Delete data with sliding date window

• quinn.jay

  SSCertifiable

  Points: 5505

  More actions

  October 1, 2018 at 12:57 pm

  #379677

  Hello, I have a table where I have one year+ of data, and I need to delete the current month, then reload it, and when the next month comes, the delete statement will roll to the next month to delete. I was working first with a controlled delete against a table like this, but need help with the syntax of the qualify rownumber type of coding that I think is whats needed:
  
  while 1=1 begin
  delete from [database].[dbo].[TableName]
  where [actvt_ym] between 201809 and 201810
  if @@rowcount = 0 break;
  end

  Thank you

• sgmunson

  SSC Guru

  Points: 110640

  More actions

  October 1, 2018 at 1:18 pm

  #2007585

  The delete you've coded does ALL the needed deletes in one fell swoop, and having a loop isn't going to help with that DELETE.   You do need a row number equivalent to help you.  Something like this:
  WHILE 1 = 1
    BEGIN;
    WITH RowNums AS (

          SELECT TOP 1 PERCENT
            TN.PrimaryKeyColumnName
        FROM [database].[dbo].[TableName] AS TN
        WHERE    TN.actvt_ym >= 201809
            AND TN.actvt_ym < 201810
        ORDER BY TN.PrimaryKeyColumnName
    )
    DELETE TN
    FROM RowNums AS RN
    INNER JOIN [database].[dbo].[TableName] AS TN
        ON RN.PrimaryKeyColumnName = TN.PrimaryKeyColumnName;
    
    IF @@ROWCOUNT = 0
        BEGIN
        BREAK;
        END;
    END;

  You can adjust the percentage up or down based on your knowledge of the total row count for that month.

  Steve (aka sgmunson) 🙂 🙂 🙂
  Rent Servers for Income (picks and shovels strategy)

• drew.allen

  SSC Guru

  Points: 76985

  More actions

  October 1, 2018 at 1:51 pm

  #2007591

  sgmunson - Monday, October 1, 2018 1:18 PM

  The delete you've coded does ALL the needed deletes in one fell swoop, and having a loop isn't going to help with that DELETE.   You do need a row number equivalent to help you.  Something like this:
  WHILE 1 = 1
      BEGIN;
      WITH RowNums AS (

          SELECT TOP 1 PERCENT
            TN.PrimaryKeyColumnName
        FROM [database].[dbo].[TableName] AS TN
        WHERE    TN.actvt_ym >= 201809
            AND TN.actvt_ym < 201810
        ORDER BY TN.PrimaryKeyColumnName
    )
    DELETE TN
    FROM RowNums AS RN
    INNER JOIN [database].[dbo].[TableName] AS TN
        ON RN.PrimaryKeyColumnName = TN.PrimaryKeyColumnName;
    
    IF @@ROWCOUNT = 0
        BEGIN
        BREAK;
        END;
    END;

  You can adjust the percentage up or down based on your knowledge of the total row count for that month.

  I would not use PERCENT in a loop where you are deleting records,because the number of records you process each loop will decrease exponentially.  You want to process approximately the same number of records each loop, so you should just use a set value.

  I also see no reason to use a ROW_NUMBER (which requires an expensive sort).  The main reason for using a ROW_NUMBER() here would be to help track which records have already been processed, but you don't need to track them, because they've already been deleted and are no longer available to be processed on subsequent loops.

  There also no reason to do the self join.  It just adds unnecessary overhead and doesn't gain you anything.

  DECLARE
    @cur_count INTEGER,
    @batch_size INTEGER = 50000,
    @date_begin INTEGER = CONVERT(CHAR(6), DATEADD(MONTH, -1, GETDATE()), 112),
    @date_end INTEGER = CONVERT(CHAR(6), GETDATE(), 112);

  SET @cur_count = @batch_size

  WHILE @cur_count = @batch_size
  BEGIN
      DELETE TOP (@batch_size)
      FROM [database].[dbo].[TableName]
      WHERE [actvt_ym] BETWEEN @date_begin AND @date_end;

