具有 ROW_NUMBER 逻辑问题的嵌套子查询答案

【问题标题】：Nested subqueries with ROW_NUMBER logic problem具有 ROW_NUMBER 逻辑问题的嵌套子查询
【发布时间】：2019-04-22 15:21:45
【问题描述】：

(SQL Server 2008) 我正在尝试从工厂计时系统中过滤掉人工记录。这是下面的数据示例。TCODE 列有一个“O”（开启时间）或一个“OF”（关闭时间）。因此，例如 CHAVEZ 在序列 50（kitpull）上工作了 1.06 小时。

问题：当工作人员完成超过 2 次滑动的作业序列时，系统会自动添加我需要删除的最后一对开关记录（0 小时）。在序列 50 Pull Kit 和序列 400 Countersinking 的情况下，只有两次滑动（打开，然后关闭），因此没有额外的记录。但是，其他每个序列都有两个尾随的额外行。在序列 100 的情况下，TTIME=93724 (09:37:24 AM) 的最后两行是额外的行。您可以看到最后一行的完整代码 1 正式标记作业序列已完成。

MFGORDNO  SEQUENCE  DESCR                 BADGE  LABOR  TCODE  ENAME   TTIME  CompleteCode
M968460   50        Pull Kit              802    0.00   O      CHAVEZ  82300  NULL
M968460   50        Pull Kit              802    1.06   OF     CHAVEZ  92631  0
M968460   100       Load Parts into AJ    396    0.00   O      CURNEY  150014 NULL
M968460   100       Load Parts into AJ    396    1.00   OF     CURNEY  160022 0
M968460   100       Load Parts into AJ    3169   0.00   O      JONES   84612  NULL
M968460   100       Load Parts into AJ    3169   0.85   OF     JONES   93724  0
M968460   100       Load Parts into AJ    3169   0.00   O      JONES   93724  NULL
M968460   100       Load Parts into AJ    3169   0.00   OF     JONES   93724  1
M968460   200       Transfer Drill Holes  3169   0.00   O      JONES   93737  NULL
M968460   200       Transfer Drill Holes  3169   2.73   OF     JONES   132135 0
M968460   200       Transfer Drill Holes  3169   0.00   O      JONES   132135 NULL
M968460   200       Transfer Drill Holes  3169   0.00   OF     JONES   132135 1
M968460   300       TransDrill Splices    3169   0.00   O      JONES   132153 NULL
M968460   300       TransDrill Splices    3169   3.56   OF     JONES   65539  0
M968460   300       TransDrill Splices    3169   0.00   O      JONES   65539  NULL
M968460   300       TransDrill Splices    3169   0.01   OF     JONES   65539  1
M968460   400       Countersinking        3169   0.00   O      JONES   63102  NULL
M968460   400       Countersinking        3169   2.79   OF     JONES   91716  0
M968460   600       SPLICE STRAPS         3169   0.00   O      JONES   131931 NULL
M968460   600       SPLICE STRAPS         3169   1.17   OF     JONES   143040 0
M968460   600       SPLICE STRAPS         3169   0.00   O      JONES   63456  NULL
M968460   600       SPLICE STRAPS         3169   4.12   OF     JONES   105200 0
M968460   600       SPLICE STRAPS         3169   0.00   O      JONES   105200 NULL
M968460   600       SPLICE STRAPS         3169   0.00   OF     JONES   105200 1

