Selecting a date based upon non unique column data in SQL Server

Question

Any suggestion please........ This is where I completely stucked. I have no code to show what I have tried. Cause I couldn't think of a way to figure this out in sql? Somebody suggested me to use pivot (new to the concept...couldn't figure it out). 4 steps of Col5 makes a complete cycle. I have multiple steps processed and need the timing of a complete cycle. 1 step of a cycle may begin before completion of other.

Table A

Col1    Col2     Col3 Col4 Col4     Col5   Col6
 A       Date1     11     12   13     StepA   AA
 A       Date2     11     12   13     StepB   AA 
 A       Date3     11     12   13     StepC   AA
 A       Date4     11     12   13     StepD   AA
 A       dAte1     11     12   13     StepA   AC
 A       dAte2     11     12   13     StepB   AC 
 A       dAte3     11     12   13     StepC   AC
 A       daTe1     111    12   13     StepA   AB <=== Step is not completed    
 A       dAte4     11     12   13     StepD   AC <=== Step completed but other step started already
 A       daTe2     111    12   13     StepB   AB 
 A       daTe3     111    12   13     StepC   AB
 A       daTe4     111    12   13     StepD   AB

Expected OutPut Table (Result):

AA Date1  Date2  Date3  Date4
AC dAte1  dAte2  dAte3  dAte4
AB daTe1  daTe2  daTe3  daTe4 

Answer 1

The Query:

SELECT * FROM 
    (SELECT [Col1]
      ,[Col2]
      ,[Col3]
      ,[Col4]
      ,[Col41]
      ,[Col5]
      ,[Col6]
    FROM [dbo].[YourTable])p
PIVOT ( MAX(Col2) FOR Col5  IN ([StepA],[StepB],[StepC],[StepD]) ) PVT

The Output:

Col1    Col3    Col4    Col41   Col6    StepA   StepB   StepC   StepD
A       11      12      13      AA      Date1   Date2   Date3   Date4
A       11      12      13      AC      dAte1   dAte2   dAte3   dAte4
A       111     12      13      AB      daTe1   daTe2   daTe3   daTe4

You can choose which columns you require and which columns to remove in your final select.(regarding col3,Col4,Col41)

** * * Based on the logic explained in the comments * ** * ***

/* Provided sample Data */

select * from #Testdata

COL1    Date                        COL2   COL3 COL4 COL5  COL6
QMS14   2013-04-08 17:19:52.8944000 3132    95  70   RFP    GSBE 
QMS14   2013-04-08 17:19:58.1445000 3132    95  70   SR     GSBE 
QMS14   2013-04-08 17:19:58.3164000 3132    95  70   RTP    GSBE 
QMS14   2013-04-08 17:19:58.3164000 3132    95  70   SoRe   GSBE
QMS14   2013-04-08 17:39:30.0501000 3132    95  70   RFP    GSBE 
QMS14   2013-04-08 17:39:30.0501000 3132    95  70   SR     GSBE 
QMS14   2013-04-08 17:39:30.3470000 3132    95  70   RTP    GSBE 
QMS14   2013-04-08 17:39:30.3470000 3132    95  70   SoRe   GSBE

/*Query assuming each Cycle happens in the same minute */

SELECT 
     COL1 
    ,COL2 
    ,COL3 
    ,COL4 
    ,COL6 
    ,[RFP]
    ,[SR]
    ,[RTP]
    ,[SoRe]
FROM (
    SELECT *
        ,DATEPART(MINUTE,DATE) DateMinute
        ,DENSE_RANK() OVER ( PARTITION BY Col1 ORDER BY DATEPART(MINUTE,DATE)) [Rank]   FROM  #Testdata
        )a
    PIVOT ( MAX([DATE]) FOR COL5 IN ([RFP],[SR], [RTP],[SoRe] )) PVT

OUTPUT:

COL1    COL2    COL3 COL4   COL6    RFP                         SR                          RTP                         SoRe
QMS14   3132     95   70   GSBE     2013-04-08 17:19:52.8944000 2013-04-08 17:19:58.1445000 2013-04-08 17:19:58.3164000 2013-04-08 17:19:58.3164000
QMS14   3132     95   70   GSBE     2013-04-08 17:39:30.0501000 2013-04-08 17:39:30.0501000 2013-04-08 17:39:30.3470000 2013-04-08 17:39:30.3470000

Answer 2

To me, it looks like the cycles are determined by combinations of four "steps" with the same tag in col6 . The data in the example only has one cycle per code, but the comments suggest there might be more than one.

How can we identify which cycle for a given code a record belongs to? Well, one way is to assume that all the cycles with that code are complete and just enumerate the records by col6, col5 . This additional cycle sequence number is then used for aggregation. For enumerating the sequences, I further assume that the dates are increasing over time.

Although a pivot can be used in this case, I think the aggregation method might be a bit clearer:

select col6, seqnum,
       MAX(case when col5 = 'StepA' then col2 end),
       MAX(case when col5 = 'StepB' then col2 end),
       MAX(case when col5 = 'StepC' then col2 end),
       MAX(case when col5 = 'StepD' then col2 end)
from (select a.*,
             ROW_NUMBER() over (partition by col6, col5 order by col2) as seqnum
      from a
     ) a
group by col6, seqnum