SQL中涉及分组和平均的复杂查询的逻辑

我有两张桌子。
以下是获得所需输出所需的步骤。我可以继续第三步。请帮助我，因为这有点复杂，我无法理解如何进一步进行

表1

Site  Code FailFlag  Comments  ModifiedDate  ModifiedBy
ABT   A01     F      Dfasdf    10/11/2011    Anna
ABT   A01     F      dsfsdf    15/12/2012    Mand
ABT   A01            Rds       30/03/2011    Tim
ABT   A01            GHDs      02/12/2012    Andy
ABT   A02     F      dfd       09/05/2012    Anna
ABT   A02            sdada     11/02/2013    Kathy
ABT   A02            Dfg       15/05/2011    Rob
AFL   A02     F      asda      13/02/2011    Dan
AFL   A02            dsaa      24/12/2010    Ryan
TRG   A01            sdasd     16/04/2010    Richard
TRG   K05            jksdh     23/04/2012    Mark
KLD   K05     F      sd        18/05/2013    Jim
KLD   K05            dsfsd     10/03/2012    James
KLD   K05            sdsd      12/05/2011    Luther
KTY   K05     F      saq       09/09/2012    Ryan
KTY   K05            asd       04/04/2010    Kathy
KMD   C02     F      nas       29/02/2012    Rob
KMD   C02            asda      11/11/2011    Andy

表2：

Site  Code   Freq     StartDate   EndDate
ABT   A01    43       01/01/2011  01/02/2012
ABT   A02    254      01/01/2011  19/02/2011
ABT   A02    109      20/02/2011  01/01/2012
ABT   A02    12       02/01/2012  01/01/2013
AFL   A02    13       01/01/2011  01/02/2012
TRG   A01    122      01/01/2011  01/02/2012
TRG   K05    61       01/01/2011  01/02/2012
KLD   KO5    33       01/01/2011  15/05/2012
KLD   K05    79       16/05/2012  01/01/2013
KTY   K05    52       01/01/2011  01/02/2012
KMD   C02    78       01/01/2011  01/02/2012
ZYT   G01    11       01/01/2011  01/02/2012
PYN   A01    15       01/01/2011  01/02/2012
DYN   F08    122      01/01/2011  01/02/2012

步骤：

表1中“站点”和“代码”两列的组合参见表2中“站点”和“代码”两列的组合

在“Failure”列中过滤相同的内容，并找出失败的数量

以下是查询和输出：

SELECT Site,Code,COUNT(*) as [Count],
FailFlagCount= SUM(CASE WHEN F = 'F' THEN 1 ELSE 0 END) FROM Table1 
GROUP BY Site,Det


Site Code    Count   FailFlagCount
ABT   A01     4      2
ABT   A02     3      1
AFL   A02     2      1
TRG   A01     1      0
TRG   K05     1      0      
KLD   KO5     3      1
KTY   K05     2      1
KMD   C02     2      1

我们在表2中检查相同的组合。i、 例如，步骤2输出的位置和代码见表2，以获得其频率

计算：

a。CC%=[1-（故障计数/计数）]*100=[1-（2/4）]*100

b。B.P.O%=[1-（故障计数/频率）]*100=[1-（2/43）]*100

c、 预测%=

让我们假设当前月份是三月

计算过去3年的故障次数并找出平均值，比如“X”。
计算剩余月份的故障，“Y”=（X*剩余月份/12）
超过12个月的总故障=当前故障+Y

因此，预测%=[1-（12个月内的总故障/频率）]*100

在我们的示例中，假设X=2 因此Y=（2*9/12）

预测%=[1-（1+1.5）]*100

上述计算将针对现场和规范的所有可能组合进行

Site    Code    CC  B.P.O   Forecast   StartDate    EndDate
ABT   A01                                  01/01/2011  01/02/2012
ABT   A02                                  01/01/2011  19/02/2011
ABT   A02                                  20/02/2011  01/01/2012
ABT   A02                                  02/01/2012  01/01/2013
AFL   A02                                  01/01/2011  01/02/2012
TRG   A01                                  01/01/2011  01/02/2012
TRG   K05                                  01/01/2011  01/02/2012
KLD   K05                                  01/01/2011  15/05/2012
KLD   K05                                  16/05/2012  01/01/2013
KTY   K05                                  01/01/2011  01/02/2012
KMD   C02                                  01/01/2011  01/02/2012

