Need a suggestion for my SQL requirement

Question

Need a suggestion for my SQL requirement

Rocky's

Ten Centuries

Points: 1102
More actions
November 30, 2018 at 12:06 am

#369207

Hello everyone,
Need some best solution for my below requirement.
I need to Insert/update/delete some data into my target based as per my staging (meta) data.

For example -

Target Table: dbo.Employee (With some existing data)

EmpID
FirstName
MiddleName
LastName
Designation
UpdatedDate
1
FN1
MN1
LN1
D1
2018-01-01
2
FN2
MN2
LN2
D2
2018-01-01
3
FN3
MN3
LN3
D3
2018-01-01
4
FN4
MN4
LN4
D4
2018-01-01
5
FN5
MN5
LN5
D5
2018-01-01
6
FN6
MN6
LN6
D6
2018-01-01

Staging Table: staging.Employee

EmpID
ColumnName
ChangeType
Value
ChangeDate
1
FirstName
Update
FN1.1
2018-01-02
2
FirstName
Update
FN2.1
2018-01-02
2
MiddleName
Delete
-
2018-01-02
3
LastName
Update
LN3.1
2018-01-02
3
Designation
Update
D3.1
2018-01-02

After applying the staging changeto my target table, my target data should be like below.

EmpID
FirstName
MiddleName
LastName
Designation
UpdatedDate
1
FN1.1
MN1
LN1
D1
2018-01-02
2
FN2.1
-
LN2
D2
2018-01-02
3
FN3
MN3
LN3.1
D3.1
2018-01-02
4
FN4
MN4
LN4
D4
2018-01-01
5
FN5
MN5
LN5
D5
2018-01-01
6
FN6
MN6
LN6
D6
2018-01-01

I thought generating required dynamic SQL DML commands as per the staging(meta) data.
Could someone please share your thoughts if there is any best approach to achieve this?
December 28, 2018 at 1:27 am

This was removed by the editor as SPAM

Viewing 9 posts - 1 through 8 (of 8 total)

You must be logged in to reply to this topic. Login to reply

ChrisM@Work SSC Guru Points: 186127 More actions · Answer 1

IF OBJECT_ID('tempdb..#dbo_Employees') IS NOT NULL DROP TABLE #dbo_Employees;
SELECT * INTO #dbo_Employees FROM (VALUES
(1,CAST('FN1' AS VARCHAR(10)), CAST('MN1' AS VARCHAR(10)), CAST('LN1' AS VARCHAR(10)), CAST('D1' AS VARCHAR(10)), CAST('2018-01-01' AS DATE)),
(2,'FN2', 'MN2', 'LN2', 'D2', '2018-01-01'),
(3,'FN3', 'MN3', 'LN3', 'D3', '2018-01-01'),
(4,'FN4', 'MN4', 'LN4', 'D4', '2018-01-01'),
(5,'FN5', 'MN5', 'LN5', 'D5', '2018-01-01'),
(6,'FN6', 'MN6', 'LN6', 'D6', '2018-01-01')
) d (EmpID,FirstName,MiddleName,LastName,Designation,UpdatedDate)

IF OBJECT_ID('tempdb..#staging_Employees') IS NOT NULL DROP TABLE #staging_Employees;
SELECT * INTO #staging_Employees FROM (VALUES
(1, 'FirstName', 'Update', 'FN1.1', '2018-01-02'),
(2, 'FirstName', 'Update', 'FN2.1', '2018-01-02'),
(2, 'MiddleName', 'Delete', '-', '2018-01-02'),
(3, 'LastName', 'Update', 'LN3.1', '2018-01-02'),
(3, 'Designation', 'Update', 'D3.1', '2018-01-02')
) d (EmpID,ColumnName,ChangeType,Value,ChangeDate)

-- Look at what we're going to do
SELECT *
FROM #dbo_Employees de
CROSS APPLY (
SELECT
  FirstName = MAX(CASE WHEN se.ColumnName = 'FirstName' THEN se.[Value] ELSE NULL END),
  MiddleName = MAX(CASE WHEN se.ColumnName = 'MiddleName' THEN se.[Value] ELSE NULL END),
  LastName = MAX(CASE WHEN se.ColumnName = 'LastName' THEN se.[Value] ELSE NULL END),
  Designation = MAX(CASE WHEN se.ColumnName = 'Designation' THEN se.[Value] ELSE NULL END),
  UpdatedDate = MAX(se.ChangeDate)
FROM #staging_Employees se
WHERE se.EmpID = de.EmpID
GROUP BY se.EmpID
) x

