A "window function" calculates a value based on input from more than one row (like an aggregate function). The result is produced without grouping the source rows (**unlike** an aggregate function).

The input row set for the calculation, or "window", is defined by an `OVER` clause. If, and only if, an `OVER` clause is present, then the function is a window function. Otherwise it is an ordinary function (either aggregate or scalar).


[![Railroad](/image?hash=97e74e1bb698bfaa457fa807e076acfa78f9465ce32c64f398da1a1915c42dc9)](https://tabatkins.github.io/railroad-diagrams/generator.html#Diagram(%0A%20%20NonTerminal('function')%0A%20%2COptional(NonTerminal('filter-clause')%2C'skip')%0A%20%2CTerminal('OVER')%0A%20%2CTerminal('(')%0A%20%2COptional(Sequence(Terminal('PARTITION')%2CTerminal('BY')%2COneOrMore(NonTerminal('expr')%2C'%2C'))%2C'skip')%0A%20%2COptional(Sequence(Terminal('ORDER')%2CTerminal('BY')%2COneOrMore(NonTerminal('expr')%2C'%2C'))%2C'skip')%0A%20%2COptional(NonTerminal('frame-specification')%2C'skip')%0A%20%2CTerminal(')')%0A)%0A%0A)

* The optional `PARTITION BY` clause defines the set of rows to use as input (if not specified all rows are used as input).
* The optional `ORDER BY` clause defines the order in which the rows are input to the function.

# See Also

* [SQLite Documentation](https://www.sqlite.org/windowfunctions.html)
* [SQLite Tutorial](https://www.sqlitetutorial.net/sqlite-window-functions/)