Additive first-order queries

Abstract

A database query q is called additive if q(A ∪ B) = q(A) ∪ q(B) for domain-disjoint input databases A and B. Additivity allows the computation of the query result to be parallelised over the connected components of the input database. We define the “connected formulas” as a syntactic fragment of first-order logic, and show that a first-order query is additive if and only if it expressible by a connected formula. This characterisation specializes to the guarded fragment of first-order logic. We also show that additivity is decidable for formulas of the guarded fragment, establish the computational complexity, and do the same for positive-existential formulas. Our results hold when restricting attention to finite structures, as is common in database theory, but also hold in the unrestricted setting.