SQL in a Nutshell, 3rd Edition

Grouping or Windowing

Many analytic functions also allow you to specify a virtual, moving window surrounding a row within a partition, using the framing clause. Such moving windows are useful for calculations such as a running total.

The following Oracle-based example uses the framing clause on the analytic variant of SUM to calculate a running sum of the values in the first column. No partitioning clause is used, so each invocation of SUM operates over the entire result set. However, the ORDER BY clause sorts the rows for SUM in ascending order of NUM’s value, and the BETWEEN clause (which is the windowing clause) causes each invocation of SUM to include values for NUM only up through the current row. Each successive invocation of SUM includes yet another value for NUM, in order, from the lowest value of NUM to the greatest:

SELECT NUM, SUM(NUM) OVER (ORDER BY NUM ROWS
 BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW) S FROM ODD_NUMS;
NUM            S
---------      ----------
0              0
1              1
2              3
3              6

This example’s a bit too easy, as the order of the final result set happens to match the order of the running total. That doesn’t need to be the case. The following example generates the same results, but in a different order. You can see that the running total values are appropriate for each value of NUM, but the rows are presented in a different order than before. The result set ordering is completely independent of the ordering used for window function calculations:

SELECT NUM, SUM(NUM) OVER (ORDER BY NUM ROWS ...