Question

我有2个表#TblFinal和#TblData ..我需要根据唯一键TblData将＃TblFinal数据与＃34 {＃{1}}进行比较

如果＃F_U_KEY vs D_U_KEY.中不存在唯一键的＃TblData数据，则＃TblFinal数据记录将插入到＃TblData表中。 TblFinal将为1（默认值），F_COUNTER为默认值
如果＃F_IS_CLEAR中存在唯一键的＃TblData数据，那么首先我们从＃TblFinal获得密钥F_ID
< / LI>
案例A：如果表格的＃D_SEV_ID＆＃34;＃TblData＆＃34;不等于-11

然后需要更新＆＃34; F_COUNTER＆＃34; for table＆＃34;＃TblFinal＆＃34; for＆＃34; F_ID＆＃34;

案例B：，如果表格＃34;＃TblData＆＃34;的D_SEV_ID等于-11

然后为D_SEV_ID = -11和F_IS_CLEAR = 1的新条目为＆＃34; F_ID＆＃34;

这是我写的光标，需要一些优化的解决方案，因为我的实际＃TblFinal拥有庞大的数据和数据比较表＆＃34; #TblData＆＃34;总是有100条记录要比较。
```
TblFinal
```

Answer 1

由于数据顺序看起来很重要，因此基于集合的方法看起来不适合您。也就是说，有些事情可以让你的查询速度提高50％左右。

从min（D_ID）循环到max（D_ID），递增1。
只在#TblFinal中查找一次而不是在exists中查找一次，然后再次查找@FId。另外，确保#TblFinal在F_U_KEY + F_IS_CLEAR + F_SEV_ID上有索引，并且F_ID是聚簇索引的一部分或在INCLUDE子句中。理想情况下，它是聚簇索引中第一个快速更新的字段。
通过将D_VAL存储在变量中，您可以直接从变量插入#TblFinal，而不必从#TblData再次进行SELECT。

以下是代码：

DECLARE @D_ID INT
DECLARE @FId INT
DECLARE @D_SEV_ID INT
DECLARE @D_U_KEY NVARCHAR(200)
DECLARE @D_VAL NVARCHAR(20)

DECLARE @Min_D_ID INT, @Max_D_ID INT
SELECT  @Min_D_ID = min(D_ID), @Max_D_ID = max(D_ID) FROM #TblData(NOLOCK)

SELECT @D_ID = @Min_D_ID

WHILE @D_ID <= @Max_D_ID
BEGIN   
    SELECT @D_SEV_ID = D_SEV_ID, @D_U_KEY = D_U_KEY, @D_VAL = D_VAL FROM #TblData(NOLOCK) WHERE D_ID = @D_ID

    IF @D_U_KEY IS NOT NULL
    BEGIN
        SELECT @FId = F_ID FROM #TblFinal(NOLOCK) WHERE F_U_KEY = @D_U_KEY AND F_IS_CLEAR = 0 AND F_SEV_ID <> -11

        IF @FId IS NOT NULL
        BEGIN
            IF (@D_SEV_ID <> -11)
            BEGIN
                UPDATE #TblFinal
                SET F_COUNTER = F_COUNTER + 1 WHERE F_ID = @FId
            END
            ELSE IF(@D_SEV_ID = -11)
            BEGIN
                INSERT INTO #TblFinal (F_VAL, F_SEV_ID, F_U_KEY)
                SELECT  @D_VAL, @D_SEV_ID, @D_U_KEY

                UPDATE #TblFinal
                SET F_IS_CLEAR = 1 WHERE F_ID = @FId
            END
            ELSE
            BEGIN
                PRINT 'DO NOTHING'
            END
        END
        ELSE
        BEGIN
            INSERT INTO #TblFinal (F_VAL, F_SEV_ID, F_U_KEY)
            SELECT  @D_VAL, @D_SEV_ID, @D_U_KEY
        END
    END

    SELECT @D_ID = @D_ID + 1, @FId = NULL
END

Answer 2

你应该能够做到这一点。这是解决这个问题的一种方法。我唯一不理解的部分是F_IS_CLEAR。对此的业务规则有点模糊。如果您需要部分帮助，我需要了解其基本原理。

CREATE TABLE #Output
(F_ID INT IDENTITY(1,1), F_VAL NVARCHAR(20), F_SEV_ID INT, F_U_KEY NVARCHAR(200), F_COUNTER INT DEFAULT(1), F_IS_CLEAR BIT DEFAULT(0))

insert #Output
    select x.D_VAL
    , x.D_SEV_ID
    , x.D_U_KEY
    , x.F_COUNTER
    , x.F_IS_CLEAR
from 
(
    select D_VAL = MIN(D_VAL)
        , D_SEV_ID = MIN(D_SEV_ID)
        , D_U_KEY
        , F_COUNTER = COUNT(*)
        , F_IS_CLEAR = 1
        , SortOrder = 1
    from #TblData d
    where D_ID < (select d2.D_ID from #TblData d2 where d2.D_U_KEY = d.D_U_KEY and d2.D_SEV_ID = -11)
    group by d.D_U_KEY

    UNION ALL

    select D_VAL
        , D_SEV_ID
        , D_U_KEY
        , F_COUNTER = 1
        , F_IS_CLEAR = 0
        , SortOrder = 2
    from #TblData d
    where d.D_SEV_ID = -11

    UNION ALL

    select D_VAL = MIN(D_VAL)
        , D_SEV_ID = MIN(D_SEV_ID)
        , D_U_KEY
        , F_COUNTER = COUNT(*)
        , F_IS_CLEAR = 0
        , SortOrder = 3
    from #TblData d
    where D_ID > (select d2.D_ID from #TblData d2 where d2.D_U_KEY = d.D_U_KEY and d2.D_SEV_ID = -11)
    group by d.D_U_KEY

    UNION ALL

    select D_VAL = MIN(D_VAL)
        , D_SEV_ID = MIN(D_SEV_ID)
        , D_U_KEY
        , F_COUNTER = COUNT(*)
        , F_IS_CLEAR = 0
        , SortOrder = 4
    from #TblData d
    where NOT EXISTS (select d2.D_ID from #TblData d2 where d2.D_U_KEY = d.D_U_KEY and d2.D_SEV_ID = -11)
    group by d.D_U_KEY
) x

order by D_U_KEY, SortOrder

select * from #Output

寻找光标替换

2 个答案: