No alternative but row-by-row processing?

• mmcardle

  SSCommitted

  Points: 1543

  More actions

  May 21, 2018 at 9:20 am

  #356478

  I know how to do this with a loop or cursor, but wondered if there was a brilliant mind out there that could suggest a set-based operation?

  This is some example data:

  

  I'm working with the first three columns in a CTE and I'm trying to get the result in the green column to return as part of the set.

  Basically, where the code and type are the same on consecutive days, the first day is the first date of those days.

  So, code 0242628 has SICK20 running consecutively from 3rd May to 5th May and so the first day is 3rd May. There is then a break (6th, 7th and 8th May), before another set runs from 9th May to 12th May, so first date is the 9th.

  I cannot figure out a set-based way to do this.

• Phil Parkin

  SSC Guru

  Points: 246960

  More actions

  May 21, 2018 at 9:40 am

  #1991268

  mmcardle - Monday, May 21, 2018 9:20 AM

  I know how to do this with a loop or cursor, but wondered if there was a brilliant mind out there that could suggest a set-based operation?

  This is some example data:

  

  I'm working with the first three columns in a CTE and I'm trying to get the result in the green column to return as part of the set.

  Basically, where the code and type are the same on consecutive days, the first day is the first date of those days.

  So, code 0242628 has SICK20 running consecutively from 3rd May to 5th May and so the first day is 3rd May. There is then a break (6th, 7th and 8th May), before another set runs from 9th May to 12th May, so first date is the 9th.

  I cannot figure out a set-based way to do this.

  You've been here long enough to know that you should present your sample data in a consumable format.

  The absence of evidence is not evidence of absence
  - Martin Rees
  The absence of consumable DDL, sample data and desired results is, however, evidence of the absence of my response
  - Phil Parkin

• aaron.reese

  SSChampion

  Points: 13498

  More actions

  May 21, 2018 at 9:53 am

  #1991277

  looks like a gaps and islands problem to me.
  You might find it easier with a Date Tally Table too.

  You could RANK() the data and then take the RANK value away from the date.  All records with the same calculated date are consecutive days.  (NOTE this won't work if you are ignoring weekends/bank holidays) 

  you could LEFT join the data to itself using a DATEADD() offset to link today and yesterday, then RANK the joined dates to find the 1st rank dates.

• mmcardle

  SSCommitted

  Points: 1543

  More actions

  May 21, 2018 at 10:02 am

  #1991283

  Phil Parkin - Monday, May 21, 2018 9:40 AM

  You've been here long enough to know that you should present your sample data in a consumable format.

  Sorry Phil, I was rushing and the first attempt at copying and pasting from Excel went awry.

  OK, the data can be had from this:
  
DECLARE @tabData TABLE
    (
        [Code] varchar(7),
        [Date] date,
        [Type] varchar(20)
    );

  INSERT INTO @tabData([Code], [Date], [Type])
SELECT '0242589', '2018-04-20', 'HOLHRSAPR'
UNION
SELECT '0242597', '2018-04-19', 'HOLHRSAPR'
UNION
SELECT '0242628', '2018-04-02', 'HOLHRSAPR'
UNION
SELECT '0242628', '2018-04-03', 'SICK20'
UNION
SELECT '0242628', '2018-04-04', 'SICK20'
UNION
SELECT '0242628', '2018-04-05', 'SICK20'
UNION
SELECT '0242628', '2018-04-09', 'SICK20'
UNION
SELECT '0242628', '2018-04-10', 'SICK20'
UNION
SELECT '0242628', '2018-04-11', 'SICK20'
UNION
SELECT '0242628', '2018-04-12', 'SICK20'
UNION
SELECT '0242640', '2018-04-03', 'HOLHRSAPR'
UNION
SELECT '0242640', '2018-04-09', 'HOLHRSAPR';

  SELECT    [Code], [Date], [Type], '?' AS [First Day]
FROM    @tabData;

  