上表的分组根据场地类型进行，即场地的第一个字母。 然后在分组后对所有计算（cc、BPO、Forecast）进行平均 例如：-“A”代表“ABT”，T代表“TRG”。 （我假设我们为多种类型创建多个表，然后进行联合以获得下面的查询）
例如：

Site     Code                             CC     B.P.O    Forecast          
 A    A01  
 A    A02 
    [i.e.,Avg value of (ABT and A02) 
         and (AFL and A02)]  
 T    A01 
 T    K05 avg
    [i.e., Avg value of (KLD and K05) 
         and (KTY and K05)]  
 K    K05 
 K    C02 

这应该是我的最终输出。请帮忙

添加表的脚本以使帮助更容易：

表1脚本：

CREATE TABLE Table1
    (
    ID INT IDENTITY(1,1) PRIMARY KEY CLUSTERED,
    Site VARCHAR(5),
    Code VARCHAR(5),
    FailFlag CHAR(1),
    Comments VARCHAR(100),
    ModifiedDate  DATETIME,
    ModifiedBy   VARCHAR(50)
    );

INSERT INTO Table1 
      (Site,  Code,  FailFlag, Comments, ModifiedDate, ModifiedBy)
SELECT 'ABT', 'A01',   'F',    'Dfasdf', '10/11/2011', 'Anna'    UNION ALL
SELECT 'ABT', 'A01',   'F',    'dsfsdf', '15/12/2012', 'Mand'    UNION ALL
SELECT 'ABT', 'A01',   NULL,   'Rds',    '30/03/2011', 'Tim'     UNION ALL
SELECT 'ABT', 'A01',   NULL,   'GHDs',   '02/12/2012', 'Andy'    UNION ALL
SELECT 'ABT', 'A02',   'F',    'dfd',    '09/05/2012', 'Anna'    UNION ALL
SELECT 'ABT', 'A02',   NULL,   'sdada',  '11/02/2013', 'Kathy'   UNION ALL
SELECT 'ABT', 'A02',   NULL,   'Dfg',    '15/05/2011', 'Rob'     UNION ALL
SELECT 'AFL', 'A02',   'F',    'asda',   '13/02/2011', 'Dan'     UNION ALL
SELECT 'AFL', 'A02',   NULL,   'dsaa',   '24/12/2010', 'Ryan'    UNION ALL
SELECT 'TRG', 'A01',   NULL,   'sdasd',  '16/04/2010', 'Richard' UNION ALL
SELECT 'TRG', 'K05',   NULL,   'jksdh',  '23/04/2012', 'Mark'    UNION ALL
SELECT 'KLD', 'K05',   'F',    'sd',     '18/05/2013', 'Jim'     UNION ALL
SELECT 'KLD', 'K05',   NULL,   'dsfsd',  '10/03/2012', 'James'   UNION ALL
SELECT 'KLD', 'K05',   NULL,   'sdsd',   '12/05/2011', 'Luther'  UNION ALL
SELECT 'KTY', 'K05',   'F',    'saq',    '09/09/2012', 'Ryan'    UNION ALL
SELECT 'KTY', 'K05',   NULL,   'asd',    '04/04/2010', 'Kathy'   UNION ALL
SELECT 'KMD', 'C02',   'F',    'nas',    '29/02/2012', 'Rob'     UNION ALL
SELECT 'KMD', 'C02',   NULL,   'asda',   '11/11/2011', 'Andy';

表2脚本：

CREATE TABLE Table2
    (
    ID INT IDENTITY(1,1) PRIMARY KEY CLUSTERED,
    Site VARCHAR(5),
    Code VARCHAR(5),
    Freq  int,
    StartDate  DATETIME,
    EndDate  DATETIME
    );

