Tasks

Snowflake uses tasks to schedule and automate data loading and transformation. Although data movement is not tracked in relational modeling, it is an integral part of transformational modeling and is covered here for completeness.

Tasks automate data pipelines by executing SQL in serial or parallel steps. Tasks can be combined with streams for continuous ELT workflows to process recently changed table rows. This can be done serverlessly (using auto-scalable Snowflake-managed compute clusters that do not require an active warehouse) or using a dedicated user-defined warehouse.

The code for creating a task is as follows:

CREATE TASK <task_name>
...
[ AFTER <parent_task_1> [ , <parent_task_2> , ... ] ]
[ WHEN <boolean_expr> ]
AS <sql>

Tasks are simple to understand—they run a SQL command (or execute a stored procedure) on a schedule or when called as part of a parent task. The following figure shows how tasks can be chained serially...