I'm not even sure if that is the correct terminology, so let me start with some sample data. Say I have a table "data" with colums "id", "datetime", "total".

Say the data looks something like

ID DATE TOTAL
1 2009-10-01 00:00:00 10
2 2009-10-01 01:00:00 15
3 2009-10-01 02:00:00 17
4 2009-10-01 03:00:00 19
5 2009-10-01 04:00:00 21
6 2009-10-02 00:00:00 25
7 2009-10-02 01:00:00 30
8 2009-10-03 02:00:00 35
9 2009-10-03 03:00:00 45
10 2009-10-04 04:00:00 50

So, what I'd like to return is the total for each day, so what I get back would look something like:

2009-10-01 11
2009-10-02 5
2009-10-03 10
2009-10-04 10

Basically what I'm doing is storing a cumulative total that is ever growing, but I'd like to get the totals for each day. Hence, 2009-10-01 saw a growth of 11 (21 - 10).

I know how to do this through queries/programming. But I'm wondering if there's a more elegant way to do this with just one query.

Thanks!

Why do you want 2009-10-04 10 to be returned? That day has only one total-value.

For the rest of the rows the following query returns the result you are asking for:
SELECT date, MAX(total) - MIN(total)
FROM tablename
GROUP BY date


Note that DATE is a reserved word in ANSI SQL. It's a good idea to avoid reserved words as column names. (Or do at least double quote them, i.e "DATE".)