• Lynn Pettis

  SSC Guru

  Points: 442467

  More actions

  May 21, 2018 at 11:11 am

  #1991304

  Hope this helps:
  
DECLARE @tabData TABLE
    (
        [Code] varchar(7),
        [Date] date,
        [Type] varchar(20)
    );

  INSERT INTO @tabData([Code], [Date], [Type])
  SELECT '0242589', '2018-04-20', 'HOLHRSAPR'
  UNION
  SELECT '0242597', '2018-04-19', 'HOLHRSAPR'
  UNION
  SELECT '0242628', '2018-04-02', 'HOLHRSAPR'
  UNION
  SELECT '0242628', '2018-04-03', 'SICK20'
  UNION
  SELECT '0242628', '2018-04-04', 'SICK20'
  UNION
  SELECT '0242628', '2018-04-05', 'SICK20'
  UNION
  SELECT '0242628', '2018-04-09', 'SICK20'
  UNION
  SELECT '0242628', '2018-04-10', 'SICK20'
  UNION
  SELECT '0242628', '2018-04-11', 'SICK20'
  UNION
  SELECT '0242628', '2018-04-12', 'SICK20'
  UNION
  SELECT '0242640', '2018-04-03', 'HOLHRSAPR'
  UNION
  SELECT '0242640', '2018-04-09', 'HOLHRSAPR';

  WITH base AS (
  SELECT
   

  .[Code]
    ,

  .[Date]
    ,

  .[Type]
    , '?' AS [First Day]
    , rn = ROW_NUMBER() OVER (PARTITION BY

  .[Code],

  .[Type] ORDER BY

  .[Date])
  --  , [td2].*
  FROM
    @tabData td
  --  LEFT OUTER JOIN @tabData AS [td2]
  --    ON [td2].[Code] =

  .[Code] AND [td2].[Type] =

  .[Type] AND DATEADD(DAY, 1,[td2].[Date]) =

  .[Date]
  )
  SELECT
    .[Code]
    , .[Date]
    , .[Type]
    , [First Day] = MIN(.[Date]) OVER (PARTITION BY .[Code], .[Type], DATEADD(DAY,-.[rn],.[Date]))
    , .[rn]
    , [grpDate] = DATEADD(DAY,-.[rn],.[Date])
  FROM
    [base] ;
  GO
  

• itmasterw 60042

  SSCrazy

  Points: 2858

  More actions

  May 21, 2018 at 11:20 am

  #1991307

  If you look up  dim date table, it can have this field in it or you can create it. This table can be created and used in a number of projects. it can be created in a few minutes and will hold all the dates you will ever need. And will emirate the need for  most formulas, You will just query the table for where the date  matches the date in this table.

• Lynn Pettis

  SSC Guru

  Points: 442467

  More actions

  May 21, 2018 at 11:42 am

  #1991310

  itmasterw 60042 - Monday, May 21, 2018 11:20 AM

  If you look up  dim date table, it can have this field in it or you can create it. This table can be created and used in a number of projects. it can be created in a few minutes and will hold all the dates you will ever need. And will emirate the need for  most formulas, You will just query the table for where the date  matches the date in this table.

  Care to demonstrate how this would work in this situation?

• Phil Parkin

  SSC Guru

  Points: 246960

  More actions

  May 21, 2018 at 11:45 am

  #1991313

  Lynn Pettis - Monday, May 21, 2018 11:42 AM

  itmasterw 60042 - Monday, May 21, 2018 11:20 AM

  If you look up  dim date table, it can have this field in it or you can create it. This table can be created and used in a number of projects. it can be created in a few minutes and will hold all the dates you will ever need. And will emirate the need for  most formulas, You will just query the table for where the date  matches the date in this table.

  Care to demonstrate how this would work in this situation?

  +1, exactly what I was thinking.

