T-SQL：如何从存储过程返回0行，以及如何使用XACT_ABORT和TRY / CATCH

Question

我正在写一个存储过程，我想在出现故障时返回0条记录。 我似乎无法弄清楚如何只返回0行？ 我用SELECT NULL但这返回1列有NULL行1列1，我也试图不指定任何SELECT在我的错误代码路径声明，但测试的价值时， @@ROWCOUNT后调用的SP，返回1.我认为这可能是因为@@ROWCOUNT从未在SP之前的SELECT语句中重置（在EXISTS() ）。 任何意见，将不胜感激。

此外，我已将XACT_ABORT设置为ON，但我还使用了TRY/CATCH块以确保从存储过程返回正确的错误“返回值”。 这个可以吗？ 如果有错误， XACT_ABORT会覆盖TRY/CATCH或者我的错误代码路径是否仍会导致返回正确的返回值？

-- Setup
SET NOCOUNT ON; -- SET NOCOUNT ON added to prevent extra result sets from interfering with SELECT statements.
SET XACT_ABORT ON; -- SET XACT_ABORT ON rollback transactions on errors
DECLARE @return int; SET @return = 1; -- Default to general error

-- Start transaction
BEGIN TRANSACTION
    BEGIN TRY

        IF NOT EXISTS(SELECT NULL FROM [MyTable] WHERE [Check] = 1) 
        BEGIN

            -- Insert new record    
            INSERT INTO [MyTable] (Check, Date) VALUES (1, GETDATE());
            SELECT SCOPE_IDENTITY() AS [MyValue]; -- Return 1 row
            SET @return = 0; -- Success

        END
        ELSE
        BEGIN

            -- Fail
            SELECT NULL AS [MyValue]; -- Want to return 0 rows not 1 row with NULL
            SET @return = 2; -- Fail error

        END

    END TRY
    BEGIN CATCH

        -- Error
        ROLLBACK TRANSACTION;
        SELECT NULL AS [MyValue]; -- Want to return 0 rows not 1 row with NULL
        SET @return = 1; -- General error

    END CATCH

-- End transaction and return
COMMIT TRANSACTION
RETURN @return;

Answer 1

要返回0行，您可以执行以下操作：

SELECT TOP 0 NULL AS MyValue

就个人而言，我会使用这个sproc的OUTPUT参数来返回ID，而不是返回结果集 - 这只是我的偏好。 然后将该输出参数设置为例如-1作为默认值以指示什么都没做。

Answer 2

这是我怎么做的：

CREATE PROCEDURE YourProcedure
AS
(   @NewMyValue  int OUTPUT   --<<<<<use output parameter and not a result set
)
BEGIN TRY

    --<<<<put everything in the BEGIN TRY!!!

    -- Setup
    SET NOCOUNT ON; -- SET NOCOUNT ON added to prevent extra result sets from interfering with SELECT statements.
    SET XACT_ABORT ON; -- SET XACT_ABORT ON rollback transactions on errors
    DECLARE @return int

    --<<init multiple variables in a select, it is faster than multiple SETs
    --set defaults
    SELECT @return = 1       -- Default to general error
          ,@NewMyValue=NULL

    -- Start transaction
    BEGIN TRANSACTION  --<<<put the transaction in the BEGIN TRY

    --<<<lock rows for this transaction using UPDLOCK & HOLDLOCK hints
    IF NOT EXISTS(SELECT NULL FROM [MyTable] WITH (UPDLOCK, HOLDLOCK) WHERE [Check] = 1)
    BEGIN
            -- Insert new record    
        INSERT INTO [MyTable] (Check, Date) VALUES (1, GETDATE());
        SELECT @NewMyValue=SCOPE_IDENTITY()  --<<<set output parameter, no result set
              ,@return = 0; -- Success
    END
    ELSE
    BEGIN
        -- Fail
        --<<no need for a result set!!! output parameter was set to a default of NULL
        SET @return = 2; -- Fail error
    END

    COMMIT TRANSACTION  --<<<commit in the BEGIN TRY!!!
END TRY
BEGIN CATCH
    -- Error
    IF XACT_STATE()!=0  --<<<only rollback if there is a bad transaction
    BEGIN
        ROLLBACK TRANSACTION
    END
    --<<any insert(s) into log tables, etc
    --<<no need for a result set!!! output parameter was set to a default of NULL
    SET @return = 1; -- General error
END CATCH

-- End transaction and return
RETURN @return;
GO