2 Week 2: Datalog Implementation

2.1 Tuesday, September 1

Trie-based Datalog engines. We will discuss Souffle, the current state-of-the-art Datalog solver based on high-performance tries.

SOUFFLE´ : On Synthesis of Program Analyzers

2.2 Thursday, September 3

Trie-based Datalog engines achieve efficiency and scalability by keeping multiple copies of relations, indexed according to their usage within the program. This allows for an extremely fast implementation of joins, the core (interest) primitive from relational algebra.

Automatic Index Selection for Large-Scale Datalog Computation