组合表变量的更新答案

【问题标题】：Combining Updates on Table Variable组合表变量的更新
【发布时间】：2018-06-20 22:01:12
【问题描述】：

如果可能的话，我想知道如何组合这两个 UPDATE 语句以及这是否会节省性能。查询目前在 3 秒内执行，但会继续增长。

如果您知道我的 CASE 和 WHERE 语句中有更简洁、更优雅的过滤，则可以加分。

谢谢

DECLARE @PartRevision TABLE
( 
OriginalPart VARCHAR(50) NULL,
PartNum VARCHAR(50) NULL,
Revision VARCHAR(50) NULL,
[Language] VARCHAR(50) NULL,
Hierarchy VARCHAR(50) NULL
)
INSERT INTO @PartRevision
(
    OriginalPart,
    PartNum,
    Revision,
    [Language],
    Hierarchy
)
SELECT
part_no
,''
,''
,''
,''
FROM dbo.PartMaster
;

UPDATE @PartRevision
SET PartNum = SUBSTRING(OriginalPart,1,6) 
    ,Revision = SUBSTRING(RIGHT(OriginalPart, LEN(OriginalPart) - 6),1,2)
    ,[Language] = CASE 
                    WHEN LEN(RIGHT(OriginalPart, LEN(OriginalPart) - 6)) = 2 THEN ''
                    WHEN RIGHT(OriginalPart, LEN(OriginalPart) - 6) = 'DEV' THEN ''
                    ELSE RIGHT(RIGHT(OriginalPart, LEN(OriginalPart) - 6), (LEN(RIGHT(OriginalPart, LEN(OriginalPart) - 6)) -2))
                    END
FROM @PartRevision
--business designed/approved filter--
WHERE LEFT(OriginalPart, 1) NOT IN ('a', 'b', 'c', 'd', 'e', 'f', 'g', 'h', 'i', 'j', 'k', 'l', 'm', 'n', 'o', 'p', 'q', 'r', 's', 't', 'u', 'v', 'w', 'x', 'y', 'z')
         AND OriginalPart NOT IN ('40T', '45T', '46AJ', '2044', '16NA', '12NA', '86TCE', '86TNA', '86TAS', '82TNA', '66TNA', '643TH', '843TH', '99000', '61TNA', '3215NA', '4626NA', '7127NA', '9250NA', '63AJNA', '1056TH', '8841NA', '3219NA', '1044TH', '3226NA', '4632NA', '4740NA', '8831NA', '208922', '7135NA', '85AJNA', '3220NA', '201975B', '201520A', '200989B', '109271B', '1044THS', '200465B', '202207A', '201436A', '2PREPAY', '200545A', '1056THS', '111169B', '30AJENA', '202204B', '202140B', '200551B', '201452B', '310142B', '2p5 ARA', '202100A', '111170B', '218999A', '1256THS', '202061B', '107212F', '202698A', '201974B', '201114B')       
;
UPDATE @PartRevision
SET PartNum = CONCAT(PartNum, [Language])
FROM @PartRevision 
;

SELECT * FROM @PartRevision

【问题讨论】：

LEN(RIGHT(OriginalPart, LEN(OriginalPart) - 6)) = 2 很像LEN(OriginalPart) - 6) = 2，不是吗？
LEFT(OriginalPart, 1) NOT IN ('a', 'b', 'c', 'd', 'e', 'f', 'g', 'h', 'i', 'j', 'k', 'l', 'm', 'n', 'o', 'p', 'q', 'r', 's', 't', 'u', 'v', 'w', 'x', 'y', 'z') 可以写成OriginalPart not like '[a-z]%'。不确定它会对性能产生什么影响。
具有适当索引的临时表，例如在OriginalPart 上，相对于没有索引的表变量，可能会提高性能。
RIGHT(RIGHT(OriginalPart, LEN(OriginalPart) - 6), (LEN(RIGHT(OriginalPart, LEN(OriginalPart) - 6)) -2)) 像之前的评论一样受到一些冗余的影响。您不需要继续创建子字符串（使用Right()）只是为了获取它的长度（使用Len()）并执行一些算术。开始解开它，看看你能不能把它归结为一个Right() 和一个Len()。（我的第一条评论有一个额外的) 浮动。我的错。）
太棒了，感谢@HABO的提示