期望的结果：简而言之，我正在尝试使用单个 SQL 语句来获得这个结果数据。

MFGORDNO  SEQUENCE  DESCR                 BADGE  LABOR  TCODE  ENAME   TTIME  CompleteCode
M968460   50        Pull Kit              802    0.00   O      CHAVEZ  82300  NULL
M968460   50        Pull Kit              802    1.06   OF     CHAVEZ  92631  0
M968460   100       Load Parts into AJ    396    0.00   O      CURNEY  150014 NULL
M968460   100       Load Parts into AJ    396    1.00   OF     CURNEY  160022 0
M968460   100       Load Parts into AJ    3169   0.00   O      JONES   84612  NULL
M968460   100       Load Parts into AJ    3169   0.85   OF     JONES   93724  0
M968460   200       Transfer Drill Holes  3169   0.00   O      JONES   93737  NULL
M968460   200       Transfer Drill Holes  3169   2.73   OF     JONES   132135 0
M968460   300       TransDrill Splices    3169   0.00   O      JONES   132153 NULL
M968460   300       TransDrill Splices    3169   3.56   OF     JONES   65539  0
M968460   400       Countersinking        3169   0.00   O      JONES   63102  NULL
M968460   400       Countersinking        3169   2.79   OF     JONES   91716  0
M968460   600       SPLICE STRAPS         3169   0.00   O      JONES   131931 NULL
M968460   600       SPLICE STRAPS         3169   1.17   OF     JONES   143040 0
M968460   600       SPLICE STRAPS         3169   0.00   O      JONES   63456  NULL
M968460   600       SPLICE STRAPS         3169   4.12   OF     JONES   105200 0

我最初开始向最里面的子查询添加一个行号，以便稍后可以像这样引用 MAX(rn)：

SELECT
    *, 
    ROW_NUMBER() OVER (ORDER by TTIME) rn 
FROM
    MySmallData

从上面的子查询中，我希望从原始数据中选择所有记录，删除“CompleteCode 中为 1 的行和正上方的行”。

这更像是一个策略问题而不是语法问题。我希望得到一些建议。

谢谢，约翰

编辑：抱歉，这里有一些示例数据：

CREATE TABLE MySmallData 
(
    [MFGORDNO] NVARCHAR(7),
    [SEQUENCE] INT,
    [DESCR] NVARCHAR(20),
    [BADGE] INT,
    [LABOR] NUMERIC(3, 2),
    [TCODE] NVARCHAR(2),
    [TSTAMP] NVARCHAR(26),
    [ENAME] NVARCHAR(19),
    [TTIME] INT,
    [CompleteCode] INT
);

INSERT INTO MySmallData 
VALUES
    ('M968460',0050,'Pull Kit',802,0,'O','2019-02-26 08:23:00.000004','CHAVEZ',82300,NULL),
    ('M968460',0050,'Pull Kit',802,1.06,'OF','2019-02-26 09:26:30.999995','CHAVEZ',92631,0),
    ('M968460',0100,'Load Parts into AJ',396,0,'O','2019-03-05 15:00:13.999997','CURNEY',150014,NULL),
    ('M968460',0100,'Load Parts into AJ',396,1,'OF','2019-03-05 16:00:22.000001','CURNEY',160022,0),
    ('M968460',0100,'Load Parts into AJ',3169,0,'O','2019-03-06 08:46:12.000003','JONES',84612,NULL),
    ('M968460',0100,'Load Parts into AJ',3169,0.85,'OF','2019-03-06 09:37:23.999998','JONES',93724,0),
    ('M968460',0100,'Load Parts into AJ',3169,0,'O','2019-03-06 09:37:23.999998','JONES',93724,NULL),
    ('M968460',0100,'Load Parts into AJ',3169,0,'OF','2019-03-06 09:37:23.999998','JONES',93724,1),
    ('M968460',0200,'Transfer Drill Holes',3169,0,'O','2019-03-06 09:37:37.000001','JONES',93737,NULL),
    ('M968460',0200,'Transfer Drill Holes',3169,2.73,'OF','2019-03-06 13:21:35.000001','JONES',132135,0),
    ('M968460',0200,'Transfer Drill Holes',3169,0,'O','2019-03-06 13:21:35.000001','JONES',132135,NULL),
    ('M968460',0200,'Transfer Drill Holes',3169,0,'OF','2019-03-06 13:21:35.000001','JONES',132135,1),
    ('M968460',0300,'TransDrill Splices',3169,0,'O','2019-03-06 13:21:52.999998','JONES',132153,NULL),
    ('M968460',0300,'TransDrill Splices',3169,3.56,'OF','2019-03-06 16:55:39','JONES',165539,0),
    ('M968460',0300,'TransDrill Splices',3169,0,'O','2019-03-06 16:55:39','JONES',165539,NULL),
    ('M968460',0300,'TransDrill Splices',3169,0.01,'OF','2019-03-06 16:55:39','JONES',165539,1),
    ('M968460',0400,'Countersinking',3169,0,'O','2019-03-07 06:31:01.999998','JONES',63102,NULL),
    ('M968460',0400,'Countersinking',3169,2.79,'OF','2019-03-07 09:17:15.999996','JONES',91716,0),
    ('M968460',0600,'SPLICE STRAPS',3169,0,'O','2019-03-08 13:19:30.999999','JONES',131931,NULL),
    ('M968460',0600,'SPLICE STRAPS',3169,1.17,'OF','2019-03-08 14:30:39.999997','JONES',143040,0),
    ('M968460',0600,'SPLICE STRAPS',3169,0,'O','2019-03-12 06:34:56.000003','JONES',63456,NULL),
    ('M968460',0600,'SPLICE STRAPS',3169,4.12,'OF','2019-03-12 10:52:00.000002','JONES',105200,0),
    ('M968460',0600,'SPLICE STRAPS',3169,0,'O','2019-03-12 10:52:00.000002','JONES',105200,NULL),
    ('M968460',0600,'SPLICE STRAPS',3169,0,'OF','2019-03-12 10:52:00.000002','JONES',105200,1);

