Code Structure

This is an older illustration of Clash’s structure, however, it should still give a hint on how things are organized.

The current intra-submodule dependencies are illustrated on the bottom of this page. Remarkable notes about the submodules:

• query: the query contains the abstract description of relations (because queries are relations) and the parser for transforming SQL to a relation.
• physicalgraph: this is the target of optimization, and basically an abstract version of a storm topology. It contains the nodes, edges, and rules that define a processing strategy.
• optimizer: this is responsible for creating a physical graph from the query
• tpch and join-order-benchmark are quality-of-life modules that contain queries and data characteristics
• documents are the elements that are sent through topologies and are contained inside relations
• workers are the actual implementations of functionality local to bolts (e.g. the join or aggregation components)
• local is a single-threaded simulation of storm which should make it easier to test new implementations