标签： tsql ssms-2016

【解决方案1】：

最终结果。解决了这个问题，但我仍然对优化更新感到好奇。

BEGIN
DECLARE @PartRevision TABLE
( 
OriginalPart VARCHAR(50) NULL,
PartNum VARCHAR(50) NULL,
Revision VARCHAR(50) NULL,
[Language] VARCHAR(50) NULL,
Hierarchy VARCHAR(50) NULL
)

DECLARE @HierLookup TABLE
(
Hierarchy INT IDENTITY(1,1) PRIMARY KEY
,RevisionCode VARCHAR(10) NOT NULL
)

DECLARE @RevisionGen TABLE
(
Alphabet VARCHAR(10) NOT NULL,
BaseRevision VARCHAR(10) NULL
)

DECLARE @iAlphabet INT = 65
DECLARE @iBaseRevision INT = 65

--Start build of Hierarchy Lookup Table (@HierLookup)
--INSERT A-Z into @RevisionGen
WHILE @iAlphabet < 91
BEGIN
    INSERT INTO @RevisionGen
    (
        Alphabet
    )
    VALUES
    (
        CHAR(@iAlphabet)
    )
    SET @iAlphabet = @iAlphabet + 1;
END
;

--Final INSERTS and UPDATES to @HierLookup
WHILE @iBaseRevision < 91
BEGIN
    UPDATE @RevisionGen
    SET BaseRevision = CHAR(@iBaseRevision);
    ;
    INSERT INTO @HierLookup
    (
        RevisionCode
        --,Hierarchy
    )
    SELECT CONCAT(BaseRevision, Alphabet)
    FROM @RevisionGen;
    SET @iBaseRevision = @iBaseRevision + 1
    ;
END
;

--Starting INSERTS and UPDATES to @PartRevision
--Get base Part data from ETL Data Warehouse
INSERT INTO @PartRevision
(
    OriginalPart,
    PartNum,
    Revision,
    [Language],
    Hierarchy
)
SELECT
PART_NO
,''
,''
,''
,''
FROM man.Stage_Gue_InventoryPart
;

--Start parsing Parts data
UPDATE @PartRevision
SET PartNum = SUBSTRING(OriginalPart,1,6) 
    ,Revision = SUBSTRING(RIGHT(OriginalPart, LEN(OriginalPart) - 6),1,2)
    ,[Language] = CASE 
                    WHEN LEN(OriginalPart) - 6 = 2 THEN ''
                    WHEN RIGHT(OriginalPart, LEN(OriginalPart) - 6) = 'DEV' THEN ''
                    ELSE RIGHT(RIGHT(OriginalPart, LEN(OriginalPart) - 6), (LEN(OriginalPart) - 6) -2)
                    END
FROM @PartRevision
--business approved filter--
WHERE LEFT(OriginalPart, 1) NOT LIKE '[a-z]%'
         AND OriginalPart NOT IN ('40T', '45T', '46AJ', '2044', '16NA', '12NA', '86TCE', '86TNA', '86TAS', '82TNA', '66TNA', '643TH', '843TH', '99000', '61TNA', '3215NA', '4626NA', '7127NA', '9250NA', '63AJNA', '1056TH', '8841NA', '3219NA', '1044TH', '3226NA', '4632NA', '4740NA', '8831NA', '208922', '7135NA', '85AJNA', '3220NA', '201975B', '201520A', '200989B', '109271B', '1044THS', '200465B', '202207A', '201436A', '2PREPAY', '200545A', '1056THS', '111169B', '30AJENA', '202204B', '202140B', '200551B', '201452B', '310142B', '2p5 ARA', '202100A', '111170B', '218999A', '1256THS', '202061B', '107212F', '202698A', '201974B', '201114B', '216610')         
;

--Final update to Part Number
UPDATE @PartRevision
SET PartNum = CONCAT(PartNum, [Language])
FROM @PartRevision 
;

--INSERT Hierarchy numbers from @HierLookup table
UPDATE @PartRevision
SET Hierarchy = CASE 
                WHEN l.Hierarchy = '' THEN 1
                ELSE l.Hierarchy
                END 
FROM @PartRevision AS p
LEFT JOIN @HierLookup AS l
ON l.RevisionCode = p.Revision
;

--UPDATE ETL, Pre-Merge Table
UPDATE man.Stage_Gue_InventoryPart
SET PART_NO = CASE 
                WHEN PartNum = '' THEN OriginalPart
                ELSE PartNum
                END,
    ENG_REVISION = Revision
    --REV_SEQ = Hierarchy   
FROM @PartRevision
WHERE PART_NO = OriginalPart
;

END

【讨论】：