  The absence of evidence is not evidence of absence
  - Martin Rees
  The absence of consumable DDL, sample data and desired results is, however, evidence of the absence of my response
  - Phil Parkin

• sestell1

  SSChampion

  Points: 10230

  More actions

  May 21, 2018 at 12:35 pm

  #1991327

  Here's an example using just table joins:

  
;WITH TabData (Code, Date, Type)
AS
(
    SELECT
        Code, Date, type
    FROM
        @tabData
)
SELECT
    Full_Data.Code,
    Full_Data.Date,
    Full_Data.type,
    Max(Series_Start.Date) AS Key_Date
FROM
    TabData Full_Data
Join
    TabData Series_Start
LEFT JOIN
    TabData Previous_Date
        ON
            Series_Start.Code = Previous_Date.Code
        AND Series_Start.Type = Previous_Date.Type
        AND Series_Start.Date = DateAdd(Day, 1, Previous_Date.Date)
        ON
            Full_Data.Code = Series_Start.Code
        AND Full_Data.Type = Series_Start.Type
        AND Full_Data.Date >= Series_Start.Date
        AND Previous_Date.Date IS Null
GROUP BY
    Full_Data.Code,
    Full_Data.Date,
    Full_Data.type
ORDER BY
    Full_Data.Code,
    Full_Data.Date

• Lynn Pettis

  SSC Guru

  Points: 442467

  More actions

  May 21, 2018 at 12:37 pm

  #1991331

  sestell1 - Monday, May 21, 2018 12:35 PM

  Here's an example using just table joins:

  
;WITH TabData (Code, Date, Type)
AS
(
    SELECT
        Code, Date, type
    FROM
        @tabData
)
SELECT
    Full_Data.Code,
    Full_Data.Date,
    Full_Data.type,
    Max(Series_Start.Date) AS Key_Date
FROM
    TabData Full_Data
Join
    TabData Series_Start
LEFT JOIN
    TabData Previous_Date
        ON
            Series_Start.Code = Previous_Date.Code
        AND Series_Start.Type = Previous_Date.Type
        AND Series_Start.Date = Previous_Date.Date + 1
        ON
            Full_Data.Code = Series_Start.Code
        AND Full_Data.Type = Series_Start.Type
        AND Full_Data.Date >= Series_Start.Date
        AND Previous_Date.Date IS Null
GROUP BY
    Full_Data.Code,
    Full_Data.Date,
    Full_Data.type
ORDER BY
    Full_Data.Code,
    Full_Data.Date

  Your code does not run.
  

  Msg 206, Level 16, State 2, Line 56

  Operand type clash: date is incompatible with int

• itmasterw 60042

  SSCrazy

  Points: 2858

  More actions

  May 21, 2018 at 1:37 pm

  #1991343

  Well with out putting this together it is hard to fully demonstrate. But unless I am misunderstanding the request, you should be able to do an inner join between his table and the dime data table and return the FirstDayOfMonth. So if his table DateTb, for example then 
  he would just do something like this:

  SELECT FirstDayOfMonth
                                    From DimDate  dt INNER JOIN DateTB db
                                     ON   dt.date = db.Date

• Lynn Pettis

  SSC Guru

  Points: 442467

  More actions

  May 21, 2018 at 1:40 pm

  #1991344

  itmasterw 60042 - Monday, May 21, 2018 1:37 PM

  Well with out putting this together it is hard to fully demonstrate. But unless I am misunderstanding the request, you should be able to do an inner join between his table and the dime data table and return the FirstDayOfMonth. So if his table DateTb, for example then 
  he would just do something like this:

  SELECT FirstDayOfMonth
                                    From DimDate  dt INNER JOIN DateTB db
                                     ON   dt.date = db.Date

  But that isn't what the OP is looking for in this case.  Go back at look at the original post.

