在 SQL Server 2016 中执行 "Cascading Insert" 是否有更简单的方法?

Is there an easier way to do a "Cascading Insert" in SQL Server 2016?

我继承了 SQL 我需要处理的代码,其设置类似于以下内容:

CREATE TABLE [dbo].[Ni](
    [FooID] [int] IDENTITY(1,1) NOT NULL,
    [Bar] [nvarchar](60) NULL,
    [LocationID] [int] NULL,
    [Thing1] [float] NULL
CONSTRAINT [PK_Ni] PRIMARY KEY CLUSTERED 
(
    [FooID] ASC
);

CREATE UNIQUE NONCLUSTERED INDEX [UQ_LocationBar] ON [dbo].[Ni]
(
    [LocationID] ASC,
    [Bar] ASC
);

CREATE TABLE [dbo].[Ni_Two](
    [ID] [int] IDENTITY(1,1) NOT NULL,
    [FooID] [int] NOT NULL,
    [Thing2] [int] NOT NULL,
    [Thing3] [int] NOT NULL
CONSTRAINT [PK_Ni_Two] PRIMARY KEY CLUSTERED 
(
    [ID] ASC
);

ALTER TABLE [dbo].[Ni_Two]  WITH CHECK ADD  CONSTRAINT [FK_NiTwo_FooID] FOREIGN KEY([FooID])
REFERENCES [dbo].[Ni] ([FooID]);

CREATE TABLE [dbo].[KillMe](
    [ID] [int] NOT NULL,
    [FooID] [int] NULL,
    [Thing4] [int] NOT NULL,
    [Thing5] [int] NOT NULL
PRIMARY KEY CLUSTERED 
(
    [ID] ASC
);

ALTER TABLE [dbo].[KillMe]  WITH NOCHECK ADD  CONSTRAINT [FK_KillMe_FooID] FOREIGN KEY([FooID])
REFERENCES [dbo].[Ni] ([FooID]);


CREATE TABLE [dbo].[PleaseStop](
    [ID] [int] NOT NULL,
    [KillMeID] [int] NOT NULL,
    [Thing7] [int] NOT NULL,
    [Thing8] [int] NOT NULL
PRIMARY KEY CLUSTERED 
(
    [Id] ASC
);

ALTER TABLE [dbo].[PleaseStop]  WITH CHECK ADD  CONSTRAINT [FK_PleaseStop_KillMe] FOREIGN KEY([KillMeID])
REFERENCES [dbo].[KillMe] ([ID]);

问题是这个设计是 [Ni].dbo.[Bar]。该唯一约束是作为要求放在那里的。每个 FooID 都是唯一的,分配给 LocationID 的每个 Bar 也必须是唯一的。

现在要求已经改变了。每个季度导入都会有一些条目,其中 Bar 字段必须更新。

我试过:

   UPDATE dbo.[Ni]
   SET   Bar                 = Imp.Bar
       , LocationID          = Imp.LocationID
       , Thing1              = Imp.Thing1
   FROM dbo.[Ni]
       INNER JOIN ImportData Imp ON [Ni].FooID = Imp.FooID

这会给我一个 Violation of UNIQUE KEY constraint 错误。

我不想更改架构,因为我不知道它会对代码产生什么其他影响。该程序的作者此后离开了公司。 . .我在这里。

作为维护程序的一部分,该程序每季度运行一次(即一年四次)。

我可以在不使用 WHILE 语句的情况下执行此操作吗?因为那会很痛苦。

谢谢!

因此要么在单个查询中更新它们,例如

CREATE TABLE [dbo].[Ni](
    [FooID] [int] IDENTITY(1,1) NOT NULL,
    [Bar] [nvarchar](60) NULL,
    [LocationID] [int] NULL,
    [Thing1] [float] NULL
CONSTRAINT [PK_Ni] PRIMARY KEY CLUSTERED 
(
    [FooID] ASC
)
);

CREATE UNIQUE NONCLUSTERED INDEX [UQ_LocationBar] ON [dbo].[Ni]
(
    [LocationID] ASC,
    [Bar] ASC
);

insert into Ni(bar) values ('a'),('b'),('c');

with newValues as
(
  select * from (values (1,'c'),(3,'x')) newValues (FooId, Bar)
),
toUpdate as
(
  select ni.FooId, ni.Bar, NewValues.Bar NewBar
  from Ni
  join NewValues 
    on ni.FooID = newValues.FooId
)  
update toUpdate set Bar = NewBar

或禁用并重建唯一索引

begin transaction
alter index [UQ_LocationBar] on ni disable
update ni set bar = 'b' where fooid = 1
update ni set bar = 'a' where fooid = 2
alter index [UQ_LocationBar] on ni rebuild
commit transaction

Am I allowed to disable and re-enable the constraint in a stored procedure?

当然,它需要额外的权限,但对 运行 存储过程中的 DDL 没有限制,并且在 SQL 服务器 DDL 是完全事务性的,因此您可以 commit/rollback以防止部分更新并防止其他会话看到部分结果。