INSERT INTO Table2 (Site, Code, Freq, StartDate, EndDate)
SELECT 'ABT', 'A01', 43,  '01/01/2011', '01/02/2012' UNION ALL
SELECT 'ABT', 'A02', 254, '01/01/2011', '19/02/2011' UNION ALL
SELECT 'ABT', 'A02', 109, '20/02/2011', '01/01/2012' UNION ALL
SELECT 'ABT', 'A02', 12,  '02/01/2012', '01/01/2013' UNION ALL
SELECT 'AFL', 'A02', 13,  '01/01/2011', '01/02/2012' UNION ALL
SELECT 'TRG', 'A01', 122, '01/01/2011', '01/02/2012' UNION ALL
SELECT 'TRG', 'K05', 61,  '01/01/2011', '01/02/2012' UNION ALL
SELECT 'KLD', 'KO5', 33,  '01/01/2011', '15/05/2012' UNION ALL
SELECT 'KLD', 'K05', 79,  '16/05/2012', '01/01/2013' UNION ALL
SELECT 'KTY', 'K05', 52,  '01/01/2011', '01/02/2012' UNION ALL
SELECT 'KMD', 'C02', 78,  '01/01/2011', '01/02/2012' UNION ALL
SELECT 'ZYT', 'G01', 11,  '01/01/2011', '01/02/2012' UNION ALL
SELECT 'PYN', 'A01', 15,  '01/01/2011', '01/02/2012' UNION ALL
SELECT 'DYN', 'F08', 122, '01/01/2011', '01/02/2012';

“有点复杂。”是的

从顶部开始，这是您的第2步。 这可能有一个关键的缺陷，因为它没有限制任何东西持续3年或12个月。从你的问题中不清楚这是什么意图，所以我们将使用这个

SELECT Site,Code,COUNT(*) as [Count],
FailFlagCount= SUM(CASE WHEN FailFlag = 'F' THEN 1 ELSE 0 END)
FROM Table1 
GROUP BY Site,Code;

需要将其与表2合并以获得频率场。我将其包装在一个子选择中，并将其称为子查询S2，因为这是第2步

SELECT s2.Site, s2.Code,s2.[Count], S2.FailFlagCount, Table2.Freq
FROM
(
SELECT Site,Code,COUNT(*) as [Count],
FailFlagCount= SUM(CASE WHEN FailFlag = 'F' THEN 1 ELSE 0 END)
FROM Table1 
GROUP BY Site,Code
) S2
LEFT JOIN Table2 on S2.Site = table2.Site and S2.Code = table2.Code
;

计算CC和BPO相对容易。上面的代码变成了一个名为S3的子查询

注意，我在这里防止了被零除的错误

SELECT Site, Code, [Count], FailFlagCount, Freq,
CASE WHEN [Count]=0 THEN 0 ELSE  ( 1-(FailFlagCount / [Count]))*100 END  CC,
CASE WHEN Freq = 0 THEN 0 ELSE ( 1-(FailFlagCount / Freq))*100 END BPO

FROM
(
SELECT s2.Site, s2.Code,s2.[Count], S2.FailFlagCount, ISNULL(Table2.Freq, 1) AS Freq
FROM   
(
SELECT Site,Code,COUNT(*) as [Count],
FailFlagCount= SUM(CASE WHEN FailFlag = 'F' THEN 1 ELSE 0 END)
FROM Table1 
GROUP BY Site,Code
) S2
LEFT JOIN Table2 on S2.Site = table2.Site and S2.Code = table2.Code
) s3
;

好的，这里有一点实验，围绕SQL month函数，如何计算还有多少个月，以及今年的第一个月

select GETDATE(), month(GETDATE()), 12-MONTH(GETDATE()),
DATEADD(YEAR, -3, GETDATE()),
DATEFROMPARTS (YEAR(GETDATE()),1,1)
;

SELECT s3.Site, s3.Code, [Count], FailFlagCount, Freq,
CASE WHEN [Count]=0 THEN 0 ELSE  ( 1-(FailFlagCount / [Count]))*100 END  CC,
CASE WHEN Freq = 0 THEN 0 ELSE ( 1-(FailFlagCount / Freq))*100 END BPO,
ISNULL(YBIt.Y, 0) Y,
isnull(currBit.currFails, 0) as currFails

FROM
(
SELECT s2.Site, s2.Code,s2.[Count], S2.FailFlagCount, ISNULL(Table2.Freq, 1) AS Freq
FROM   
(
SELECT Site,Code,COUNT(*) as [Count],
FailFlagCount= SUM(CASE WHEN FailFlag = 'F' THEN 1 ELSE 0 END)
FROM Table1 
GROUP BY Site,Code
) S2
LEFT JOIN Table2 on S2.Site = table2.Site and S2.Code = table2.Code
) s3

