0

“进入@gutsTVP”之后的左括号打破了这一点。
错误在第 12 行的正确括号上。
如果删除它运行的第一个联合周围的第一组括号，但它返回错误的答案，因为然后在第一个联合之前应用了相交

DECLARE @gutsTVP AS TABLE ( sID INT PRIMARY KEY); 
insert into @gutsTVP
(
        select distinct [ftsIndexWordOnce].[sID] 
        from [ftsIndexWordOnce] with (nolock)
        where [ftsIndexWordOnce].[wordID] in (1,2)
    union
        select distinct [ftsIndexWordOnceB].[sID] 
        from [ftsIndexWordOnceB] with (nolock)
        where [ftsIndexWordOnceB].[wordID] in (5,6)
)
intersect
(
        select distinct [ftsIndexWordOnce].[sID] 
        from [ftsIndexWordOnce] with (nolock)   
        where [ftsIndexWordOnce].[wordID] in (9,10,11,12)
    union
        select distinct [ftsIndexWordOnceB].[sID] 
        from [ftsIndexWordOnceB] with (nolock)  
        where [ftsIndexWordOnceB].[wordID] in (13,14,15,16)
)
select [guts].[sID] from @gutsTVP as [guts] 
join [docSVsys] with (nolock)
  on [docSVsys].[sID] = [guts].[sID]
order by [docSVsys].[sParID], [docSVsys].[sID]

包括最终连接，因为这是 TABLE 的目的。
如果我加入派生表，它不知道 PK 并且慢得多。

是的，我知道可以将其重新考虑为没有领先的左括号。
这是一个简化的查询。
需要能够处理前导左括号。

下面的技巧使语法起作用。
但现在加入速度较慢，因为它不知道 PK。
不能将 sID 作为 PK 并接受 null

DECLARE @gutsTVP AS TABLE ( sID INT PRIMARY KEY); 
insert into @gutsTVP
select 0  -- real PK starts at 1
union
(     
        select distinct [ftsIndexWordOnce].[sID] 
        from [ftsIndexWordOnce] with (nolock)
        where [ftsIndexWordOnce].[wordID] in (1,2)
    union
        select distinct [ftsIndexWordOnceB].[sID] 
        from [ftsIndexWordOnceB] with (nolock)
        where [ftsIndexWordOnceB].[wordID] in (5,6)
)
intersect
(
        select distinct [ftsIndexWordOnce].[sID] 
        from [ftsIndexWordOnce] with (nolock)   
        where [ftsIndexWordOnce].[wordID] in (9,10,11,12)
    union
        select distinct [ftsIndexWordOnceB].[sID] 
        from [ftsIndexWordOnceB] with (nolock)  
        where [ftsIndexWordOnceB].[wordID] in (13,14,15,16)
)
select [guts].[sID] from  @gutsTVP as [guts]
join [docSVsys] with (nolock)
  on [docSVsys].[sID] = [guts].[sID]
where [guts].[sID] > 0
order by [docSVsys].[sParID], [docSVsys].[sID]
4

1 回答 1

2

改为尝试 CTE：

DECLARE @gutsTVP AS TABLE ( sID INT PRIMARY KEY); 
; WITH a AS (
        select [ftsIndexWordOnce].[sID] 
        from [ftsIndexWordOnce] with (nolock)
        where [ftsIndexWordOnce].[wordID] in (1,2)
    union
        select [ftsIndexWordOnceB].[sID] 
        from [ftsIndexWordOnceB] with (nolock)
        where [ftsIndexWordOnceB].[wordID] in (5,6)
)
, b AS (
        select [ftsIndexWordOnce].[sID] 
        from [ftsIndexWordOnce] with (nolock)   
        where [ftsIndexWordOnce].[wordID] in (9,10,11,12)
    union
        select [ftsIndexWordOnceB].[sID] 
        from [ftsIndexWordOnceB] with (nolock)  
        where [ftsIndexWordOnceB].[wordID] in (13,14,15,16)
)
insert into @gutsTVP
SELECT * FROM a
    intersect
SELECT * FROM b;

select [guts].[sID] from @gutsTVP as [guts] 
join [docSVsys] with (nolock)
  on [docSVsys].[sID] = [guts].[sID]
order by [docSVsys].[sParID], [docSVsys].[sID]

另请注意，您可以丢弃distincts，因为 s 已经暗示了它们union。

于 2013-05-12T19:56:31.903 回答