This example covers easy queries of type

SELECT r.key, avg(r.val1), sum(r.val2) ...
FROM relation r
WHERE filter_condition(r)
GROUP BY r.key

The physical graph for such a select-project-aggregate query consists of three nodes:

• InputStub
• SelectProjectNode
• AggregateNode

The InputStub resembles the relation relation. The (stateless) SelectProjectNode is responsible for transforming tuples of relation to tuples of a relation where the filter-criterion is satisfied and only attributes are left that are needed for the aggregation. Finally, the (stateful) AggregateNode computes all aggregates simultaneously.

The relations involved here are the following:

SELECT * FROM relation r

For tuples arriving at and emitted from the InputStub. The SelectProjectNode transforms these to this relation:

SELECT r.key, r.val1, r.val2
FROM relation r
WHERE filter_condition(r)

And finally, the AggregateNode produces the resulting relation from the top of this page.