LEFT JOIN
(
SELECT SITE, CODE, count(*) tot, cast(count(*) as float)/36 avg, (cast(count(*) as float)/36) * ((12-MONTH(GETDATE()))/12) Y
FROM Table1 
WHERE FailFlag = 'F'
AND ModifiedDate >= DATEADD(YEAR, -3, GETDATE())
GROUP BY SITE, CODE
) YBit on S3.Site = Ybit.site AND S3.code = YBit.code 

LEFT JOIN
(
SELECT SITE, CODE, COUNT(*) currFails 
FROM Table1 
WHERE FailFlag = 'F'
AND ModifiedDate >= DATEFROMPARTS (YEAR(GETDATE()),1,1)
GROUP BY SITE, CODE
) currBit ON s3.site = currBit.site and s3.code = currBit.code;

;

现在我们可以计算Y数字了——这是表1的全部内容，但在过去36个月内进行了过滤，然后计算出月平均值。当你在这个月中旬运行时，隐性错误可能会在这里蔓延——它不会在月份边界上断裂，它是在过去3年完成的，而不是最后36个月完成的，不包括这个月。

请注意，我必须将count（*）强制转换为浮点，否则它是一个整数，在除法中四舍五入为0

SELECT SITE, CODE, count(*) tot, cast(count(*) as float)/36 avg, (cast(count(*) as float)/36) * ((12-MONTH(GETDATE()))/12) Y
FROM Table1 
WHERE FailFlag = 'F'
AND ModifiedDate >= DATEADD(YEAR, -3, GETDATE())
GROUP BY SITE, CODE;

我们用同样的方法计算今年的失败率

SELECT SITE, CODE, COUNT(*) currFails 
FROM Table1 
WHERE FailFlag = 'F'
AND ModifiedDate >= DATEFROMPARTS (YEAR(GETDATE()),1,1)
GROUP BY SITE, CODE;

因此，现在我们可以结合上述三个步骤，得到一个包含Y和当前年份的表。注意，这里的ISNULL和左连接用于补偿今年没有失败的站点

select GETDATE(), month(GETDATE()), 12-MONTH(GETDATE()),
DATEADD(YEAR, -3, GETDATE()),
DATEFROMPARTS (YEAR(GETDATE()),1,1)
;

SELECT s3.Site, s3.Code, [Count], FailFlagCount, Freq,
CASE WHEN [Count]=0 THEN 0 ELSE  ( 1-(FailFlagCount / [Count]))*100 END  CC,
CASE WHEN Freq = 0 THEN 0 ELSE ( 1-(FailFlagCount / Freq))*100 END BPO,
ISNULL(YBIt.Y, 0) Y,
isnull(currBit.currFails, 0) as currFails

FROM
(
SELECT s2.Site, s2.Code,s2.[Count], S2.FailFlagCount, ISNULL(Table2.Freq, 1) AS Freq
FROM   
(
SELECT Site,Code,COUNT(*) as [Count],
FailFlagCount= SUM(CASE WHEN FailFlag = 'F' THEN 1 ELSE 0 END)
FROM Table1 
GROUP BY Site,Code
) S2
LEFT JOIN Table2 on S2.Site = table2.Site and S2.Code = table2.Code
) s3

LEFT JOIN
(
SELECT SITE, CODE, count(*) tot, cast(count(*) as float)/36 avg, (cast(count(*) as float)/36) * ((12-MONTH(GETDATE()))/12) Y
FROM Table1 
WHERE FailFlag = 'F'
AND ModifiedDate >= DATEADD(YEAR, -3, GETDATE())
GROUP BY SITE, CODE
) YBit on S3.Site = Ybit.site AND S3.code = YBit.code 

LEFT JOIN
(
SELECT SITE, CODE, COUNT(*) currFails 
FROM Table1 
WHERE FailFlag = 'F'
AND ModifiedDate >= DATEFROMPARTS (YEAR(GETDATE()),1,1)
GROUP BY SITE, CODE
) currBit ON s3.site = currBit.site and s3.code = currBit.code;

