asked on

UPDATING 300 Million Rows With A Cursor Is There A Better Method? 500 POINTS!

I have to update a table with 300 million rows, and there is no unique ID on the table. I have written the cursor below to do the update:

CREATE PROCEDURE dbo.usp_CleanADFTradeline

AS

DECLARE @dynCR_ID INT
DECLARE Curs_tblTasks_Update CURSOR FOR
SELECT DISTINCT CR_ID AS dynCR_ID FROM dbo.tblTasks

OPEN Curs_tblTasks_Update
FETCH NEXT FROM Curs_tblTasks_Update INTO @dynCR_ID

WHILE (@@FETCH_STATUS = 0)
BEGIN
UPDATE tblTasks
      SET Task_Num = CASE
      WHEN DATALENGTH(Task_Num)= 9 THEN SUBSTRING(Task_Num,1,6)
      WHEN DATALENGTH(Task_Num)> 9 THEN SUBSTRING(Task_Num,1,8)
      ELSE Task_Num
      END
      WHERE CR_ID = @dynCR_ID

FETCH NEXT FROM Curs_tblTasks_Update INTO @dynCR_ID
END
CLOSE Curs_tblTasks_Update
DEALLOCATE Curs_tblTasks_Update

GO

Is there a better code that will do this without the performance headache of cursors?

DeMyu

MartinCMS

I don't understand why you're using a cursor for what you're trying to do!

you should be able to just run you Update query straightup.

UPDATE tblTasks
SET Task_Num = CASE
WHEN DATALENGTH(Task_Num)= 9 THEN SUBSTRING(Task_Num,1,6)
WHEN DATALENGTH(Task_Num)> 9 THEN SUBSTRING(Task_Num,1,8)
ELSE Task_Num
END

DeMyu

ASKER

I am trying to run the update in batches and each CR_ID is a batch. Out of the 300 million records there are 13 million distinct CR_ID.

DeMyu

MartinCMS

The Cursor you're using run the same as an Update Statement without the Where Clause. Unless you want to run one distinct CR_ID at the time.... then

UPDATE tblTasks
SET Task_Num = CASE
WHEN DATALENGTH(Task_Num)= 9 THEN SUBSTRING(Task_Num,1,6)
WHEN DATALENGTH(Task_Num)> 9 THEN SUBSTRING(Task_Num,1,8)
ELSE Task_Num
END
Where CR_ID = InputyourCR_ID

I guest I am not 100% understanding what you're trying to accomplish here. Further clarity would be helpful.

Scott Pletcher

Since cursors are so much overhead, personally I would use SET ROWCOUNT to do the batching. Also, if there are 13M distinct CR_IDs, then your batches will be way too small (avg. 24 rows each):

SET ROWCOUNT 100000 --or whatever other batch size you want
SELECT '1'
-- WARNING: as coded, this will run until all rows are updated!!
WHILE @@ROWCOUNT > 0
BEGIN
UPDATE tblTasks
SET Task_Num = SUBSTRING(Task_Num, 1,
CASE DATALENGTH(Task_Num) WHEN 9 THEN 6 ELSE 8 END)
WHERE DATALENGTH(Task_Num) >= 9
END --WHILE

MartinCMS

Or you can do something like this....

UPDATE tblTasks
SET Task_Num = CASE
WHEN DATALENGTH(Task_Num)= 9 THEN SUBSTRING(Task_Num,1,6)
WHEN DATALENGTH(Task_Num)> 9 THEN SUBSTRING(Task_Num,1,8)
ELSE Task_Num END
WHERE CR_ID in (SELECT DISTINCT CR_ID AS dynCR_ID FROM dbo.tblTasks)

Jay Toops

Setup a control table of
CRID for doing a
WHERE CRID > 0 and CRID < 20000000

make your main loop select these values
and have your update statement update more rows at one time.
This will minimize your update statments to whatever level of
granularity you wish

Id do a MAX(CRID) and Min(crid) and divide it by the number of batches you want.

(IF YOU want to get fancy) You can even have a boolean value in that table
so that you can select the particular range for that processing execution.

JAY

Scott Pletcher

NOTE:

That code will tend to run very long near the end since it will have to scan most of the table to find qualifying rows. Although not generally preferred, in this situation you might want to consider an updateable cursor that does the change, for example:

DECLARE Curs_tblTasks_Update CURSOR FOR
SELECT task_num
FROM dbo.tblTasks
WHERE DATALENGTH(task_num) >= 9
FOR UPDATE OF task_num

OPEN Curs_tblTasks_Update
FETCH NEXT FROM Curs_tblTasks_Update INTO @task_num

WHILE (@@FETCH_STATUS = 0)
BEGIN
UPDATE tblTasks
SET Task_Num = SUBSTRING(Task_Num, 1,
CASE DATALENGTH(Task_Num) WHEN 9 THEN 6 ELSE 8 END)
WHERE CURRENT OF Curs_tblTasks_Update
FETCH NEXT FROM Curs_tblTasks_Update INTO @task_num
END
CLOSE Curs_tblTasks_Update
DEALLOCATE Curs_tblTasks_Update
GO

DeMyu

ASKER

ScottPletcher will it update the table in batches of 100000 as you stated above. I guess at the end of the last record @@ROWCOUNT equals 0 and the loop will be exited. Is this right?

Scott Pletcher

Sorry, but this code:

UPDATE tblTasks
SET Task_Num = CASE
WHEN DATALENGTH(Task_Num)= 9 THEN SUBSTRING(Task_Num,1,6)
WHEN DATALENGTH(Task_Num)> 9 THEN SUBSTRING(Task_Num,1,8)
ELSE Task_Num END
WHERE CR_ID in (SELECT DISTINCT CR_ID AS dynCR_ID FROM dbo.tblTasks)

is a waste of time. The cr_id will *always* be present in the list of distinct cr_id's in the same table, so the whole table will always be updated. You might as well just code:

UPDATE tblTasks
SET Task_Num = CASE
WHEN DATALENGTH(Task_Num)= 9 THEN SUBSTRING(Task_Num,1,6)
WHEN DATALENGTH(Task_Num)> 9 THEN SUBSTRING(Task_Num,1,8)
ELSE Task_Num END

ASKER CERTIFIED SOLUTION

Scott Pletcher

membership

This solution is only available to members.

To access this solution, you must be a member of Experts Exchange.

Start Free Trial

DeMyu

ASKER

ScottPletcher, the @rowcount variable in the code above is never incremented. My goal would be to update 100000 rows at a time. Any suggestions?

Scott Pletcher

The @rowcount is how many rows will be *updated*, not just read, so the first UPDATE should do the first 100K rows, the next UPDATE a different 100000 rows, and so on.

DeMyu

ASKER

Thank you.