating_table
2 as
3 g_vehicle_ids t_vehicle_ids := t_vehicle_ids()
4 ;
5 procedure reset_vehicleids
6 is
7 begin
8 g_vehicle_ids.delete;
9 end reset_vehicleids
10 ;
11 procedure store_vehicleid (p_vehicle_id in vehicles.id%type)
12 is
13 begin
14 g_vehicle_ids.extend;
15 g_vehicle_ids(g_vehicle_ids.count) := p_vehicle_id;
16 end store_vehicleid
17 ;
18 procedure adjust_vehicle_last_allocation
19 is
20 begin
21 update vehicles v
22 set ( v.lastallocationusername
23 , v.lastallocationdate
24 , v.lastallocationid
25 ) =
26 ( select max(a.username) keep (dense_rank last order by a.mydate)
27 , max(a.mydate)
28 , max(a.id) keep (dense_rank last order by a.mydate)
29 from allocations a
30 where a.vehicleid = v.id
31 )
32 where v.id in (select column_value from table(cast(g_vehicle_ids as t_vehicle_ids)))
33 ;
34 end adjust_vehicle_last_allocation
35 ;
36 end allocations_mutating_table;
37 /
Package body created.
然后 3 个数据库触发器将更新代码从行级别移动到语句级别，从而规避了 mutating table 错误：
SQL> create trigger allocations_bsiud
2 before insert or update or delete on allocations
3 begin
4 allocations_mutating_table.reset_vehicleids;
5 end;
6 /
Trigger created.
SQL> create trigger allocations_ariud
2 after insert or update or delete on allocations
3 for each row
4 begin
5 allocations_mutating_table.store_vehicleid(nvl(:new.vehicleid,:old.vehicleid));
6 end;
7 /
Trigger created.
SQL> create trigger allocations_asiud
2 after insert or update or delete on allocations
3 begin
4 allocations_mutating_table.adjust_vehicle_last_allocation;
5 end;
6 /
Trigger created.
并进行一个小测试以验证它是否可以在单个用户环境中工作：
SQL> select * from vehicles
2 /
ID REGISTRATIONNUMBER LA LASTALLOCATIONDATE LASTALLOCATIONID
---------- ------------------ -- ------------------- ----------------
1 1 Me 13-05-2010 14:03:43 2
2 2 Me 14-05-2010 14:03:43 3
2 rows selected.
SQL> insert into allocations values (4, 1, 'Me', sysdate)
2 /
1 row created.
SQL> select * from vehicles
2 /
ID REGISTRATIONNUMBER LA LASTALLOCATIONDATE LASTALLOCATIONID
---------- ------------------ -- ------------------- ----------------
1 1 Me 14-05-2010 14:03:43 4
2 2 Me 14-05-2010 14:03:43 3
2 rows selected.
SQL> update allocations
2 set mydate = mydate - 2
3 where id = 4
4 /
1 row updated.
SQL> select * from vehicles
2 /
ID REGISTRATIONNUMBER LA LASTALLOCATIONDATE LASTALLOCATIONID
---------- ------------------ -- ------------------- ----------------
1 1 Me 13-05-2010 14:03:43 2
2 2 Me 14-05-2010 14:03:43 3
2 rows selected.
SQL> delete allocations
2 where id in (2,4)
3 /
2 rows deleted.
SQL> select * from vehicles
2 /
ID REGISTRATIONNUMBER LA LASTALLOCATIONDATE LASTALLOCATIONID
---------- ------------------ -- ------------------- ----------------
1 1 Me 12-05-2010 14:03:43 1
2 2 Me 14-05-2010 14:03:43 3
2 rows selected.
现在您所要做的就是添加一些序列化以使其在多用户环境中 100% 工作。但希望这个例子足够可怕。
问候，罗布。
在 Oracle 中使用另一个表“更新”的最简单方法是使用 MERGE。
MERGE INTO vehicle v
USING (
-- subquery to get info you need
) s ON (v.id = s.vehicleId)
WHEN MATCHED THEN UPDATE SET
username = s.username
...
您是否正在寻找触发器内的更新？
CREATE TRIGGER ALLOCATION_I
AFTER INSERT ON ALLOCATION
REFERENCING NEW AS NEW
FOR EACH ROW
Begin
UPDATE Vehicle
set LastAllocationUserName = :NEW.Username
,LastAllocationDate = :NEW.date
,LastAllocationId = :NEW.id
WHERE Id = :NEW.VehicleId;
END;
UPDATE VEHICLE V
SET (LastAllocationId, LastAllocationDate, LastAllocationUserName) =
(SELECT a.id
,a.date
,a.username
FROM ALLOCATIONS a
where a.VehicleId = V.id
and a.date = ( select max(Last_a.date) from ALLOCATIONS Last_a
where Last_a.VehicleId = V.id )
)
你是对的。带有历史表的视图很慢。没有快速“加入最后一个记录”这样的东西。触发器是最好的解决方案。如果可以，第一次使用 PL 填充。它更容易理解和维护。
DECLARE
Last_date DATE;
Last_User Varchar2(100);
Last_ID number;
Begin
FOR V IN ( Select * from VEHICLE )
LOOP
select max(date) into Last_date
from ALLOCATIONS Last_a
where Last_a.VehicleId = V.id;
IF Last_date is NULL then
Last_User := NULL;
Last_ID := NULL;
else
select Id,UserName into Last_id, Last_user
from ALLOCATIONS Last_a
where Last_a.VehicleId = V.id
and Last_a.date = Last_date;
END IF;
UPDATE Vehicle
set LastAllocationUserName = Last_User
,LastAllocationDate = Last_date
,LastAllocationId Last_id
Where id = V.id;
END LOOP;
End;
警告：写在这里，未经测试。