;

因此，我们可以最终计算预测

SELECT Site, code, [Count], FAilFlagCount, Freq,
   CC, BPO, Y, currFails, Y+currFails totF12,
    CASE WHEN Freq = 0 then 0 else (1-(( Y+currFails)/ Freq))*100 END Forecast
 FROM 
(
SELECT s3.Site, s3.Code, [Count], FailFlagCount, Freq,
CASE WHEN [Count]=0 THEN 0 ELSE  ( 1-(FailFlagCount / [Count]))*100 END  CC,
CASE WHEN Freq = 0 THEN 0 ELSE ( 1-(FailFlagCount / Freq))*100 END BPO,
ISNULL(YBIt.Y, 0) Y,
isnull(currBit.currFails, 0) as currFails

FROM
(
SELECT s2.Site, s2.Code,s2.[Count], S2.FailFlagCount, ISNULL(Table2.Freq, 1) AS Freq
FROM   
(
SELECT Site,Code,COUNT(*) as [Count],
FailFlagCount= SUM(CASE WHEN FailFlag = 'F' THEN 1 ELSE 0 END)
FROM Table1 
GROUP BY Site, Code
) S2
LEFT JOIN Table2 on S2.Site = table2.Site and S2.Code = table2.Code
) s3

LEFT JOIN
(
SELECT SITE, CODE, count(*) tot, cast(count(*) as float)/36 avg, (cast(count(*) as float)/36) * ((12-MONTH(GETDATE()))/12) Y
FROM Table1 
WHERE FailFlag = 'F'
AND ModifiedDate >= DATEADD(YEAR, -3, GETDATE())
GROUP BY SITE, CODE
) YBit on S3.Site = Ybit.site AND S3.code = YBit.code 

LEFT JOIN
(
SELECT SITE, CODE, COUNT(*) currFails 
FROM Table1 
WHERE FailFlag = 'F'
AND ModifiedDate >= DATEFROMPARTS (YEAR(GETDATE()),1,1)
GROUP BY SITE, CODE
) currBit ON s3.site = currBit.site and s3.code = currBit.code

) S4A
;

好的，我们突然需要开始日期和结束日期。它们在表2中，这似乎也是站点+代码的主列表。让我们做一个主节点，并用左连接将我们的计算连接到主节点。我们还将计算站点类型

这是您对每个站点+代码的详细计算。

select left (t2.site, 1) siteType, t2.site, t2.code, 
  [Count], FAilFlagCount, s4.Freq,
   CC, BPO, Forecast,
    T2.StartDate, T2.EndDAte

FROM 
 Table2 T2 
LEFT JOIN
(

SELECT Site, code, [Count], FAilFlagCount, Freq,
   CC, BPO, Y, currFails, Y+currFails totF12,
    CASE WHEN Freq = 0 then 0 else (1-(( Y+currFails)/ Freq))*100 END Forecast
 FROM 
(
SELECT s3.Site, s3.Code, [Count], FailFlagCount, Freq,
CASE WHEN [Count]=0 THEN 0 ELSE  ( 1-(FailFlagCount / [Count]))*100 END  CC,
CASE WHEN Freq = 0 THEN 0 ELSE ( 1-(FailFlagCount / Freq))*100 END BPO,
ISNULL(YBIt.Y, 0) Y,
isnull(currBit.currFails, 0) as currFails

FROM
(
SELECT s2.Site, s2.Code,s2.[Count], S2.FailFlagCount, ISNULL(Table2.Freq, 1) AS Freq
FROM   
(
SELECT Site,Code,COUNT(*) as [Count],
FailFlagCount= SUM(CASE WHEN FailFlag = 'F' THEN 1 ELSE 0 END)
FROM Table1 
GROUP BY Site, Code
) S2
LEFT JOIN Table2 on S2.Site = table2.Site and S2.Code = table2.Code
) s3

LEFT JOIN
(
SELECT SITE, CODE, count(*) tot, cast(count(*) as float)/36 avg, (cast(count(*) as float)/36) * ((12-MONTH(GETDATE()))/12) Y
FROM Table1 
WHERE FailFlag = 'F'
AND ModifiedDate >= DATEADD(YEAR, -3, GETDATE())
GROUP BY SITE, CODE
) YBit on S3.Site = Ybit.site AND S3.code = YBit.code 

