Question

I have a database that looks like this:

label_id, scan_type, scan_cost

And some rows that look like this:

001, A40, 70
001, A40, 70
002, A40, 85
003, A40, 85
003, A40, 85

I need to produce a result set that looks like this:

001, A40, 70
001, A40, 0
002, A40, 85
003, A40, 85
003, A40, 0

That is, any multiple occurrence of the same label_id then the scan_cost column needs to be set to 0, but the first occurrence of each label_id the value needs to remain untouched.

The label_id are not sequential if that changes anything.

Is it possible to achieve this behaviour in SQL? Note the SQL dialect is T-SQL, Microsoft SQL Server 2008

Answer 1

This query returns what you are looking for:

SELECT
    label_id
,   scan_type
,   CASE WHEN row_number() OVER (PARTITION BY label_id ORDER BY scan_type)=1
        THEN scan_cost
        ELSE 0
    END
FROM test
ORDER BY label_id, scan_type

The idea behind this solution is to partition the data by label_id, and use the ROW_NUMBER function to decide which data to keep.

I used scan_type in the ORDER BY sub-clause, which is not ideal. If your actual table has a column with data that is better suited to determine what row is first, e.g. a timestamp column, you should use that other column instead. The second column in the outer order by needs to be the same as the column in the inner order by.

Here is a link to this query on sqlfiddle.