Apart from grouping and aggregation, PostgreSQL provides another way to perform computations based on the values of several records. It can be done using Window functions. Grouping and aggregation implies the output of a single record for every group of input records. Window functions can do similar things, but they are executed for every record, and the number of records in the output and the input is the same:

In the preceding diagram, the rectangles represent the records of a table. Let's assume that the color of the rectangles indicates the value of a field used to group the records. When GROUP BY is used in a query, each distinct value of that field will create a group and each group will become a single record in the results of the query. This was explained in Chapter 05, SQL Language. Window functions provide access to the values of all the records of the group (which is called a partition in this case), although the number of records in the result set stays the same. When window functions are used, no grouping is necessary, although possible.

Window functions are evaluated after grouping and aggregation. For this reason, the only places in the SELECT query where the window functions are allowed are select-list and the ORDER BY clause.