LEFT JOIN
(
SELECT SITE, CODE, COUNT(*) currFails 
FROM Table1 
WHERE FailFlag = 'F'
AND ModifiedDate >= DATEFROMPARTS (YEAR(GETDATE()),1,1)
GROUP BY SITE, CODE
) currBit ON s3.site = currBit.site and s3.code = currBit.code

) S4A
) S4 
on t2.site = s4.site and t2.code = s4.code
;

从那里，我们可以很容易地创建平均值

SELECT siteType, '' as site, '' as code, avg([count]) [count], avg(FailFlagCount) FailFlagCount, avg(Freq) Freq, 
    avg (CC) CC, avg(BPO) BPO, avg(Forecast) Forecast, '' StartDAte, '' EndDate
FROM 
(

select left (t2.site, 1) siteType, t2.site, t2.code, 
  [Count], FAilFlagCount, s4.Freq,
   CC, BPO, Forecast,
    T2.StartDate, T2.EndDAte

FROM 
 Table2 T2 
LEFT JOIN
(

SELECT Site, code, [Count], FAilFlagCount, Freq,
   CC, BPO, Y, currFails, Y+currFails totF12,
    CASE WHEN Freq = 0 then 0 else (1-(( Y+currFails)/ Freq))*100 END Forecast
 FROM 
(
SELECT s3.Site, s3.Code, [Count], FailFlagCount, Freq,
CASE WHEN [Count]=0 THEN 0 ELSE  ( 1-(FailFlagCount / [Count]))*100 END  CC,
CASE WHEN Freq = 0 THEN 0 ELSE ( 1-(FailFlagCount / Freq))*100 END BPO,
ISNULL(YBIt.Y, 0) Y,
isnull(currBit.currFails, 0) as currFails

FROM
(
SELECT s2.Site, s2.Code,s2.[Count], S2.FailFlagCount, ISNULL(Table2.Freq, 1) AS Freq
FROM   
(
SELECT Site,Code,COUNT(*) as [Count],
FailFlagCount= SUM(CASE WHEN FailFlag = 'F' THEN 1 ELSE 0 END)
FROM Table1 
GROUP BY Site, Code
) S2
LEFT JOIN Table2 on S2.Site = table2.Site and S2.Code = table2.Code
) s3

LEFT JOIN
(
SELECT SITE, CODE, count(*) tot, cast(count(*) as float)/36 avg, (cast(count(*) as float)/36) * ((12-MONTH(GETDATE()))/12) Y
FROM Table1 
WHERE FailFlag = 'F'
AND ModifiedDate >= DATEADD(YEAR, -3, GETDATE())
GROUP BY SITE, CODE
) YBit on S3.Site = Ybit.site AND S3.code = YBit.code 

LEFT JOIN
(
SELECT SITE, CODE, COUNT(*) currFails 
FROM Table1 
WHERE FailFlag = 'F'
AND ModifiedDate >= DATEFROMPARTS (YEAR(GETDATE()),1,1)
GROUP BY SITE, CODE
) currBit ON s3.site = currBit.site and s3.code = currBit.code

) S4A
) S4 
on t2.site = s4.site and t2.code = s4.code
) S5
GROUP BY siteType
;

或者，您可以用详细的计算填充一个表，然后计算平均值，而不是使用我的最终计算

---

我将把最后一个联合留给你。

“3.我们检查表2中的相同组合。”这意味着什么？我的意思是，我们在表2中寻找相同的“站点”和“代码”组合，以获得完整书面问题的频率+1。但表2中没有“ABT/A02”的频率。在这种情况下，有很多——三个。“ABT/A02”的正确频率值是多少？如果您可以观察到这3行的开始日期和结束日期是不同的。因此，我们考虑所有3行，出于同样的原因，我的第5步将有日期谢谢您GregHNZ如此努力帮助我。这正是我需要的东西。我添加的几个tweek是在计算CC和BPO时，我必须将它们转换为float以获得准确的值。再次感谢。还有一个更新是，我已经完成了按站点类型和代码的最后一组，并在我的查询中包含了Det，因为我们需要获得站点类型和代码的固定组合。我希望我是对的。