Implementation Summary

Changes

• Added the MinimumDummyActivitiesPert graph model with DAG validation, brute-force dimensions 2^num_arcs, and evaluation logic that merges selected predecessor-finish / successor-start event pairs, rejects cyclic or reachability-changing encodings, and minimizes the number of distinct dummy activity arcs.
• Added dedicated unit coverage for creation, invalid cyclic input, the corrected issue example with optimum 2, an invalid spurious-reachability merge, brute-force optimality, serialization, and the paper example path.
• Registered the model in the graph module exports, crate prelude, and example-db coverage so it appears in schema/example exports.
• Extended pred create and CLI help with MinimumDummyActivitiesPert parsing and example support.
• Added a paper problem-def entry with the canonical example and figure.

Deviations from Plan

• Used a direct-precedence merge encoding with one binary choice per input arc rather than a larger event-assignment search space. This keeps the implementation aligned with the issue's intended 2^num_arcs brute-force complexity and makes the canonical example exhaustively solvable.

Open Questions

• None.

Codecov Report

❌ Patch coverage is 99.60630% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 97.63%. Comparing base (adfe9ae) to head (69e123c).
⚠️ Report is 1 commits behind head on main.

@@           Coverage Diff            @@
##             main     #751    +/-   ##
========================================
  Coverage   97.62%   97.63%            
========================================
  Files         437      439     +2     
  Lines       53997    54251   +254     
========================================
+ Hits        52715    52968   +253     
- Misses       1282     1283     +1     

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

Agentic Review Report

Structural Check

Structural Review: model MinimumDummyActivitiesPert

Structural Completeness

Deterministic Checks

• Whitelist: FAIL — the packet flagged files outside the current model whitelist (problemreductions-cli/src/cli.rs, problemreductions-cli/src/commands/create.rs, src/lib.rs, src/models/graph/mod.rs, src/models/mod.rs, src/unit_tests/example_db.rs). These changes look related to registration/CLI/docs rather than autogenerated outputs, but the whitelist still reports them.
• Completeness: PASS
• Blacklisted auto-generated files: PASS — none of the forbidden generated files are in the diff.

Build Status

• make test: PASS
• make clippy: PASS

Semantic Review

• evaluate() correctness: ISSUE — src/models/graph/minimum_dummy_activities_pert.rs:131 unconditionally materializes one dummy arc for every precedence arc whose config bit is 0. That makes the model overcount DAGs with transitive arcs. Reproduction in the checked-out branch: create MinimumDummyActivitiesPert on arcs 0>1,1>2,0>2, then run pred solve ... --solver brute-force; the solver returns Valid(1) with solution [1,1,0]. Under the issue definition, zero dummy arcs suffice by merging 0+=1- and 1+=2-, because 0+ already reaches 2- through task 1.
• dims() correctness: OK — binary choice per input arc.
• Size getter consistency: OK — num_arcs() matches the declared complexity string.
• Weight handling: OK — not applicable for this unweighted directed-graph model.

Issue Compliance (linked issue #301)

Summary

• 16/16 structural completeness checks passed.
• 6/8 issue-compliance checks passed.
• ISSUE — the evaluator overcounts dummy arcs on DAGs with transitive precedence edges, so the model does not implement the full problem from issue [Model] MinimumDummyActivitiesPert #301.
• FAIL (deterministic) — file whitelist check reported extra touched files outside the current model whitelist.

Quality Check

Quality Review

Design Principles

• DRY: OK — the model/CLI/doc wiring follows existing patterns without obvious copy-paste debt.
• KISS: ISSUE — the simplified one-bit-per-direct-precedence encoding is easy to brute-force, but it changes the optimization problem’s semantics; the simplification is not behavior-preserving.
• HC/LC: OK — model logic, CLI support, tests, and paper docs remain separated by module.

HCI (CLI changed)

• Error messages: OK — cyclic-input rejection is explicit and actionable.
• Discoverability: ISSUE — the paper/example flow tells users to run pred solve minimum-dummy-activities-pert.json, but that command fails for this model unless the user already knows to add --solver brute-force.
• Consistency: OK — pred create help and examples are wired in the usual way.
• Least surprise: ISSUE — pred create --example MinimumDummyActivitiesPert followed by the documented solve command fails with No reduction path from MinimumDummyActivitiesPert to ILP... Try --solver brute-force.
• Feedback: OK — the CLI clearly reports the failure reason and the workaround.

