Reset a running total if field value is hit - sql-server-2008

Question

I am able to keep a running total with the below query, and it does that just fine. What I really want, is when the gap field is greater than or equal to the date_diff field, the running total should reset back to the current hrly_qty. I'm sure I could achieve my results with a cursor, but I wanted to know of possible other ways. Ideas?
Example:
WITH CTE AS
(
SELECT a.*, SUM(b.hrly_qty) AS running_total, c.gap
FROM #tmpTrxhist2 a
INNER JOIN #tmpTrxhist2 b ON a.people_id = b.people_id
AND b.sequence_id <= a.sequence_id
INNER JOIN incent_level c ON a.owner_division_id = c.owner_division_id
GROUP BY a.date_diff, a.owner_division_id, a.people_id, a.sequence_id,
a.hrly_qty, c.gap
)
SELECT * FROM CTE
ORDER BY people_id, sequence_id

Answer 1

You seldom need a cursor! A simple case statement should suffice. Of the top of my head something like:
WITH CTE AS
(
SELECT a.*, SUM(b.hrly_qty) AS running_total, c.gap
FROM #tmpTrxhist2 a
INNER JOIN #tmpTrxhist2 b ON a.people_id = b.people_id
AND b.sequence_id <= a.sequence_id
INNER JOIN incent_level c ON a.owner_division_id = c.owner_division_id
GROUP BY a.date_diff, a.owner_division_id, a.people_id, a.sequence_id,
a.hrly_qty, c.gap
)
SELECT *,
CASE
WHEN gap >= date_diff then hrly_qty else gap
END as comp_gap
FROM CTE
ORDER BY people_id, sequence_id
Run the execution plans between the two but SQL is almost always much happier trying to optimise non-cursored code. In my field at least, you usually find cursors are over used and abused by 'proper' C/C++ programmers because they get in their comfort zone when see something that looks a bit like a while-loop rather than think about sets of data. There is a place for cursors, but this isn't it!