Constrained Higher-Order RPO & Reduction Pairs

• The paper introduces novel HORPO variants and constrained reduction pairs that prove termination for higher-order rewriting systems by integrating logical constraints and theory entailments.
• Constrained orderings are defined frameworks that incorporate logical side conditions and curried, partially applied functions to enhance termination proofs in LCSTRSs and LCTRSs.
• The methodology employs filtered argument comparisons and structured constraint propagation to ensure well-founded, stable orderings critical for dependency pair frameworks.

Constrained orderings, specifically higher-order recursive path orderings (HORPO) adapted for logical constraints and reduction pair frameworks, provide foundational tools for proving termination in higher-order rewriting systems enriched with background theories and logical side conditions. This class of techniques is crucial for the analysis of logically constrained simply typed rewriting systems (LCSTRSs) and higher-order logically constrained term rewriting systems (LCTRSs), accommodating both curried application and first-order logical constraints in the ordering (Kop, 2024, Guo et al., 2023).

1. Logically Constrained Rewriting and the Need for Constrained Orderings

LCSTRSs and higher-order LCTRSs generalize classical rewriting frameworks by incorporating:

• Theory or value symbols: Certain function symbols’ ground instances are interpreted in a background theory (such as integer arithmetic), allowing terms like $0$, $\text{succ}$, $\text{pred}$ to refer to values within a fixed model.
• Logical constraints: Rewrite rules and comparisons are equipped with side conditions $\varphi$, first-order formulae in the background theory, restricting the applicability of rewriting and orientation.
• Curried, higher-order signatures: Function symbols may be partially applied, necessitating orderings that handle terms like $f\ x$ vs. $f\ x\ y$ in the presence of partial application.

Classical HORPO is inadequate in these settings, as it assumes maximally applied function symbols and is unconstrained by logical side-conditions. The principal technical problems are orienting partially applied terms and integrating theory reasoning with syntactic ordering.

2. HORPO-Variant for Constrained, Curried Systems

A specialized HORPO-variant for LCSTRSs employs two main mechanisms (Kop, 2024):

• Argument filter $\pi$: Assigns to each non-theory symbol $f$ of arity $m$ a set $\pi(f) \subseteq \{1,\ldots,m\}$ of active argument positions, restricting recursive comparison to these selected arguments. This enables orientation in the curried setting where partial applications may arise.
• Constrained reduction pair: All strict comparisons $s \succ t$ are augmented to $s \succ_{(\varphi)} t$, requiring the constraint $\varphi$ to be provable in the theory for the comparison to hold. The ordering is systematically embedded as a pair $(\succeq, \succ)$, with both components augmented by logical constraints.

Basic Structure

Given types (base sort $\iota$ and arrow types $\eta\rightarrow \theta$), non-value symbols (with partial application and argument filters), and background orders $\succ_T$ on the theory, the following inductive relations are defined:

• Equivalence $\approx$ of curried terms via $\pi$ and precedence $\geq_F$,
• Unconstrained covering orderings $\geq_0$ (quasi-order) and $>_0$ (well-founded strict order), using rules for curried application, filtered-argument comparison, and precedence,
• Constrained reduction pair, where $s \succeq_{(\varphi)} t$ and $s \succ_{(\varphi)} t$ carry constraint $\varphi$ and a value-variable set $L$, enforcing theory-order entailment for pure theory terms and stable syntactic comparison elsewhere.

The resulting $(\succeq, \succ)$ forms a constrained reduction pair, with:

• $\succeq$ reflexive, monotonic, stable,
• $\succ$ well-founded, stable,
• strict compatibility: $\succ \subseteq \succeq$, $\succeq \circ \succ \subseteq \succ$, $\succ \circ \succeq \subseteq \succ$, ensuring suitability for termination proofs in the dependency pair (DP) framework.

3. Higher-Order Constrained HORPO for LCTRSs

The constrained HORPO for higher-order LCTRSs extends classical HORPO by threading logical constraints $C$ through each comparison and by supporting both lexicographic and multiset status (Guo et al., 2023). Essential ingredients are:

• Sorts and types: Including distinguished theory sorts $S^{th}$,
• Theory orders $>_A$: Well-founded orders on interpretation domains $X_A$,
• Precedence and status: Precedence $>$ on function symbols and theory-symbol separation, status map $stat(f)\in\{\mathrm{lex}, m_k\}$,
• Constraint propagation: All relations $>_C$ and $\geq_C$ are defined inductively, carrying constraint $C$ and closed under context and substitution.