【问题讨论】：

标签： sql-server subquery

【解决方案1】：

我认为你可以使用exists：

select msd.*
from mysmalldata msd
where not ((msd.completecode is null and
            exists (select 1
                    from mysmalldata msd2
                    where msd2.MFGORDNO = msd.MFGORDNO and
                          msd2.sequence = msd.sequence and
                          msd2.badge = msd.badge and
                          msd2.ttime = msd.ttime and
                          msd2.completecode = 1
                   )
            ) or
           (coalesce(msd.completecode, -1) = 1 and
            exists (select 1
                    from mysmalldata msd2
                    where msd2.MFGORDNO = msd.MFGORDNO and
                          msd2.sequence = msd.sequence and
                          msd2.badge = msd.badge and
                          msd2.ttime = msd.ttime and
                          msd2.completecode is null
                   )
            )
           );

Here 是一个 dbfiddle。

【讨论】：

感谢@Gordon Linoff。是的，这就是我遇到麻烦的地方。以上仅返回 CompleteCode = 0 的记录。它删除了所有空行。仅应删除与 CompleteCode = 1 行配对的空行。其他空行需要保留在结果中。
@JohnJoseph 。 . .这是我很长一段时间以来不得不发现的更微妙的错误之一。 OR 的第二个条件是返回 NULL，而 completecode 是 NULL。这导致整个表达式为 NULL 并被过滤掉。我已经替换了逻辑。

【解决方案2】：

开始制作我自己的版本，想“不，选择@Gordon，他是怎么做到的”，遇到了一些问题，所以接受它 (+1) 并对其进行了一些修改：

SELECT msd.*
 from mysmalldata msd
 where msd.completecode = 0  --  All "completeCode=0" rows are accepted
  or (--  All "NULL" rows for which there is no matching "completeCode=1" row are accepted
      msd.completecode is null
       and not exists (select 1
                        from mysmalldata msd2
                         where msd2.MFGORDNO = msd.MFGORDNO
                          and msd2.sequence = msd.sequence
                          and msd2.badge = msd.badge
                          and msd2.ttime = msd.ttime
                          and msd2.completecode = 1
                      )
     )
 --  Note that no "completeCode=1" rows are accepted

【讨论】：

@PhillipKelley 也很棒。不错的减法。感谢您的帮助
未来读者：@Gordon 已经根据我的修改修改了他的代码。一种价格的两种可能的解决方案！