      SET @cur_count = @@ROWCOUNT;
end

  Drew

  J. Drew Allen
  Business Intelligence Analyst
  Philadelphia, PA

• quinn.jay

  SSCertifiable

  Points: 5505

  More actions

  October 1, 2018 at 1:54 pm

  #2007592

  But lets say we were in Oct18 and now Nov18, how would the code know that the month changed, and would know to change from Oct18? I would think that needs to be dynamic, and this doesn't look like that, or I'm not getting it

• quinn.jay

  SSCertifiable

  Points: 5505

  More actions

  October 1, 2018 at 1:55 pm

  #2007593

  quinn.jay - Monday, October 1, 2018 1:54 PM

  But lets say we were in Oct18 and now Nov18, how would the code know that the month changed, and would know to change from Oct18? I would think that needs to be dynamic, and this doesn't look like that, or I'm not getting it

  Something very simple like this perhaps?:
  
  delete from [database].[dbo].[TableName]where [actvt_ym] = FORMAT(GETDATE(),'yyyyMM')*1

• Jeffrey Williams

  SSC Guru

  Points: 90303

  More actions

  October 1, 2018 at 2:08 pm

  #2007597

  You don't really need a row number - nor do you need a break...

  Declare @rowsAffected int = 0
  , @batchSize int = 200000   -- adjust for your system
  , @activityYearMonth int = cast(convert(char(6), getdate(), 112) As int);

  While @rowsAffected > 0
Begin

  Delete Top(@batchSize)
 From [database].[dbo].[TableName]
Where [actvt_ym] = @activityYearMonth;

    Set @rowsAffected = @@rowcount;
  
   Checkpoint;  -- Or - issue a tlog backup here if database is full recovery
  End;

  In your original code - you used between and 2 periods.  Because BETWEEN is inclusive that delete would remove rows for both periods and I assume you only want it to remove data for the current period.
  

  Jeffrey Williams
  “We are all faced with a series of great opportunities brilliantly disguised as impossible situations.”

  ― Charles R. Swindoll

  How to post questions to get better answers faster
  Managing Transaction Logs

• GaryV

  Hall of Fame

  Points: 3307

  More actions

  October 2, 2018 at 5:09 am

  #2007649

  Jeffrey Williams 3188 - Monday, October 1, 2018 2:08 PM

  You don't really need a row number - nor do you need a break...

  Declare @rowsAffected int = 0
  , @batchSize int = 200000   -- adjust for your system
  , @activityYearMonth int = cast(convert(char(6), getdate(), 112) As int);

  While @rowsAffected > 0
Begin

  Delete Top(@batchSize)
 From [database].[dbo].[TableName]
Where [actvt_ym] = @activityYearMonth;

    Set @rowsAffected = @@rowcount;
  
   Checkpoint;  -- Or - issue a tlog backup here if database is full recovery
  End;

  In your original code - you used between and 2 periods.  Because BETWEEN is inclusive that delete would remove rows for both periods and I assume you only want it to remove data for the current period.
  

  Check the @rowsAffected initial setting - the loop never executes.

• Phil Parkin

  SSC Guru

  Points: 246960

  More actions

  October 2, 2018 at 5:28 am

  #2007653

  quinn.jay - Monday, October 1, 2018 1:55 PM

  quinn.jay - Monday, October 1, 2018 1:54 PM

  But lets say we were in Oct18 and now Nov18, how would the code know that the month changed, and would know to change from Oct18? I would think that needs to be dynamic, and this doesn't look like that, or I'm not getting it

  Something very simple like this perhaps?:
  
  delete from [database].[dbo].[TableName]where [actvt_ym] = FORMAT(GETDATE(),'yyyyMM')*1

  If all that you want is 'current' year/month in YYYYMM format (as an INT), you can use this:

  YEAR(GETDATE()) * 100 + MONTH(GETDATE())

  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

• Jeffrey Williams

  SSC Guru

  Points: 90303

  More actions

  October 2, 2018 at 7:19 am

  #2007675

  gvoshol 73146 - Tuesday, October 2, 2018 5:09 AM

  Jeffrey Williams 3188 - Monday, October 1, 2018 2:08 PM

  You don't really need a row number - nor do you need a break...

