Datalog with First-Class Facts

Abstract: Datalog is a popular logic programming language for deductive reasoning tasks in a wide array of applications, including business analytics, program analysis, and ontological reasoning. However, Datalog's restriction to flat facts over atomic constants leads to challenges in working with tree-structured data, such as derivation trees or abstract syntax trees. To ameliorate Datalog's restrictions, popular extensions of Datalog support features such as existential quantification in rule heads (Datalog$^\pm$, Datalog$^\exists$) or algebraic data types (Souffl\'e). Unfortunately, these are imperfect solutions for reasoning over structured and recursive data types, with general existentials leading to complex implementations requiring unification, and ADTs unable to trigger rule evaluation and failing to support efficient indexing. We present DL$^{\exists!}$, a Datalog with first-class facts, wherein every fact is identified with a Skolem term unique to the fact. We show that this restriction offers an attractive price point for Datalog-based reasoning over tree-shaped data, demonstrating its application to databases, artificial intelligence, and programming languages. We implemented DL$^{\exists!}$ as a system \slog{}, which leverages the uniqueness restriction of DL$^{\exists!}$ to enable a communication-avoiding, massively-parallel implementation built on MPI. We show that Slog outperforms leading systems (Nemo, Vlog, RDFox, and Souffl\'e) on a variety of benchmarks, with the potential to scale to thousands of threads.

• The paper introduces an innovative Datalog extension that assigns unique Skolem terms to facts for improved logical reasoning.
• It implements Slog, a massively parallel Datalog engine that outperforms traditional systems in handling complex data.
• Applications in provenance, abstract interpretation, and type systems demonstrate the method’s versatility and practical impact.

An Expert Overview of "Datalog with First-Class Facts"

The paper "Datalog with First-Class Facts" presents an advanced extension to the conventional Datalog programming language, targeting its syntactic limitations in handling tree-structured data. This work introduces a novel variant of Datalog that employs first-class facts with unique Skolem terms to identify each fact, enhancing its applicability in domains such as databases, artificial intelligence, and programming languages.

Motivation and Challenges

Datalog is a long-standing logic programming language adept at recursive query processing and is widely used in diverse arenas such as business analytics, program analysis, and ontological reasoning. However, traditional Datalog is restricted to flat facts involving atomic constants, which imposes limitations on its expressibility, especially when reasoning about structured and recursive data types like derivation trees or abstract syntax trees. Existing enhancements to Datalog, such as Datalog$^\pm$ and Datalog$^\exists$, offer partial solutions using existential quantification or algebraic data types (ADTs). However, these extensions encounter performance and complexity bottlenecks, particularly when triggering rule evaluations and indexing efficiency.

Contributions

The paper's primary contribution is the introduction of a Datalog extension that facilitates reasoning with first-class facts. This is achieved by assigning each fact a unique identity via a Skolem term. This extension simplifies implementations by avoiding unification challenges and making indexing more efficient, thus offering superior support for parallel computations:

1. Introduction of First-Class Facts: The paper formalizes syntax and semantics for a variant of Datalog with unique existential quantification, ensuring that every fact is identified by its unique Skolem term. With a chase-based semantics, the system efficiently handles fact generation and the avoidance of cyclic references.
2. Implementation of Slog: The authors introduced a practical system named Slog, an experimental language that extends the Datalog engine by leveraging their methodology. Slog utilizes a massively parallel implementation built on MPI, showcasing significant performance improvements over existing systems like Nemo and RDFox.
3. Applications Demonstrating Versatility: Various applications illustrate the system's broad applicability, including provenance, abstract interpretation, and implementing type systems. The paper shows that Slog effectively models complex systems due to its flexible handling of nested facts and its suitability for functional programming paradigms.

Keywords

• Provenance: Slog's capability to handle provenance or lineage is particularly noteworthy. It supports both eager and on-demand calculation of why-provenance, crucial for understanding the origins of computed facts.
• Functional Programming: By allowing first-class facts, Slog facilitates ad-hoc polymorphic rules similar to those found in functional programming.
• Parallelism and Performance: Slog's highly parallel architecture circumvents Datalog's conventional performance bottlenecks. It yields an efficient execution on multi-core and distributed systems, showing optimal speed-ups with increased thread counts during benchmarks.

Implications and Future Directions

The advancements introduced in this paper hold substantial implications for artificial intelligence and database systems where scalable, efficient reasoning over complex data shapes is essential. The strong numerical results, especially regarding Slog's capacity to outperform current leading systems, substantiate the claims of the paper. As the authors speculate, future work could further explore optimizing data parallelism and extending the syste-syntactic structures for comprehensive coverage of recursive data and computation patterns.

In summary, the paper "Datalog with First-Class Facts" contributes an innovative approach to extending the expressiveness of Datalog, making it highly relevant for modern logical reasoning and programming applications. Future work may include refining its parallelism further and exploring its integration with broader data management and AI tools.