Codecov Report

❌ Patch coverage is 95.37954% with 14 lines in your changes missing coverage. Please review.
✅ Project coverage is 97.58%. Comparing base (a01930d) to head (e938f2d).
⚠️ Report is 1 commits behind head on main.

@@            Coverage Diff             @@
##             main     #730      +/-   ##
==========================================
- Coverage   97.60%   97.58%   -0.03%     
==========================================
  Files         409      411       +2     
  Lines       50893    51196     +303     
==========================================
+ Hits        49676    49958     +282     
- Misses       1217     1238      +21     

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

Implementation Summary

Changes

• Added the new StackerCrane model in src/models/misc/stacker_crane.rs, including validation, evaluation via shortest connector paths in the mixed graph, registry metadata, and a canonical example.
• Wired StackerCrane into the library exports and example database, with unit tests covering construction, serialization, the issue witness, and the global B = 19 unsatisfiability claim.
• Extended pred create support for StackerCrane, including examples, validation tests, and schema-driven help coverage for the documented --arcs, --graph, --arc-costs, --edge-lengths, and --bound flags.
• Added a paper entry and bibliography references for Stacker Crane in docs/paper/reductions.typ and docs/paper/references.bib.

Deviations from Plan

• During verification, the schema-driven pred create StackerCrane help path was still using generic flag names (--biedges, --arc-lengths, --edge-weights). I fixed that by adding StackerCrane-specific help-flag overrides and an integration test for the real CLI help path.
• I also strengthened the test coverage beyond the original witness check by brute-forcing the issue instance at B = 19 to confirm it is globally unsatisfiable, as stated in issue [Model] MinimumStackerCrane #245.

Open Questions

• No associated rule issue exists yet, so this model still enters the catalog as an orphan until a future reduction is added.

Agentic Review Report

Structural Check

Structural Review: model StackerCrane

Structural Completeness

Build Status

• make test: PASS
• make clippy: PASS

Semantic Review

• evaluate(): OK — the implemented decision model is internally coherent: non-permutations, unreachable connectors, and overflow are rejected before the bound check in src/models/misc/stacker_crane.rs:227.
• dims(): OK for the implemented encoding — [num_arcs; num_arcs] enumerates arc-order candidates and evaluate() filters to valid permutations in src/models/misc/stacker_crane.rs:268.
• Edge cases: ISSUE — try_new() allows num_vertices == 0, and then dims() is empty while closed_walk_length([]) returns Some(0), so the empty mixed graph is accepted as satisfiable even though there is no start vertex for a closed walk in src/models/misc/stacker_crane.rs:76 and src/config.rs:44.
• Size getter consistency: OK — num_vertices, num_arcs, and num_edges align with the declared complexity fields in src/models/misc/stacker_crane.rs:121 and src/models/misc/stacker_crane.rs:279.
• Brute-force assumptions: OK for the current encoding — brute force is sound over arc-order tuples, but it is a different search space from the linked issue’s walk-sequence model.
• Complexity metadata: ISSUE — the model declares num_vertices * 2^num_arcs in src/models/misc/stacker_crane.rs:279, which diverges from the linked issue’s num_vertices^2 * 2^num_arcs.

Issue Compliance

Summary

• 18/19 structural checks passed.
• 1/8 issue compliance checks passed.
• FAIL — supplied review packet reports deterministic whitelist failure.
• ISSUE — the PR is internally consistent as the decision problem StackerCrane, but it does not comply with the saved linked-issue spec for MinimumStackerCrane.
• ISSUE — empty mixed-graph instances with num_vertices == 0 are accepted as satisfiable.
• ISSUE — complexity metadata does not match the linked issue and is not justified by an implemented exact algorithm.

Quality Check

Quality Review

Design Principles

• DRY: ISSUE — the StackerCrane CLI contract is duplicated across the after-help table, example text, usage text, and parser defaults, and those copies already disagree about whether --arc-costs / --edge-lengths are required. See problemreductions-cli/src/cli.rs:269, problemreductions-cli/src/commands/create.rs:590, problemreductions-cli/src/commands/create.rs:1508, problemreductions-cli/src/commands/create.rs:3947, and problemreductions-cli/src/commands/create.rs:4836.
• KISS: ISSUE — the create path composes parse_directed_graph and parse_graph, then rejects mismatched inferred vertex counts in problemreductions-cli/src/commands/create.rs:1513. A single mixed-graph vertex-count resolution would be simpler and would avoid the current valid-input failure mode.
• HC/LC: ISSUE — the StackerCrane branch is tightly coupled to the inference behavior of two unrelated helpers instead of owning one mixed-graph input contract. That coupling is what makes otherwise valid inputs fail in problemreductions-cli/src/commands/create.rs:1520.

HCI

• Error messages: ISSUE — the failure in problemreductions-cli/src/commands/create.rs:1520 says the directed and undirected inputs must “agree on --num-vertices”, but the real problem is that the CLI does not infer a combined vertex count from both sides.
• Discoverability: ISSUE — the help line marks --num-vertices as optional in problemreductions-cli/src/cli.rs:269, but omission only works when arcs and edges happen to mention the same maximum vertex index.
• Consistency: ISSUE — help/usage present --arc-costs and --edge-lengths as required in problemreductions-cli/src/cli.rs:269 and problemreductions-cli/src/commands/create.rs:1508, while the implementation silently defaults them in problemreductions-cli/src/commands/create.rs:3947 and problemreductions-cli/src/commands/create.rs:4836.
• Least surprise: ISSUE — pred create StackerCrane --arcs "5>0" --graph "0-1" --arc-costs 1 --edge-lengths 1 --bound 3 fails today unless --num-vertices 6 is added, even though the mixed input already determines that size.
• Feedback: OK — successful creation emits the expected JSON shape, as covered in problemreductions-cli/src/commands/create.rs:5994.