  Declare @rowsAffected int = 0
  , @batchSize int = 200000   -- adjust for your system
  , @activityYearMonth int = cast(convert(char(6), getdate(), 112) As int);

  While @rowsAffected > 0
Begin

  Delete Top(@batchSize)
 From [database].[dbo].[TableName]
Where [actvt_ym] = @activityYearMonth;

    Set @rowsAffected = @@rowcount;
  
   Checkpoint;  -- Or - issue a tlog backup here if database is full recovery
  End;

  In your original code - you used between and 2 periods.  Because BETWEEN is inclusive that delete would remove rows for both periods and I assume you only want it to remove data for the current period.
  

  Check the @rowsAffected initial setting - the loop never executes.

  Typo, it should be initialized to 1...

  Jeffrey Williams
  “We are all faced with a series of great opportunities brilliantly disguised as impossible situations.”

  ― Charles R. Swindoll

  How to post questions to get better answers faster
  Managing Transaction Logs

• sgmunson

  SSC Guru

  Points: 110640

  More actions

  October 2, 2018 at 8:17 am

  #2007700

  drew.allen - Monday, October 1, 2018 1:51 PM

  sgmunson - Monday, October 1, 2018 1:18 PM

  The delete you've coded does ALL the needed deletes in one fell swoop, and having a loop isn't going to help with that DELETE.   You do need a row number equivalent to help you.  Something like this:
  WHILE 1 = 1
      BEGIN;
      WITH RowNums AS (

          SELECT TOP 1 PERCENT
            TN.PrimaryKeyColumnName
        FROM [database].[dbo].[TableName] AS TN
        WHERE    TN.actvt_ym >= 201809
            AND TN.actvt_ym < 201810
        ORDER BY TN.PrimaryKeyColumnName
    )
    DELETE TN
    FROM RowNums AS RN
    INNER JOIN [database].[dbo].[TableName] AS TN
        ON RN.PrimaryKeyColumnName = TN.PrimaryKeyColumnName;
    
    IF @@ROWCOUNT = 0
        BEGIN
        BREAK;
        END;
    END;

  You can adjust the percentage up or down based on your knowledge of the total row count for that month.

  I would not use PERCENT in a loop where you are deleting records,because the number of records you process each loop will decrease exponentially.  You want to process approximately the same number of records each loop, so you should just use a set value.

  I also see no reason to use a ROW_NUMBER (which requires an expensive sort).  The main reason for using a ROW_NUMBER() here would be to help track which records have already been processed, but you don't need to track them, because they've already been deleted and are no longer available to be processed on subsequent loops.

  There also no reason to do the self join.  It just adds unnecessary overhead and doesn't gain you anything.

  DECLARE
    @cur_count INTEGER,
    @batch_size INTEGER = 50000,
    @date_begin INTEGER = CONVERT(CHAR(6), DATEADD(MONTH, -1, GETDATE()), 112),
    @date_end INTEGER = CONVERT(CHAR(6), GETDATE(), 112);

  SET @cur_count = @batch_size

  WHILE @cur_count = @batch_size
  BEGIN
      DELETE TOP (@batch_size)
      FROM [database].[dbo].[TableName]
      WHERE [actvt_ym] BETWEEN @date_begin AND @date_end;

      SET @cur_count = @@ROWCOUNT;
end

  Drew

  Oops!   Kinda forgot about the impact of what using a percentage would do.  Here's my corrected code, based on yours, that addresses the date issue:
  DECLARE
    @RowCount    AS int    = 1,
  @BatchSize    AS int    = 50000,
    @Today        AS date    = '20180901';
DECLARE
  @StartMonth    AS int    = CONVERT(char(6), DATEFROMPARTS(YEAR(@Today), MONTH(@Today), 1), 112),
  @EndMonth    AS int    = CONVERT(char(6), DATEADD(month, 1, @Today), 112);

  WHILE @RowCount > 0
    BEGIN
  DELETE TOP (@BatchSize)
  FROM [database].[schema].[TableName] AS TN
  WHERE    TN.actvt_ym >= @StartMonth
        AND TN.actvt_ym < @EndMonth;

    SET @RowCount = @@ROWCOUNT;
    END;

  Steve (aka sgmunson) 🙂 🙂 🙂
  Rent Servers for Income (picks and shovels strategy)