Window Function Overview

Window Functions are special functions where the result is dependent on the order and grouping of rows. The “window” for a Window Function is defined by the grouping at that level. The result of a window function is determined only by the rows within the window. For example, given a table of city populations, if you grouped by State and then Rank'ed the cities by Population, you would have an independent ranking for each State. If there is no grouping, the “window” is the entire table.

For most Window Functions, the results are dependent on how the rows within the window are sorted. Many Window Functions will only work properly if the table is uniquely sorted by a column in the same “window” as the Window Function. To ensure that there is a unique sort order without any duplicate values, it is often useful to use multi-column sorting criteria to clearly define how to handle duplicate values.

The complexity of Window Functions makes them a little harder to use, but they are also a very powerful tool.

Cumulative Window Functions

Cumulative Window Functions are aggregate functions that apply to all of the rows up-to and including the current row.

The window that Cumulative Window Functions are applied in must be uniquely sorted for the function to work properly.

Moving Window Functions

Moving Window Functions are aggregate functions which apply to a “window” bracketing the current row. A common moving Window Function is a “moving average”.

The window is specified as a certain number of rows above and below the current row, constrained by the grouping.

The window that Moving Window Functions are applied in must be uniquely sorted for the function to work properly.

Shifting Window Functions

Shifting Window functions either shift the values in a column or repeat a specified value. The functions are useful when calculating month over month data.

The window that Shifting Window Functions are applied in must be uniquely sorted for the function to work properly.