我运行了一个小型基准测试 - 在这种情况下，使用 UDF 的方法运行速度慢了近 100 倍.

I ran a small benchmark - in this case the approach with a UDF runs almost 100 times slower.

这是 Sql 代码...

Here's the Sql code...

--建立一个128K行的辅助表Numbers:

-- set up an auxiliary table Numbers with 128K rows:

CREATE TABLE dbo.Numbers(n INT NOT NULL PRIMARY KEY)
GO
DECLARE @i INT;
SET @i = 1;
INSERT INTO dbo.Numbers(n) SELECT 1;
WHILE @i<128000 BEGIN
  INSERT INTO dbo.Numbers(n)
    SELECT n + @i FROM dbo.Numbers;
  SET @i = @i * 2;
END;
GO

-- 表格

CREATE TABLE dbo.Animals
(AnimalId INT NOT NULL IDENTITY PRIMARY KEY,
AnimalType TINYINT NOT NULL, -- 1: Mammal, 2:Reptile, etc..
Name VARCHAR(30))
GO
ALTER TABLE dbo.Animals
ADD CONSTRAINT UNQ_Animals UNIQUE(AnimalId, AnimalType)
GO
CREATE FUNCTION dbo.GetAnimalType(@AnimalId INT)
RETURNS TINYINT
AS
BEGIN
DECLARE @ret TINYINT;
SELECT @ret = AnimalType FROM dbo.Animals
  WHERE AnimalId = @AnimalId;
RETURN @ret;
END
GO
CREATE TABLE dbo.Mammals
(AnimalId INT NOT NULL PRIMARY KEY,
SomeOtherStuff VARCHAR(10),
CONSTRAINT Chk_AnimalType_Mammal CHECK(dbo.GetAnimalType(AnimalId)=1)
);
GO

--- 填充 UDF:

--- populating with UDF:

INSERT INTO dbo.Animals
  (AnimalType, Name)
SELECT 1, 'some name' FROM dbo.Numbers;
GO
SET STATISTICS IO ON
SET STATISTICS TIME ON
GO
INSERT INTO dbo.Mammals
(AnimalId,SomeOtherStuff)
SELECT n, 'some info' FROM dbo.Numbers;

结果是:

SQL Server parse and compile time: 
CPU time = 0 ms, elapsed time = 2 ms.
Table 'Mammals'. Scan count 0, logical reads 272135, 
    physical reads 0, read-ahead reads 0, lob logical reads 0, 
    lob physical reads 0, lob read-ahead reads 0.
Table 'Numbers'. Scan count 1, logical reads 441, physical reads 0, 
    read-ahead reads 0, lob logical reads 0, lob physical reads 0, 
    lob read-ahead reads 0.

SQL Server Execution Times:
    CPU time = 7750 ms,  elapsed time = 7830 ms.

(131072 row(s) affected)

--- 填充 FK:

CREATE TABLE dbo.Mammals2
(AnimalId INT NOT NULL PRIMARY KEY,
AnimalType TINYINT NOT NULL,
SomeOtherStuff VARCHAR(10),
CONSTRAINT Chk_Mammals2_AnimalType_Mammal CHECK(AnimalType=1),
CONSTRAINT FK_Mammals_Animals FOREIGN KEY(AnimalId, AnimalType)
  REFERENCES dbo.Animals(AnimalId, AnimalType)
);

INSERT INTO dbo.Mammals2
(AnimalId,AnimalType,SomeOtherStuff)
SELECT n, 1, 'some info' FROM dbo.Numbers;

结果是:

SQL Server parse and compile time: 
   CPU time = 93 ms, elapsed time = 100 ms.
Table 'Animals'. Scan count 1, logical reads 132, physical reads 0,
    read-ahead reads 0, lob logical reads 0, lob physical reads 0, 
    lob read-ahead reads 0.
Table 'Mammals2'. Scan count 0, logical reads 275381, physical reads 0,
   read-ahead reads 0, lob logical reads 0, lob physical reads 0, 
   lob read-ahead reads 0.
Table 'Numbers'. Scan count 1, logical reads 441, physical reads 0,
   read-ahead reads 0, lob logical reads 0, lob physical reads 0, 
   lob read-ahead reads 0.

SQL Server Execution Times:
   CPU time = 375 ms,  elapsed time = 383 ms.

-- 没有任何完整性的填充:

-- populating without any integrity:

CREATE TABLE dbo.Mammals3
(AnimalId INT NOT NULL PRIMARY KEY,
SomeOtherStuff VARCHAR(10)
);
INSERT INTO dbo.Mammals3
(AnimalId,SomeOtherStuff)
SELECT n,  'some info' FROM dbo.Numbers;

结果是:
SQL Server 解析和编译时间:CPU 时间 = 1 毫秒，经过的时间 = 1 毫秒.

results are:
SQL Server parse and compile time: CPU time = 1 ms, elapsed time = 1 ms.

SQL Server Execution Times:
   CPU time = 0 ms,  elapsed time = 66 ms.
Table 'Mammals3'. Scan count 0, logical reads 272135, physical reads 0,
    read-ahead reads 0, lob logical reads 0, lob physical reads 0,
    lob read-ahead reads 0.
Table 'Numbers'. Scan count 1, logical reads 441, physical reads 0, 
    read-ahead reads 0, lob logical reads 0, lob physical reads 0, 
    lob read-ahead reads 0.

SQL Server Execution Times:
   CPU time = 297 ms,  elapsed time = 303 ms.

(131072 row(s) affected)

FK 在 CPU 时间中的开销 = 375 ms - 297 ms = 78 ms
CPU 时间中 UDF 的开销 = 7750 ms - 297 ms = 7453 ms

The overhead of an FK in CPU time = 375 ms - 297 ms = 78 ms
The overhead of an UDF in CPU time = 7750 ms - 297 ms = 7453 ms