SQL Server中的Try/Catch块不处理错误

Question

我们在SQL Server存储过程中使用以下错误处理模式:

ALTER PROCEDURE [dbo].[USP_Districtdata_Import]
    @DistrictData DistrictData Readonly
AS
BEGIN
    SET NOCOUNT ON;
    BEGIN TRY
    BEGIN TRAN

    --Insert the new records into BudgetDistrict Table.
    INSERT INTO [dbo].[BudgetDistrict]
    (
     DistrictID,
     [Year],
     Season,
     BudgetStateID,
     ProjectedReturnCountIsCalc,
     RowStatus,
     CreatedBy,
     CreatedDate,
     LastModifiedBy,
     LastModifiedDate,
     EnableBudgetLock
     )

    SELECT
    DISTINCT list.[District Id],list.[Year],list.[Season],1,0,'A',@CreatedBy,@Updtime,@CreatedBy,@Updtime,0
    FROM @DistrictData liston]
        AND bud.RowStatus = 'A'
        )
    LEFT OUTER JOIN [dbo].[BudgetDistrict] bud
      ON (bud.DistrictID  = list.[District Id]
        AND bud.[Year]    = list.[Year]
        AND bud.[Season]  = list.[Seas
    WHERE bud.DistrictID IS NULL


    --Update the existing pending budgets
    UPDATE wk
    SET  wk.Budget         = list.[Budget], 
         wk.BudgetAdjusted = list.[Budget],
         wk.ProjectedReturnCount = list.[ProjectedReturn Count], 
         wk.CreatedBy      = @CreatedBy, 
         wk.CreatedDate    = @Updtime,
         wk.LastModifiedBy = @CreatedBy, 
         wk.LastModifiedDate = @Updtime  
    FROM @DistrictData list
    INNER JOIN [dbo].[BudgetDistrict] bud
      ON (bud.DistrictID  = list.[District Id]
        AND bud.[Year]    = list.[Year]
        AND bud.[Season]  = list.[Season])
    INNER JOIN [dbo].[BudgetDistrictWeekly] wk
      ON (wk.NationalBudgetID = bud.BudgetDistrictID
       AND wk.[WeekDate]      = list.[Week])
    WHERE bud.RowStatus  = 'A'
       AND wk.RowStatus  = 'A'
       AND bud.BudgetStateID = 1

    --Insert the new budgets
    INSERT INTO [dbo].[BudgetDistrictWeekly]
    (
    WeekDate,
    Budget,
    BudgetAdjusted,
    RowStatus,
    CreatedBy,
    CreatedDate,
    LastModifiedBy,
    LastModifiedDate,
    ProjectedReturnCount
    )

    SELECT LIST.[Week],list.[Budget],list.[Budget],'A',@CreatedBy,@Updtime,@CreatedBy,@Updtime,[ProjectedReturn Count]
    FROM @DistrictData list
    LEFT JOIN [dbo].[BudgetDistrict] bud
      ON (bud.DistrictID  = list.[District Id]
        AND bud.[Year]    = list.[year]
        AND bud.[Season]  = list.Season
        AND bud.RowStatus = 'A')
    WHERE bud.DistrictID IS NULL

       IF @@ERROR = 0
       BEGIN
              COMMIT TRAN;
      END
    END TRY
    BEGIN CATCH
        SELECT 
            ERROR_NUMBER() AS ErrorNumber,  
            ERROR_LINE() AS ErrorLine, 
            ERROR_MESSAGE() AS ErrorMessage;
        ROLLBACK TRAN;
    END CATCH

SET NOCOUNT OFF;
END

但是当存储过程中出现以下错误时，try/catch块不起作用。

错误详细信息：存储过程试图将NULL值插入非NULL列。

在执行存储过程中，我得到了以下错误

EXECUTE之后的事务计数表示BEGIN和COMMIT语句的数量不匹配。上一个计数=1，当前计数=0。

msg 3903， Level 16， State 1， Line 30
ROLLBACK TRANSACTION请求没有对应的BEGIN TRANSACTION。

为什么不处理异常？请帮忙

Answer 1

有一种方法，如果你能够使用动态sql，那么尝试捕获工作

Answer 2

根据留档（错误不受TRY… CATCH构造部分的影响），有些错误不会被CATCH语句捕获。

尤其：

• 阻止批处理运行的编译错误，如语法错误
• 语句级重新编译过程中发生的错误，如编译后由于延迟名称解析而发生的对象名称解析错误

非常典型的情况是，存储过程访问在创建存储过程后删除的表列（或访问对象）。

请参见下面的示例:

create table #test (id int, somecolumn uniqueidentifier)
GO
create procedure #testSP
as
begin try
    set nocount on;

    begin tran;

    insert into #test (id, somecolumn)
    values (1, 0x);

    commit;

end try
begin catch
    rollback;
    print 'Error happened';
end catch
GO
exec #testSP
GO
alter table #test
    drop column somecolumn
GO
exec #testSP
GO

上述代码生成

消息 207，级别 16，状态 1，过程 #testSP_..._00053EAF，第 33 行列名称“某个列”无效。

消息 266，级别 16，状态 2，过程 #testSP_..._00053EAF，执行后的第 33 行事务计数指示 BEGIN 和 COMMIT 语句的数量不匹配。上一个计数 = 1，当前计数 = 2。

如您所见，尽管严重性级别为 16，但不会捕获错误，并且不会打印消息“发生错误”。