T-SQL: If parameter1 is null, use parameter2

Question

This is a simple one but I can't seem to figure it out.

I have two parameters, 1 and 2. If one parameter is passed a null value, use the other parameter in there WHERE clause.

Ex:

SELECT ...
 WHERE CASE
    WHEN @Parameter1 IS NULL
    THEN Field2 = @Parameter2
    WHEN @Parameter2 IS NULL
    THEN Field1 = @Parameter1
  END

I know this is something simple or maybe I'm using my logic wrong. Any help or direction is appreciated.

Answer 1

The simplest answer to your question involves telling SQL Server what to look for by placing brackets around grouped requirements. So, for instance, if we only want to check Field2 when @parameter1 IS NULL, we could do this:

WHERE (@parameter1 IS NULL AND Field2 = @parameter2)
    OR (Field1 = @parameter1)

SQL Server will evaluate Field2 = @parameter2 when @parameter1 IS NULL. When '@Parameter1` is not null SQL Server will only return rows that match it.

If you have a lot of rows in the source table, you may want to use a stored procedure to help optimize the possible choices. The following example creates a table, populates a couple of rows, and creates a procedure to query the table.

I'm doing this in tempdb since this is just an example:

USE tempdb;

Create the table, and populate it with a couple of rows:

IF OBJECT_ID('dbo.SomeTable') IS NOT NULL
DROP TABLE dbo.SomeTable;
GO
CREATE TABLE dbo.SomeTable
(
    SomeID INT NOT NULL
        CONSTRAINT PK_SomeTable
        PRIMARY KEY CLUSTERED
        IDENTITY(1,1)
    , Column1 NVARCHAR(30) NULL
    , Column2 NVARCHAR(30) NULL
    , Column3 NVARCHAR(200) NOT NULL
);
INSERT INTO dbo.SomeTable (Column1, Column2, Column3)
VALUES (NULL, 'col2_value', 'some data for row 1')
    , ('col1_value', NULL, 'some data for row 2');

Since I know the precise patterns my queries against this table will take, I can create a couple of simple covering indexes that will allow SQL Server to very quickly find the relevant rows:

CREATE INDEX IX_SomeTable_c1 ON dbo.SomeTable(Column1) INCLUDE (Column2, Column3);
CREATE INDEX IX_SomeTable_c2 ON dbo.SomeTable(Column2) INCLUDE (Column1, Column3);
GO

This will add an extra 10,000 rows to the table just to give us enough data for the indexes to be considered by the query optimizer:

INSERT INTO dbo.SomeTable (Column1, Column2, Column3)
SELECT TOP(10000) LEFT(o1.name, 30), RIGHT(o2.name, 30), o3.name
FROM sys.objects o1
    , sys.objects o2
    , sys.objects o3

Here, I'm creating a procedure that will search our table using the contents of either Column1 or Column2 depending on the parameters passed in.

IF OBJECT_ID('dbo.MySearch') IS NOT NULL
DROP PROCEDURE dbo.MySearch;
GO
CREATE PROCEDURE dbo.MySearch
(
    @parameter1 NVARCHAR(30) = NULL
    , @parameter2 NVARCHAR(30) = NULL
)
AS
BEGIN
    SET NOCOUNT ON;
    DECLARE @sql NVARCHAR(MAX);
    DECLARE @where NVARCHAR(100);
    DECLARE @params NVARCHAR(100);
    SET @sql = 'SELECT t.Column1
    , t.Column2
    , t.Column3
FROM dbo.SomeTable t
';
    IF @parameter1 IS NULL
    BEGIN
        SET @params = '@p2 NVARCHAR(30)';
        SET @where = 'WHERE t.Column2 = @p2';
        SET @sql = @sql + @where;
        EXEC sys.sp_executesql @sql, @params, @p2 = @parameter2;
    END
    IF @parameter2 IS NULL
    BEGIN
        SET @params = '@p1 NVARCHAR(30)';
        SET @where = 'WHERE t.Column1 = @p1';
        SET @sql = @sql + @where;
        EXEC sys.sp_executesql @sql, @params, @p1 = @parameter1;
    END
END;
GO

If you pass in values for both @parameter1 and @parameter2, the stored procedure will return no rows. You may want to add a third IF statement to the stored procedure to handle that situation, something like:

    IF @parameter1 IS NOT NULL AND @parameter2 IS NOT NULL
    BEGIN
        SET @params = '@p1 NVARCHAR(30), @p2 NVARCHAR(30)';
        SET @where = 'WHERE t.Column1 = @p1 AND t.Column2 = @p2';
        SET @sql = @sql + @where;
        EXEC sys.sp_executesql @sql, @params, @p1 = @parameter1, @p2 = @parameter2;
    END

Without knowing your exact requirements, it's difficult to say if you need that.

Two samples that show the usage of the stored proc, along with the execution plans:

EXEC dbo.MySearch @parameter1 = 'col1_value';

EXEC dbo.MySearch @parameter2 = 'col2_value';

Using a stored procedure in combination with dynamic SQL allows the query designer (us) to help SQL Server consistently choose the best plan for the job by providing two independent cachable plans that are optimized for each scenario. Depending on the cardinality of each column involved, the complexity of the actual query design, along with the presence of desirable indexes, SQL Server may decide to create vastly different plans for the two queries, which can be a Very Good Thing™.

Just for fun, using the "simple" WHERE clause I show at the top of my answer like this:

DECLARE @parameter1 NVARCHAR(30) = 'col1_value';
DECLARE @parameter2 NVARCHAR(30);
SELECT t.Column1
    , t.Column2
    , t.Column3
FROM dbo.SomeTable t
WHERE (@parameter1 IS NULL AND t.Column2 = @parameter2)
    OR (t.Column1 = @parameter1);

Results in a plan like this:

This clearly shows SQL Server searching both indexes for Column1 and Column2. This is rather obviously not optimal, and would certainly not be the preferred route for larger sets of data.

Answer 2

如果我理解你，这应该可以完成工作：

SELECT ...
 WHERE 
    (Field2 = @Parameter2 AND @Parameter1 IS NULL) OR 
    (Field1 = @Parameter1 AND @Parameter2 IS NULL) 

但是，由于潜在的性能问题，我宁愿不做这样的事情。此查询的执行计划在很大程度上取决于参数值。例如，传递NULLas @Parameter1should使优化器倾向于在 Field2 上使用索引，传递NULL为 @Parameter2` - Field1 上的索引。

为了克服这个问题，您需要在每次查询运行时强制重新编译。即使 SQL Server 的更高版本具有类似于 Oracle 绑定感知的功能，拥有 2 个单独的查询也比依赖优化器魔术要容易得多。

例如，您可以修改您的过程/函数并将逻辑放在那里：

IF  @Parameter1 IS NULL 
 BEGIN
    SELECT ... WHERE Field2 = @Parameter2;
    ...
 END;
ELSE 
 BEGIN
    SELECT ... WHERE Field1 = @Parameter1;
    ...
 END;

进一步阅读：

Erland Sommarskog在 T?SQL 中的动态搜索条件