-- Do it
UPDATE de SET
  FirstName = ISNULL(x.FirstName,de.FirstName),
  MiddleName = ISNULL(x.MiddleName,de.MiddleName),
  LastName = ISNULL(x.LastName,de.LastName),
  Designation = ISNULL(x.Designation,de.Designation),
  UpdatedDate = x.UpdatedDate
FROM #dbo_Employees de
CROSS APPLY (
SELECT
  FirstName = MAX(CASE WHEN se.ColumnName = 'FirstName' THEN se.[Value] ELSE NULL END),
  MiddleName = MAX(CASE WHEN se.ColumnName = 'MiddleName' THEN se.[Value] ELSE NULL END),
  LastName = MAX(CASE WHEN se.ColumnName = 'LastName' THEN se.[Value] ELSE NULL END),
  Designation = MAX(CASE WHEN se.ColumnName = 'Designation' THEN se.[Value] ELSE NULL END),
  UpdatedDate = MAX(se.ChangeDate)
FROM #staging_Employees se
WHERE se.EmpID = de.EmpID
GROUP BY se.EmpID
) x

-- Check if it has worked
SELECT * FROM #dbo_Employees

^{“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

Jonathan AC Roberts SSCoach Points: 18361 More actions · Answer 2

Jonathan AC Roberts

SSCoach

Points: 18361

November 30, 2018 at 5:35 am

#2015092

ChrisM's solution looks good.

fahey.jonathan Hall of Fame Points: 3617 More actions · Answer 3

You could also look at the MERGE statement, which will do inserts, updates, and deletes from a source data set to a target data set in a single statement. There are clauses for what to do when matching records are found (UPDATE), what to do when the source has records and the target does not (INSERT), and what to do when the source does not have records but the target does (DELETE). These can be combined so any combination of those can be used.

Documentation is found at Microsoft's website: https://docs.microsoft.com/en-us/sql/t-sql/statements/merge-transact-sql?view=sql-server-2017

fahey.jonathan Hall of Fame Points: 3617 More actions · Answer 4

fahey.jonathan - Friday, November 30, 2018 11:10 AM
You could also look at the MERGE statement, which will do inserts, updates, and deletes from a source data set to a target data set in a single statement. There are clauses for what to do when matching records are found (UPDATE), what to do when the source has records and the target does not (INSERT), and what to do when the source does not have records but the target does (DELETE). These can be combined so any combination of those can be used.
Documentation is found at Microsoft's website: https://docs.microsoft.com/en-us/sql/t-sql/statements/merge-transact-sql?view=sql-server-2017

On second look, your source data is not normalized, so it may be difficult to get your data into a usable format for MERGE. You could do one column at a time, like this:
WITH BaseData AS ( SELECT EmpID, Value AS FirstName, ROW_NUMBER() OVER (PARTITION BY EmpID ORDER BY ChangeDate DESC) AS RowNum FROM #staging_Employees WHERE ColumnName = 'FirstName' ) MERGE #dbo_Employees u USING ( SELECT EmpID, FirstName, RowNum FROM BaseData WHERE RowNum = 1 ) x ON x.EmpID = u.EmpID WHEN MATCHED THEN UPDATE SET FirstName = x.FirstName WHEN NOT MATCHED BY TARGET THEN INSERT (EmpID, FirstName) VALUES (x.EmpID, x.FirstName);
You could create a more complex query to get your staging data into a normalized format. I think that would be more difficult, because you have to account of multiple changes to the same item and select only the latest.

Rocky's Ten Centuries Points: 1102 More actions · Answer 5

fahey.jonathan, ChrisM@Work - Thanks for your inputs. I will consider these both option and choose one on their performance as we are having huge data in staging and target.

andycadley SSCertifiable Points: 5296 More actions · Answer 6

I'm not sure ChrisM's solution works in the general case - the various MAX's operate independently and so unless you can guarantee the values all ascend in time (which seems unlikely) you won't necessarily get the right end values. Try adding the line:

(1, 'FirstName', 'Update','A2','2018-02-02')

to the list and you'll see what I mean. You would need to rank the rows by date and take the most recent for any given field first to ensure you're getting the row values you actually require.

Jason A. Long SSC-Insane Points: 23786 More actions · Answer 7

@rockys - Is it possible that the same EmpID could have multiple updates for the same column, at the same time? If so, how would you determine which of the multiple values gets updated to the dbo.Employee table?
Looking at the Staging table, it gives the impression that it's wiped and reloaded on daily basis. If the UpdatedDate really does lack a time component, it could be difficult (more likely impossible) to determine which is the most resent.