Test Quality

• Naive test detection: ISSUE
  • src/unit_tests/models/graph/minimum_dummy_activities_pert.rs:49 through src/unit_tests/models/graph/minimum_dummy_activities_pert.rs:87 cover the canonical example, cyclic input, and one spurious merge, but they never exercise a DAG with transitive arcs. That omission lets the objective-function bug above pass unnoticed.

Issues

Critical (Must Fix)

• src/models/graph/minimum_dummy_activities_pert.rs:131: non-merged direct precedence arcs are always counted as dummy arcs, even when the precedence is already implied transitively in the event network. This makes the solver return suboptimal values on valid DAG inputs such as 0>1,1>2,0>2.

Important (Should Fix)

• docs/paper/reductions.typ:1867: the documented pred solve minimum-dummy-activities-pert.json command fails on this PR because the model has no ILP reduction path; the example needs --solver brute-force.
• src/unit_tests/models/graph/minimum_dummy_activities_pert.rs:49: missing regression coverage for transitive precedence arcs.

Minor (Nice to Have)

• None beyond the deterministic whitelist noise already noted above.

Summary

• ISSUE — the core search space is too restrictive for the mathematical problem, so the model can return a non-optimal dummy count.
• ISSUE — the published example solve command is not runnable as written.
• ISSUE — tests miss the transitive-arc case that exposes the correctness bug.

Agentic Feature Tests

Feature Test Report: problem-reductions

Date: 2026-03-22
Project type: CLI + Rust library
Features tested: MinimumDummyActivitiesPert model
Profile: ephemeral
Use Case: Discover the new model from the CLI, create the canonical example, and solve it from docs/examples alone.
Expected Outcome: The model appears in the catalog, pred show explains it, pred create --example writes a valid instance, and the documented solve flow works end-to-end.
Verdict: fail
Critical Issues: 1

Summary

Per-Feature Details

MinimumDummyActivitiesPert

• Role: downstream CLI user trying the newly added model.
• Use Case: find the model in the catalog, inspect it, create the example instance, and solve it without reading the Rust source.
• What they tried: pred list, pred show MinimumDummyActivitiesPert, pred create --example MinimumDummyActivitiesPert, pred solve minimum-dummy-activities-pert-example.json, pred solve ... --solver brute-force, and pred evaluate ... --config 1,0,0,1,1.
• Discoverability: good. The model appears in pred list and pred show prints the expected description and field list.
• Setup: good. No extra setup was required inside the PR worktree.
• Functionality: partial. Example creation and evaluation work. Default pred solve fails with No reduction path from MinimumDummyActivitiesPert to ILP or ILP solver found no solution. Try --solver brute-force. Adding --solver brute-force succeeds on the canonical example.
• Expected vs Actual Outcome: not met. The paper/example flow implies pred solve should work directly, but a real user must discover --solver brute-force on their own.
• Blocked steps: none.
• Friction points: the documented solve command is wrong for this model.
• Doc suggestions: update the paper/example command to pred solve minimum-dummy-activities-pert.json --solver brute-force.

Expected vs Actual Outcome

• Catalog and show output matched expectations.
• Example creation matched expectations.
• Solving did not match expectations because the documented command fails without an explicit solver selection.

Issues Found

• Confirmed critical: on a custom transitive DAG instance (0>1,1>2,0>2), pred solve ... --solver brute-force returns Valid(1) even though zero dummy arcs suffice under the issue’s reachability-only definition.
• Confirmed important: the paper’s documented pred solve minimum-dummy-activities-pert.json command fails unless the user adds --solver brute-force.

Suggestions

• Fix the evaluator/search space so transitive precedence arcs can be satisfied without forcing an extra dummy arc.
• Add a regression test for a DAG with redundant transitive precedence edges.
• Correct the paper/example solve command to include --solver brute-force.

Generated by review-pipeline