Question

I am looking to update a calculated sum in sql

Basically I have a table:

ImportID  SeiralNumber   Day   Hour     value    Difference   Complete
1             123         1     1         6         NULL        0
2             123         1     2         8         NULL        0
3             123         1     5         21        NULL        0
4             123         1     6         28        NULL        0
5             222         2     2         12        NULL        0
6             222         2     5         18        NULL        0
7             222         2     4         16        NULL        0
8             222         1     12        8         NULL        0

For each serial number there will be a day 1-365 and hour through 1-12, all I want to do is calculate the difference filed from the record before

So take ImportID 6, I need to get the record which is on the same day and the hour before (importID 7) then I need to update the Difference using the value field which is 18 -17 = 1.

N.B. There may be gaps in the sequence and if there is no previous record then the difference should stay as NULL. Once they have been calculated they need to be inserted into a new table only when the difference is now not null and it doesn’t exist in the table already, on a successful insert they get marked as complete. Also a record before can be a previous day (day 1 hour 12) is the record before (day 2, hour 1)

Currently I am using a loop to select the null values, get the previous record, update the record, if its OK insert into other table, update the Completed field.

My issue is that this is working on a million records and it is taking a long while to Select the applicable records (completed = 0) into a temp table and loop through each.

Is there any quicker way to mass process these as an update statement? Or separate statements?

The result should be

ImportID  SeiralNumber   Day   Hour     value    Difference   Complete
1             123         1     1         6         NULL        0
2             123         1     2         8         2           1
3             123         1     5         21        NULL        0
4             123         1     6         28        7           1
5             222         2     1         12        4           1
6             222         2     5         18        2           1
7             222         2     4         16        NULL        0
8             222         1     12        8         NULL        0

Thanks in advance

Answer 1

I think this is basically it isn’t it?

DECLARE @TABLE TABLE
(
    ImportId INT,
    SerialNumber INT,
    Day INT,
    Hour INT,
    Value INT,
    Difference INT,
    Complete INT
)

INSERT INTO @TABLE VALUES
(1,123,1,1,6,NULL,0),
(2,123,1,2,8,NULL,0),
(3,123,1,5,21,NULL,0),
(4,123,1,6,28,NULL,0),
(5,222,2,1,12,NULL,0),
(6,222,2,5,18,NULL,0),
(7,222,2,4,16,NULL,0),
(8,222,1,12,8,NULL,0)

SELECT * FROM @Table

UPDATE T
    SET T.Difference = T.Value - TT.Value,
    Complete = 1
FROM
(
    SELECT *
    ,ROW_NUMBER() OVER (PARTITION BY SerialNumber ORDER BY Day ASC, Hour ASC) AS RowCounter
    FROM @TABLE
    WHERE Complete = 0 --Ignore completed ones  
)AS T
INNER JOIN
(
    SELECT *
    ,ROW_NUMBER() OVER (PARTITION BY SerialNumber ORDER BY Day ASC, Hour ASC) AS RowCounter
    FROM @TABLE
)AS TT
ON T.SerialNumber = TT.SerialNumber
WHERE
(
    T.RowCounter = TT.RowCounter + 1
    AND
    T.Day = TT.Day 
    AND
    T.Hour = TT.Hour + 1
)
OR
(
    T.Day = TT.Day + 1 
    AND
    T.Hour = 1
    AND
    TT.Hour = 12
)

SELECT * FROM @TABLE