Logical Foundations and Complexity of 4QL, a Query Language with Unrestricted Negation

The paper discusses properties of a DATALOG$^{\neg\neg}$-like query language 4QL, originally outlined in [MS10]. Negated literals in heads of rules naturally lead to inconsistencies. On the other hand, rules do not have to attach meaning to some literals. Therefore 4QL is founded on a four-valued semantics, employing the logic introduced in [MSV08, VMS09] with truth values: ’true’, ‘false’, ‘inconsistent’ and ‘unknown’. 4QL allows one to use rules with negation in heads and bodies of rules, it is based on a simple and intuitive semantics and provides uniform tools for “lightweight” versions of known forms of nonmonotonic reasoning. In addition, 4QL is tractable as regards its data complexity and captures PTIME queries. Even if DATALOG$^{\neg\neg}$ is known as a concept for the last 30 years, to our best knowledge no existing approach enjoys these properties. In the current paper we: - investigate properties of well-supported models of 4QL - prove the correctness of the algorithm for computing well-supported models - show that 4QL has PTIME data complexity and captures PTIME.

💡 Research Summary

The paper introduces 4QL, a DATALOG‑like query language that permits unrestricted negation both in rule bodies and heads. Traditional DATALOG⁽¬¬⁾ approaches encounter consistency problems when negated literals appear in heads, because the semantics of classical two‑valued logic cannot accommodate contradictory information. 4QL resolves this by adopting a four‑valued semantics originally proposed in the works of M. S. Vardi and collaborators (MSV08, VMS09). The four truth values are true, false, inconsistent, and unknown. An atom can simultaneously carry a truth and a falsity component, allowing the system to represent contradictions without collapsing.

The authors formalize the notion of a well‑supported model (WSM). A WSM is a minimal interpretation that respects the dependency graph of the program: every assigned value must be justified by a rule whose premises are already justified, and cycles are broken by assigning the value inconsistent where necessary. They prove that for any finite 4QL program a unique WSM exists, establishing both existence and uniqueness.

To compute the WSM, the paper presents a deterministic algorithm that proceeds in two phases per stratum of the program’s topological ordering:

1. Normalization – Detects contradictions within the current stratum and marks the affected atoms as inconsistent.
2. Propagation – Propagates the newly assigned values to higher strata, updating the truth status of dependent atoms.

The algorithm iterates over all strata, guaranteeing termination after a number of steps linear in the size of the database plus the number of rules. The authors prove its correctness by showing that the final interpretation coincides with the unique WSM defined earlier.

Complexity analysis yields two central results:

• Data complexity – For a fixed set of rules, the evaluation runs in polynomial time with respect to the size of the extensional database (O(n·|P|), where n is the number of ground facts and |P| the number of rules). Hence 4QL belongs to PTIME.
• Expressive power – The language captures exactly PTIME on ordered databases. The authors construct a translation from any deterministic polynomial‑time Turing machine (or, equivalently, any PTIME query expressed in first‑order logic with a least‑fixed‑point operator) into a 4QL program that yields the same result on every input. This demonstrates that 4QL is both tractable and expressive enough to represent all PTIME queries.

Beyond the theoretical contributions, the paper argues that 4QL offers a “lightweight” framework for non‑monotonic reasoning. Because contradictions are first‑class citizens, applications such as data integration, knowledge‑graph querying, and reasoning over incomplete or noisy information can exploit unrestricted negation without risking global inconsistency. The four‑valued semantics provides a clear, intuitive handling of unknown and inconsistent states, enabling developers to reason about the provenance of data and to design policies for conflict resolution.

The authors conclude with several avenues for future work: extending 4QL with aggregate functions and set operations, implementing a prototype engine to evaluate performance on real‑world datasets, and comparing 4QL’s approach to other non‑monotonic formalisms such as answer‑set programming, well‑founded semantics, and paraconsistent logics. Overall, the paper establishes 4QL as a novel, theoretically sound, and practically promising query language that uniquely combines unrestricted negation, four‑valued semantics, PTIME data complexity, and full PTIME expressiveness.