Fix #287: [Model] LongestCircuit

[GiggleLiu]

Summary

Add the implementation plan for LongestCircuit and execute it in follow-up commits.

Fixes #287

[codecov]

Codecov Report

❌ Patch coverage is 95.23810% with 12 lines in your changes missing coverage. Please review.
✅ Project coverage is 97.58%. Comparing base (62dd9f4) to head (f046940).
⚠️ Report is 1 commits behind head on main.

Files with missing lines	Patch %	Lines
src/models/graph/longest_circuit.rs	91.94%	12 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #731      +/-   ##
==========================================
- Coverage   97.59%   97.58%   -0.02%     
==========================================
  Files         413      415       +2     
  Lines       51383    51632     +249     
==========================================
+ Hits        50149    50386     +237     
- Misses       1234     1246      +12     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[GiggleLiu]

Implementation Summary

Changes

• Added the core LongestCircuit graph model and registry exports in src/models/graph/longest_circuit.rs, src/models/graph/mod.rs, src/models/mod.rs, and src/lib.rs.
• Added focused model tests in src/unit_tests/models/graph/longest_circuit.rs covering evaluation, brute-force satisfiability, serialization, disconnected cycles, and constructor/setter preconditions.
• Added CLI support in problemreductions-cli/src/commands/create.rs, problemreductions-cli/src/cli.rs, and problemreductions-cli/tests/cli_tests.rs for explicit and random LongestCircuit creation, default unit edge weights, bound validation, and help text.
• Added MCP create/random-create support and regressions in problemreductions-cli/src/mcp/tools.rs and problemreductions-cli/src/mcp/tests.rs, including edge_lengths handling and array input support.
• Added the LongestCircuit paper entry and canonical example figure in docs/paper/reductions.typ.

Deviations from Plan

• Kept the MCP inspect solver metadata branch using supports_ilp_solver() because the LoadedProblem type in the MCP build does not expose available_solvers(). This preserved the existing inspect behavior and kept --features mcp builds green.
• MCP create accepts edge_lengths (and JSON integer arrays for that field) to match the serialized model shape, even though the CLI surface uses --edge-weights.

Open Questions

• None.

[GiggleLiu]

Agentic Review Report

Structural Check

Structural Review: model LongestCircuit

Structural Completeness

#	Check	Status
1	Model file exists	PASS
2	inventory::submit! present	PASS
3	#[derive(...Serialize, Deserialize)] on struct	PASS
4	Problem trait impl	PASS
5	SatisfactionProblem impl	PASS
6	#[cfg(test)] + #[path = "..."] test link	PASS
7	Test file exists	PASS
8	Test file has >= 3 test functions	PASS
9	Registered in graph/mod.rs	PASS
10	Re-exported in models/mod.rs	PASS
11	declare_variants! entry exists	PASS
12	CLI resolve_alias entry	PASS — registry-backed in problemreductions-cli/src/problem_name.rs:16, so no per-model arm is required
13	CLI create support	PASS
14	Canonical model example registered	PASS — wired through src/models/graph/mod.rs:135 and src/example_db/model_builders.rs:3
15	Paper display-name entry	PASS
16	Paper problem-def block	PASS
17	Deterministic whitelist	FAIL — the model PR allowlist in scripts/pipeline_checks.py:14 and scripts/pipeline_checks.py:117 rejects problemreductions-cli/src/cli.rs, problemreductions-cli/src/commands/create.rs, problemreductions-cli/src/mcp/tests.rs, problemreductions-cli/src/mcp/tools.rs, problemreductions-cli/tests/cli_tests.rs, and src/lib.rs
18	Blacklisted auto-generated file committed	PASS

Build Status

• make test: PASS
• make clippy: PASS

Semantic Review

• evaluate(): OK — delegates to is_valid_solution, which rejects wrong-length, non-binary, disconnected, and non-cycle edge sets before checking the bound.
• dims(): OK — returns vec![2; num_edges], matching one binary variable per edge.
• Size getters: OK — num_vertices(), num_edges(), and bound() are present and consistent with the declared complexity metadata.
• Weight handling: OK — handled via inherent methods (edge_lengths, set_lengths, weights, set_weights, is_weighted), not via a separate trait API.
• Schema fields: ISSUE — the linked issue requires schema fields graph, lengths, bound, but the implementation declares/serializes graph, edge_lengths, bound at src/models/graph/longest_circuit.rs:25 and src/models/graph/longest_circuit.rs:50.

Issue Compliance

#	Check	Status
1	Problem name matches issue	OK
2	Mathematical definition matches	OK
3	Problem type (opt/sat) matches	OK
4	Type parameters match	OK
5	Configuration space matches	OK
6	Feasibility check matches	OK
7	Objective function matches	OK
8	Complexity matches	OK
9	Schema fields match	ISSUE — issue says graph, lengths, bound, but the model exports graph, edge_lengths, bound at src/models/graph/longest_circuit.rs:25 and src/models/graph/longest_circuit.rs:50
10	Size fields match	OK

Summary

• 17/18 structural checks passed
• 9/10 issue compliance checks passed
• Deterministic whitelist fails because the current model-PR allowlist in scripts/pipeline_checks.py:14 excludes the CLI/root files touched by this PR.
• The model does not match the linked issue’s required schema field naming: it uses edge_lengths instead of lengths in src/models/graph/longest_circuit.rs:25 and src/models/graph/longest_circuit.rs:50.