Test Quality

• Naive test detection: ISSUE.
• problemreductions-cli/tests/cli_tests.rs:136 only checks that help text contains flag names; it does not verify actual optional/required semantics, so the --arc-costs / --edge-lengths drift is invisible to tests.
• problemreductions-cli/src/commands/create.rs:5994 covers one symmetric happy path and two parse errors, but it does not cover the valid asymmetric inference path where arcs and edges touch different highest-numbered vertices, which is where the new create logic breaks.
• src/unit_tests/models/misc/stacker_crane.rs:27 reuses one canonical fixture for most assertions and never exercises constructor rejection or unreachable-connector branches in src/models/misc/stacker_crane.rs:76 and src/models/misc/stacker_crane.rs:192.

Issues

Critical (Must Fix)

• None.

Important (Should Fix)

• Valid StackerCrane instances are rejected unless users manually provide --num-vertices whenever the arc list and edge list infer different maxima. The bug is in the interaction between problemreductions-cli/src/commands/create.rs:1513, problemreductions-cli/src/commands/create.rs:3854, and problemreductions-cli/src/commands/create.rs:4772.
• The added CLI tests miss that failing inference path, so the new user-facing bug slips through despite dedicated coverage. See problemreductions-cli/tests/cli_tests.rs:136 and problemreductions-cli/src/commands/create.rs:5994.

Minor (Nice to Have)

• The help contract is internally inconsistent: problemreductions-cli/src/cli.rs:269 and problemreductions-cli/src/commands/create.rs:1508 imply required length flags, while problemreductions-cli/src/commands/create.rs:3947 and problemreductions-cli/src/commands/create.rs:4836 make them optional.

Summary

• Important: pred create StackerCrane rejects valid mixed-graph inputs when arcs and edges infer different vertex counts unless --num-vertices is added manually.
• Important: the new tests do not cover that inference path, so the main CLI bug is not being exercised.
• Minor: the StackerCrane help/usage text disagrees with the implementation about whether --arc-costs and --edge-lengths are required.

Agentic Feature Tests

Verdict

partial fail for the requested end-to-end CLI path. pred list, pred show StackerCrane, and pred create --example StackerCrane work. Solving the shipped example works only with --solver brute-force; plain pred solve <instance> fails.

Test Results

• cargo build -p problemreductions-cli --bin pred: passed.
• ./target/debug/pred list: passed. Catalog includes StackerCrane * with complexity O(num_vertices * 2^num_arcs).
• ./target/debug/pred show StackerCrane: partial pass. Description, complexity, fields, and zero incoming/outgoing reductions display correctly.
• ./target/debug/pred create --example StackerCrane -o /tmp/.../stacker-crane.json: passed. Output JSON has "type": "StackerCrane" and the shipped canonical 6-vertex example.
• ./target/debug/pred solve /tmp/.../stacker-crane.json: failed. Error: No reduction path from StackerCrane to ILP or ILP solver found no solution. Try --solver brute-force.
• ./target/debug/pred solve /tmp/.../stacker-crane.json --solver brute-force: passed. Returned solution [0,2,1,4,3] with evaluation true.
• ./target/debug/pred evaluate /tmp/.../stacker-crane.json --config 0,2,1,4,3: passed. Returned true.
• Naming-friction check: ./target/debug/pred show MinimumStackerCrane and ./target/debug/pred create --example MinimumStackerCrane both fail with Unknown problem, while the linked issue still uses MinimumStackerCrane.

Findings

1. Default pred solve <instance> does not work for the shipped StackerCrane example.
   Status: confirmed
   Severity: medium
   Reproduction: create the example, then run plain pred solve on it; it fails because StackerCrane has no ILP reduction path. The same instance succeeds with --solver brute-force.
   Recommended fix: either make plain pred solve fall back to brute-force when ILP is unavailable but brute-force is supported, or surface the brute-force requirement in the primary discovery path for this model so the documented happy path matches actual behavior.

2. pred show StackerCrane omits size-field metadata.
   Status: confirmed
   Severity: low
   Reproduction: pred show StackerCrane --json returns "size_fields": [], and human-readable pred show StackerCrane shows no Size fields section.
   Recommended fix: register size-field metadata in src/models/misc/stacker_crane.rs for at least num_vertices and num_arcs, and likely num_edges, so the CLI can display them.

3. Linked-issue naming still points users to MinimumStackerCrane, which the CLI does not recognize.
   Status: confirmed
   Severity: low
   Reproduction: pred show MinimumStackerCrane and pred create --example MinimumStackerCrane both return Unknown problem.
   Recommended fix: update the linked issue title/metadata and generated review context to use StackerCrane, matching the shipped model name and the CLI catalog.

Generated by review-pipeline

Follow-up items (recorded during final review):

• Update issue [Rule] HamiltonianCircuit to MinimumStackerCrane #261 title from "HamiltonianCircuit to MinimumStackerCrane" to "HamiltonianCircuit to StackerCrane" to match the shipped model name
• Update issue [Model] MinimumStackerCrane #245 title from "[Model] MinimumStackerCrane" to "[Model] StackerCrane" for consistency