SQL Merge Statement not working with Batch Data - sql

Question

I have created a stored procedure using MERGE statement as follows.
MERGE Tasks S
USING (SELECT id, name, field1, field2, field3 FROM Tasks_Temp) SS
ON S.id=SS.id
WHEN MATCHED
THEN UPDATE SET S.name=SS.name, S.field1=SS.field1, S.field2=SS.field2, S.field3=SS.field3
WHEN NOT MATCHED
THEN INSERT (id, name, field1, field2, field3) VALUES (SS.id, SS.name, SS.field1, SS.field2, SS.field3)
Tasks_temp:
id
name
field1
field2
field3
1
Task1
NULL
NULL
STARTED
2
Task2
SUBMITTED
NULL
STARTED
1
Task1
SUBMITTED
NULL
STARTED
3
Task3
NULL
NULL
STARTED
1
Task1
APPROVED
NULL
STARTED
2
Task2
APPROVED
NULL
STARTED
Tasks:(required)
id
name
field1
field2
field3
1
Task1
APPROVED
NULL
STARTED
2
Task2
APPROVED
NULL
STARTED
3
Task3
NULL
NULL
STARTED
Now here the Tasks_Temp table is empty at first and then it's filled with a sequential form of data as displayed where for each ID there's multiple changes like the field1, field2 or field3. The merge statement is inserting all the values into the other table, even duplicate IDs i.e., exactly the records as present in Tasks_Temp.
The next time when I run it gives me errors saying
Trying to update multiple records not allowed
Is merge not able to handle batch data like this? Even if I ignore getting the latest data for each fields MERGE still inserts the same number of rows as present in Tasks_Temp table. If not, what are some other optimal approaches?
I feel that even if merge worked, it wouldn't necessarily mean it's taking the most updated field values for an id.
However, if I have a timestamp provided, how can I insert the most recent records from the Tasks_Temp table to the Tasks table?

Answer 1

CREATE TABLE Tasks_Temp(
id INTEGER NOT NULL
,name VARCHAR(100) NOT NULL
,field1 VARCHAR(100)
,field2 VARCHAR(100)
,field3 VARCHAR(100) NOT NULL
);
INSERT INTO Tasks_Temp
(id,name,field1,field2,field3) VALUES
(1,'Task1',NULL,NULL,'STARTED'),
(2,'Task2','SUBMITTED',NULL,'STARTED'),
(1,'Task1','SUBMITTED',NULL,'STARTED'),
(3,'Task3',NULL,NULL,'STARTED'),
(1,'Task1','APPROVED',NULL,'STARTED'),
(2,'Task2','APPROVED',NULL,'STARTED');
As #Larnu said you should use a criterion for choosing same id, based on my perspective it should be APPROVED>SUBMITTED>NULL for field1
hence I use ROW_NUMBER alphabetically and filter best values of Tasks_temp as follows:
SELECT id,
 NAME,
       field1,
       field2,
       field3
FROM   (SELECT id,
               NAME,
               field1,
               Row_number()
                 OVER (
                   partition BY NAME
                   ORDER BY IIF(field1 IS NULL, 'xxx', field1)) Stats,
               field2,
               field3
        FROM   tasks_temp) a
WHERE  stats = 1
That gives following table
id
name
field1
field2
field3
1
Task1
APPROVED
null
STARTED
2
Task2
APPROVED
null
STARTED
3
Task3
null
null
STARTED
Hence you should use the above written query as a subquery for use with the MERGE in order to facilitate the operation:
MERGE Tasks S
USING (SELECT id,
       NAME,
       field1,
       field2,
       field3
FROM   (SELECT id,
               NAME,
               field1,
               Row_number()
                 OVER (
                   partition BY NAME
                   ORDER BY Iif(field1 IS NULL, 'xxx', field1)) Stats,
               field2,
               field3
        FROM   tasks_temp) a
WHERE  stats = 1 ) SS
ON S.id=SS.id
WHEN MATCHED
THEN UPDATE SET S.name=SS.name, S.field1=SS.field1, S.field2=SS.field2, S.field3=SS.field3
WHEN NOT MATCHED
THEN INSERT (id, name, field1, field2, field3) VALUES (SS.id, SS.name, SS.field1, SS.field2, SS.field3)
dbfiddle