Initial commit: Build project architecture, define interfaces, write documentation, with the help of Claude

3 files changed, 211 insertions, 0 deletions

diff --git a/include/atp/strategy/calculus.h b/include/atp/strategy/calculus.h
new file mode 100644
index 0000000..ee8ee09
--- /dev/null
+++ b/include/atp/strategy/calculus.h
@@ -0,0 +1,68 @@
+/*
+ * richard-atp-engine: A saturation-based automated theorem prover for first-order logic.
+ * Copyright (C) 2026 Richard Qin
+ *
+ * This program is free software: you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation, either version 3 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <https://www.gnu.org/licenses/>.
+ */
+
+
+#pragma once
+
+/// @file calculus.h
+/// @brief A Calculus bundles the exact set of inference rules + heuristics
+///        to use for a specific problem class.
+///
+/// Think of it as the "algorithm profile" that the Prover loads.
+/// Different problems get different calculi:
+///   - Pure propositional:  resolution + factoring only
+///   - FOL without equality: + full unification
+///   - FOL with equality:   + paramodulation + KBO ordering
+///   - Arithmetic:          + theory solver integration
+///
+/// This is the single point where search space control happens.
+
+#include "atp/rewrite/ordering.h"
+#include "atp/search/clause_queue.h"
+#include "atp/strategy/inference_rule.h"
+
+#include <memory>
+#include <string_view>
+#include <vector>
+
+namespace atp {
+
+/// A Calculus = a complete specification of how to search for a proof.
+/// The Prover is parameterized by this — it never hardcodes which rules to use.
+struct Calculus {
+    /// Name for logging (e.g., "superposition", "pure_resolution").
+    std::string name;
+
+    /// Generating rules (executed in order during the main loop).
+    std::vector<std::unique_ptr<GeneratingRule>> generating_rules;
+
+    /// Simplifying rules (applied to each new clause before enqueuing).
+    std::vector<std::unique_ptr<SimplifyingRule>> simplifying_rules;
+
+    /// Clause selection heuristic.
+    ClauseComparator comparator = bfsCompare;
+
+    /// Term ordering (nullptr if equality reasoning is disabled).
+    std::unique_ptr<TermOrdering> ordering = nullptr;
+
+    /// Resource limits.
+    size_t max_clauses = 100000;
+    size_t max_iterations = 500000;
+};
+
+}  // namespace atp
diff --git a/include/atp/strategy/inference_rule.h b/include/atp/strategy/inference_rule.h
new file mode 100644
index 0000000..1c7dbba
--- /dev/null
+++ b/include/atp/strategy/inference_rule.h
@@ -0,0 +1,81 @@
+/*
+ * richard-atp-engine: A saturation-based automated theorem prover for first-order logic.
+ * Copyright (C) 2026 Richard Qin
+ *
+ * This program is free software: you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation, either version 3 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <https://www.gnu.org/licenses/>.
+ */
+
+
+#pragma once
+
+/// @file inference_rule.h
+/// @brief Abstract interface for pluggable inference rules.
+///
+/// Each inference rule (resolution, factoring, paramodulation, etc.)
+/// is a self-contained object that the Prover can selectively load.
+/// This prevents search space explosion by only enabling rules relevant
+/// to the problem at hand.
+
+#include "atp/core/clause.h"
+#include "atp/core/clause_store.h"
+#include "atp/core/term_bank.h"
+
+#include <string_view>
+#include <vector>
+
+namespace atp {
+
+/// Abstract base class for all inference rules.
+/// The Prover invokes only the rules present in its active calculus.
+class InferenceRuleBase {
+  public:
+    virtual ~InferenceRuleBase() = default;
+
+    /// Human-readable name for logging and proof output.
+    [[nodiscard]] virtual std::string_view name() const = 0;
+
+    /// Is this a generating rule (produces new clauses) or simplifying rule
+    /// (rewrites/deletes existing clauses)?
+    enum class Kind : uint8_t { kGenerating, kSimplifying };
+    [[nodiscard]] virtual Kind kind() const = 0;
+};
+
+/// Generating inference rule: given a "given clause" and the set of processed
+/// clauses, produce new clauses.
+class GeneratingRule : public InferenceRuleBase {
+  public:
+    [[nodiscard]] Kind kind() const final { return Kind::kGenerating; }
+
+    /// Generate new clauses from the given clause against all processed clauses.
+    /// @param given     The selected clause from the Unprocessed queue.
+    /// @param processed All previously processed clauses (accessed via store).
+    /// @param bank      The global term bank.
+    /// @return Newly generated clauses (not yet added to the store).
+    [[nodiscard]] virtual std::vector<Clause> generate(
+        const Clause& given, const std::vector<ClauseId>& processed,
+        const ClauseStore& store, TermBank& bank) const = 0;
+};
+
+/// Simplifying inference rule: rewrite or delete a clause in-place.
+class SimplifyingRule : public InferenceRuleBase {
+  public:
+    [[nodiscard]] Kind kind() const final { return Kind::kSimplifying; }
+
+    /// Attempt to simplify a clause. Returns true if the clause was modified.
+    /// If the clause should be deleted entirely, clear its literals.
+    virtual bool simplify(Clause& clause, const ClauseStore& store,
+                          TermBank& bank) const = 0;
+};
+
+}  // namespace atp
diff --git a/include/atp/strategy/problem_analyzer.h b/include/atp/strategy/problem_analyzer.h
new file mode 100644
index 0000000..b5fdef4
--- /dev/null
+++ b/include/atp/strategy/problem_analyzer.h
@@ -0,0 +1,62 @@
+/*
+ * richard-atp-engine: A saturation-based automated theorem prover for first-order logic.
+ * Copyright (C) 2026 Richard Qin
+ *
+ * This program is free software: you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation, either version 3 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <https://www.gnu.org/licenses/>.
+ */
+
+
+#pragma once
+
+/// @file problem_analyzer.h
+/// @brief Analyzes input clauses and selects the optimal Calculus.
+///
+/// Before the Prover starts its main loop, the ProblemAnalyzer inspects
+/// the input clause set to determine problem characteristics:
+///   - Contains equality?             → enable paramodulation
+///   - Pure Horn clauses?             → use SLD resolution
+///   - Contains arithmetic literals?  → enable theory solver
+///   - Problem size / clause density  → tune resource limits
+///
+/// This is the "auto mode" that mature ATPs use to avoid loading
+/// unnecessary inference rules that would explode the search space.
+
+#include "atp/core/clause.h"
+#include "atp/strategy/calculus.h"
+
+#include <vector>
+
+namespace atp {
+
+/// Features detected in the input problem.
+struct ProblemFeatures {
+    bool has_equality = false;        ///< Contains '=' predicate
+    bool is_purely_propositional = false;  ///< No function symbols, no variables
+    bool is_horn = false;             ///< All clauses are Horn clauses
+    bool has_existential = false;     ///< Had existential quantifiers (Skolemized)
+    size_t num_clauses = 0;
+    size_t num_predicates = 0;
+    size_t num_functions = 0;
+    size_t max_clause_size = 0;
+    size_t max_term_depth = 0;
+};
+
+/// Analyze the input clauses and extract problem features.
+ProblemFeatures analyzeFeatures(const std::vector<Clause>& clauses, const TermBank& bank);
+
+/// Select the optimal calculus based on detected features.
+/// This is the core strategy dispatch logic.
+std::unique_ptr<Calculus> selectCalculus(const ProblemFeatures& features);
+
+}  // namespace atp