Structural Clauses

• Theory-term comparison: For $s,t$ theory terms of sort $A$, $C \models s >_A t$ (entailment in background theory).
• Lexicographic/multiset extension: When comparing heads $f$, $g$, the order is extended to the argument vectors by $stat(f)$.
• Head dominance and decomposition: Precedence controls application cases and subterm descent.
• Substitution and constraint stability: If $C \vDash D$ and $s >_D t$, then $s >_C t$. The ordering is stable under substitution respecting the theory sorts.

These properties establish $(\geq_C, >_C)$ as a constrained reduction pair, compatible with abstract DP frameworks.

4. Principal Meta-Theoretical Properties

Both approaches establish:

• Well-foundedness: Ensured by a type-driven computability argument or induction on a measure combining syntactic and theory subterms—for any fixed constraint, no infinite descending chains in $>_C$ or $>_0$ exist.
• Monotonicity and stability: $\succeq$ (or $\geq_C$) is context- and substitution-monotonic; strict monotonicity is generally weakened ("weak monotonicity"), as only the covering order must be strictly monotonic for DP applicability (Kop, 2024).
• Soundness in termination frameworks: If every DP and rule is oriented (i.e., $\ell \succ_{(C)} r$ for every DP $\ell \rightarrow_{C} r$), then termination is guaranteed via reduction-pair results (Guo et al., 2023).

5. Exemplars and Comparative Features

Constraint-driven orientation:

• Example: For $$f:\mathbb{N}\rightarrow\mathbb{N}\rightarrow\mathbb{N}$$, comparing $f\ x\ y$ versus $f\ z\ z$ under constraint $x > y$, one shows $f\ x\ y \succ_{(x>y)} f\ z\ z$ by applying filtered argument comparison and checking theory entailments for each argument, reducing to background-theory inequalities (Kop, 2024).

Feature Summary Table

Aspect	HORPO-variant for LCSTRS (Kop, 2024)	Constrained HORPO for LCTRSs (Guo et al., 2023)
Constraint syntax	General 1st-order formula $\varphi$	General theory constraint $C$
Application handling	Curried, partial app (argument filter)	Application, variable-abstraction
Status (lex/mul)	Lexicographic (filter $\pi$)	Lex/multiset ($stat(f)$)
DP/Reduction pair	Weak monotonicity in strict component	Standard reduction pair framework
Theory integration	Explicit, by value variables and entailment	External entailment $C \models s >_A t$

Both frameworks contrast with the classical computability-closure based HORPO (0708.3582): the modern, syntax-driven definitions absorb the closure into the inductive structure, enabling direct decidability, quadratic complexity, and modular support for logical constraints.

6. Limitations and Practical Application

Limitations:

• The LCSTRS ordering treats only lexicographic status ($\pi$ filters) rather than full multiset status.
• Weak monotonicity of strict order can hinder orientation of context-sensitive or complex rules.
• Inference of precedence and argument filters generally requires heuristics or external (SMT) support.

Practical Use:

• Users compute precedence$\geq_F$ and argument filters$\pi$ heuristically.
• All rules and DPs are oriented using the constrained ordering, with theory side-conditions discharged via a background solver.
• If all orientation succeeds, DP-framework methods yield termination of the target LCSTRS or higher-order LCTRS (Kop, 2024, Guo et al., 2023).

7. Relation to Other Orderings and Theoretical Significance

The advanced constrained HORPOs offer:

• Direct handling of higher-order, constrained, and curried rewriting,
• Integration with dependency-pair and reduction-pair frameworks,
• Compatibility with strictly positive inductive types and recursors, facilitated by explicit abstraction-handling rules (in the case of (0708.3582)),
• Decidable, syntax-directed definitions with clearly characterized complexity and stability properties.

Their introduction broadens the applicability of syntactic termination orderings to modern rewriting systems encompassing theories, logical side-conditions, and higher-order program analysis, with clear metatheoretical soundness and practical effectiveness.