On the Expressive Power of Multiple Heads in CHR

Constraint Handling Rules (CHR) is a committed-choice declarative language which has been originally designed for writing constraint solvers and which is nowadays a general purpose language. CHR programs consist of multi-headed guarded rules which allow to rewrite constraints into simpler ones until a solved form is reached. Many empirical evidences suggest that multiple heads augment the expressive power of the language, however no formal result in this direction has been proved, so far. In the first part of this paper we analyze the Turing completeness of CHR with respect to the underneath constraint theory. We prove that if the constraint theory is powerful enough then restricting to single head rules does not affect the Turing completeness of the language. On the other hand, differently from the case of the multi-headed language, the single head CHR language is not Turing powerful when the underlying signature (for the constraint theory) does not contain function symbols. In the second part we prove that, no matter which constraint theory is considered, under some reasonable assumptions it is not possible to encode the CHR language (with multi-headed rules) into a single headed language while preserving the semantics of the programs. We also show that, under some stronger assumptions, considering an increasing number of atoms in the head of a rule augments the expressive power of the language. These results provide a formal proof for the claim that multiple heads augment the expressive power of the CHR language.

💡 Research Summary

The paper investigates the expressive power contributed by multiple heads in Constraint Handling Rules (CHR), a committed‑choice declarative language originally devised for constraint solvers but now used as a general‑purpose programming language. CHR programs consist of guarded simplification and propagation rules whose heads are multisets of user‑defined constraints. The authors provide the first formal results showing that allowing more than one atom in a rule head genuinely increases the language’s expressive capabilities.

The study is divided into two main parts. In the first part the authors examine Turing completeness of CHR relative to the underlying constraint theory (CT). They prove that if CT includes an equality predicate interpreted as pattern matching and the signature contains at least one function symbol of arity > 0, then restricting CHR to single‑head rules (denoted CHR₁) does not diminish Turing completeness. The proof proceeds by encoding Minsky’s two‑counter machine into CHR₁, using the function symbol to simulate counter increments, decrements, and zero tests. Conversely, when the signature lacks function symbols and consists only of finitely many constants, CHR₁ fails to be Turing‑complete, while full CHR (with multi‑heads) remains Turing‑complete. This establishes a clear computational gap that depends on the presence of multiple heads.

The second part shifts focus from raw computational power to language‑level expressiveness via the notion of language encoding. An encoding is required to preserve observable behaviour and to satisfy compositionality constraints on goal translation. The authors consider two standard CHR observables: (i) data‑sufficient answers (final built‑in constraint store after successful termination) and (ii) qualified answers (final built‑in store together with any remaining user‑defined constraints). Under three reasonable assumptions—(a) the source and target languages share the same CT, (b) goal translation is compositional with respect to conjunction, and (c) no restriction is placed on the translation of program rules—they prove that no encoding can map full CHR into CHR₁ while preserving either observable. The core argument is that multi‑head rules can simultaneously match and remove several constraints, a capability that cannot be simulated by any sequence of single‑head rule applications without altering the set of answers.

A further contribution concerns the hierarchy induced by limiting the number of atoms allowed in rule heads. For each n ≥ 1, CHRₙ denotes CHR where each rule head contains at most n atoms. The authors show that for any m < n, CHRₙ cannot be encoded into CHRₘ under slightly stronger compositionality assumptions. This demonstrates a strict increase in expressive power as the head arity grows.

Methodologically, the paper combines classic Turing‑machine simulation techniques, model‑theoretic properties of constraint theories, and formal language‑encoding criteria (including compositionality and preservation of data‑sufficient/qualified answers). The negative encoding results are established by constructing specific CHR programs whose answer sets cannot be reproduced by any program in the target language, thereby yielding a contradiction.

In summary, the paper delivers four principal findings: (1) CHR₁ is Turing‑complete when the underlying signature contains function symbols, but not otherwise; (2) multi‑head CHR is strictly more expressive than CHR₁ under natural observational and compositional constraints; (3) increasing the maximum number of atoms in rule heads yields a proper hierarchy of expressive power; and (4) the expressive advantage of multiple heads is not merely syntactic convenience but a fundamental computational property of CHR. These results provide the first rigorous justification for the long‑standing belief that multi‑head rules are essential for the full power of CHR.