• sestell1

  SSChampion

  Points: 10230

  More actions

  May 21, 2018 at 2:57 pm

  #1991350

  Lynn Pettis - Monday, May 21, 2018 12:37 PM

  sestell1 - Monday, May 21, 2018 12:35 PM

  Here's an example using just table joins:

  
;WITH TabData (Code, Date, Type)
AS
(
    SELECT
        Code, Date, type
    FROM
        @tabData
)
SELECT
    Full_Data.Code,
    Full_Data.Date,
    Full_Data.type,
    Max(Series_Start.Date) AS Key_Date
FROM
    TabData Full_Data
Join
    TabData Series_Start
LEFT JOIN
    TabData Previous_Date
        ON
            Series_Start.Code = Previous_Date.Code
        AND Series_Start.Type = Previous_Date.Type
        AND Series_Start.Date = Previous_Date.Date + 1
        ON
            Full_Data.Code = Series_Start.Code
        AND Full_Data.Type = Series_Start.Type
        AND Full_Data.Date >= Series_Start.Date
        AND Previous_Date.Date IS Null
GROUP BY
    Full_Data.Code,
    Full_Data.Date,
    Full_Data.type
ORDER BY
    Full_Data.Code,
    Full_Data.Date

  Your code does not run.
  

  Msg 206, Level 16, State 2, Line 56

  Operand type clash: date is incompatible with int

  Oh sorry, I was lazy and changed the date column data type to DateTime so I could subtract 1 rather than use DateAdd.  :Whistling:
  I've updated the code above to use DateAdd.  

• DesNorton

  SSC-Insane

  Points: 24386

  More actions

  May 22, 2018 at 3:41 am

  #1991379

  The following 2 links are solution to the Packing Intervals Problem by Itzik Ben-Gan.
  2011/04/13 ... http://blogs.solidq.com/en/sqlserver/packing-intervals/
  2015/08/11 ... http://sqlmag.com/sql-server/new-solution-packing-intervals-problem

  ---

  How to post data/code on a forum to get the best help.

• ChrisM@Work

  SSC Guru

  Points: 186127

  More actions

  May 22, 2018 at 4:12 am

  #1991382

  IF OBJECT_ID('tempdb..#tabData') IS NOT NULL DROP TABLE #tabData;

  CREATE TABLE #tabData ([Code] varchar(7),[Date] date,[Type] varchar(20));

  INSERT INTO #tabData ([Code], [Date], [Type])

  VALUES

  ('0242589', '2018-04-20', 'HOLHRSAPR'),

  ('0242597', '2018-04-19', 'HOLHRSAPR'),

  ('0242628', '2018-04-02', 'HOLHRSAPR'),

  ('0242628', '2018-04-03', 'SICK20'),

  ('0242628', '2018-04-04', 'SICK20'),

  ('0242628', '2018-04-05', 'SICK20'),

  ('0242628', '2018-04-09', 'SICK20'),

  ('0242628', '2018-04-10', 'SICK20'),

  ('0242628', '2018-04-11', 'SICK20'),

  ('0242628', '2018-04-12', 'SICK20'),

  ('0242640', '2018-04-03', 'HOLHRSAPR'),

  ('0242640', '2018-04-09', 'HOLHRSAPR');

  CREATE UNIQUE CLUSTERED INDEX ucx_Stuff ON #tabData ([Code], [Type], [Date]);

  SELECT [Code], [Date], [Type],

  [First Day] = MIN([Date]) OVER(PARTITION BY [Code], [Type], grp)

  FROM (

  SELECT [Code], [Date], [Type],

  grp = DATEADD(DAY,1-ROW_NUMBER() OVER(PARTITION BY [Code], [Type] ORDER BY [Date]),[Date])

  FROM #tabData

  ) d;

  EDIT: Exactly the same method as Lynn 🙂

  ^{“Write the query the simplest way. If through testing it becomes clear that the performance is inadequate, consider alternative query forms.” - Gail Shaw}

  For fast, accurate and documented assistance in answering your questions, please read this article.
  Understanding and using APPLY, (I) and (II) Paul White
  Hidden RBAR: Triangular Joins / The "Numbers" or "Tally" Table: What it is and how it replaces a loop Jeff Moden