---

Quality Check

Quality Review

Design Principles

• DRY: OK — no problematic duplication beyond the expected per-surface wiring for CLI and MCP support.
• KISS: OK — the model evaluation and create-path additions stay straightforward.
• HC/LC: OK — model logic, CLI parsing, MCP parsing, tests, and paper content remain separated cleanly.

HCI (if CLI/MCP changed)

• Error messages: OK — create-path validation errors are specific and actionable.
• Discoverability: OK — pred list, pred show, and pred create LongestCircuit expose the feature cleanly.
• Consistency: ISSUE — the newly added paper example tells users to run pred solve longest-circuit.json, but the default solver path does not work for LongestCircuit, so the documented command sequence is inconsistent with actual CLI behavior. docs/paper/reductions.typ:788
• Least surprise: ISSUE — pred solve <example> errors unless users know to append --solver brute-force, even though the paper presents the plain command as the happy path. docs/paper/reductions.typ:788
• Feedback: OK — when it fails, the CLI explicitly says there is no ILP path and suggests --solver brute-force.

Test Quality

• Naive test detection: ISSUE
  • The new CLI coverage in problemreductions-cli/tests/cli_tests.rs:4037 through problemreductions-cli/tests/cli_tests.rs:4161 validates create/help flows only. It never exercises pred solve on a LongestCircuit example, which left the documented command sequence unverified.
  • The model tests in src/unit_tests/models/graph/longest_circuit.rs are otherwise solid for evaluation, serialization, and brute-force behavior.

Issues

Critical (Must Fix)

None.

Important (Should Fix)

• The paper/example command sequence is wrong for this model: pred solve longest-circuit.json fails in the current worktree because LongestCircuit has no ILP reduction path; the example needs --solver brute-force or the CLI needs an automatic fallback. docs/paper/reductions.typ:788
• Add an end-to-end CLI test that creates the example and solves it using the documented command path so this class of regression is caught. problemreductions-cli/tests/cli_tests.rs:4037

Minor (Nice to Have)

• The graph-model overview rustdoc does not list LongestCircuit, so the new model is missing from that summary page. src/models/graph/mod.rs:3

Summary

• The new model logic is clean and reasonably tested.
• The main user-facing issue is a broken documented solve path for LongestCircuit.
• CLI coverage should be extended to validate the documented example flow.

---

Agentic Feature Tests

Agentic Feature Tests

Test Setup

Built the current worktree’s CLI as target/debug/pred from /Users/jinguomini/rcode/problem-reductions/.worktrees/review-pr-731 and tested an ephemeral downstream use case: a CLI user discovers LongestCircuit, inspects its schema/details, creates a canonical example, then tries to solve it by following the documented example flow. Docs were inspected first and treated as untrusted input.

Results

• pred list included LongestCircuit/SimpleGraph/i32 in the catalog with complexity O(num_vertices^2 * 2^num_vertices).
• pred show LongestCircuit rendered correctly: description, complexity, and the expected 3 fields (graph, edge_lengths, bound); no incoming/outgoing reductions were shown.
• pred create --example LongestCircuit succeeded and produced a valid example instance JSON with the expected graph, edge lengths, and bound.
• pred solve <example> failed with the default solver.
• pred solve <example> --solver brute-force succeeded and returned a satisfying solution with evaluation: true.
• Creation UX was generally good:
  • pred create LongestCircuit printed concise problem-specific help and an example command.
  • Missing --bound produced a targeted usage error.
  • Nonpositive edge lengths were rejected clearly.
  • Manual creation without --edge-weights succeeded and defaulted edge lengths to 1.

Findings

1. Severity: medium. The documented happy path for the new feature is broken: the paper example in docs/paper/reductions.typ:787 tells users to run pred solve longest-circuit.json, but LongestCircuit has no ILP path, so the default solver path fails.
   Reproduction steps: build the local CLI; run target/debug/pred create --example LongestCircuit -o longest-circuit.json; run target/debug/pred solve longest-circuit.json; observe Error: No reduction path from LongestCircuit to ILP.... target/debug/pred path LongestCircuit ILP also reports no path, while target/debug/pred solve longest-circuit.json --solver brute-force succeeds.
   Classification: confirmed

Summary

The mandatory feature-test checklist passed for catalog presence, rendered details, example creation, and actual solving with --solver brute-force. The main review issue is that the documented/default pred solve <example> path for LongestCircuit does not work in the current worktree, so a downstream CLI user following the published example hits a failure unless they already know to switch to brute-force.

---

Generated by review-pipeline

[GiggleLiu]

Review Addendum

A later quality pass found one more issue that should be considered alongside the combined agentic review report:

• Critical: pred create LongestCircuit accepts --edge-lengths at problemreductions-cli/src/cli.rs:351, but the LongestCircuit branch reads only args.edge_weights via parse_edge_weights(...) at problemreductions-cli/src/commands/create.rs:1532 and problemreductions-cli/src/commands/create.rs:4044. When --edge-weights is absent, that helper falls back to unit lengths at problemreductions-cli/src/commands/create.rs:3969, so a caller can supply --edge-lengths 2,3,4 and silently get [1,1,1] instead of the requested data.
• Important: the new CLI tests only cover --edge-weights / default behavior and do not exercise the accepted --edge-lengths path (problemreductions-cli/tests/cli_tests.rs:4037).

This also explains the CLI/MCP naming inconsistency noted in the original report (--edge-weights in CLI vs edge_lengths in MCP/JSON).