Fix #52: TravelingSalesman to ILP reduction (#60)

* update

* Add plan for #52: TravelingSalesman to ILP reduction

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* save makefile and .claude

* fix reductions macro missing

* feat: add TravelingSalesman to ILP reduction (#52)

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* test: add unit tests for TravelingSalesman to ILP reduction (#52)

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* docs: add TravelingSalesman to ILP reduction rule in paper (#52)

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* feat: add TravelingSalesman to ILP example (#52)

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* chore: regenerate reduction graph and fix clippy in example (#52)

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* update issue-to-pr

* fix: resolve PR review comments for TSP-to-ILP reduction

- Use exact constraint count formula (n³ - n² + 2n + 4mn) instead of approximation
- Fix missing "is" in introduction.md
- Categorize TravelingSalesman as "graph" with correct doc_path

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>

Author: GiggleLiu

.claude/CLAUDE.md

@@ -3,6 +3,9 @@
 ## Project Overview
 Rust library for NP-hard problem reductions. Implements computational problems with reduction rules for transforming between equivalent formulations.
 
+## Skills
+When resolving an issue with pull request, please use [issue-to-pr](skills/issue-to-pr.md) skill to convert the issue into a pull request with a plan.
+
 ## Commands
 ```bash
 make help           # Show all available targets

.claude/skills/issue-to-pr.md

@@ -7,12 +7,6 @@ description: Use when you have a GitHub issue and want to create a PR with an im
 
 Convert a GitHub issue into an actionable PR with a plan that auto-triggers Claude execution.
 
-## Usage
-
-```
-/issue-to-pr <issue-number-or-url>
-```
-
 ## Workflow
 
 ```dot
@@ -24,15 +18,13 @@ digraph issue_to_pr {
     "Research references" [shape=box];
     "Write plan file" [shape=box];
     "Create branch and PR" [shape=box];
-    "PR triggers [action]" [shape=doublecircle];
 
     "Receive issue number" -> "Fetch issue with gh";
     "Fetch issue with gh" -> "Check the rules to follow";
     "Check the rules to follow" -> "Verify completeness";
     "Verify completeness" -> "Research references";
     "Research references" -> "Write plan file";
     "Write plan file" -> "Create branch and PR";
-    "Create branch and PR" -> "PR triggers [action]";
 }
 ```
 
@@ -108,6 +100,7 @@ The plan MUST include an **action pipeline** section with concrete steps based o
    - Present example in tutorial style (see KColoring→QUBO section for reference)
 
 5. **Regenerate graph** — `cargo run --example export_graph`
+6. **Push and create PR** — Push the changes and create a pull request with a description of the changes.
 
 **Rules for solver implementation:**
 - Make sure at least one solver is provided in the issue template. Check if the solving strategy is valid. If not, reply under issue to ask for clarification.
@@ -140,8 +133,10 @@ The plan MUST include an **action pipeline** section with concrete steps based o
 3. **Document** — Update `docs/paper/reductions.typ`:
    - Add `display-name` entry
    - Add `#problem-def("Name")[definition...]`
+4. **Push and create PR** — Push the changes and create a pull request with a description of the changes.
 
 ### 6. Create PR
+Create a pull request with only the plan file.
 
 ```bash
 # Create branch
@@ -156,16 +151,15 @@ git commit -m "Add plan for #<number>: <title>"
 # Push
 git push -u origin issue-<number>-<slug>
 
-# Create PR with [action] at the BEGINNING
-gh pr create --title "Fix #<number>: <title>" --body "[action]
+# Create PR
+gh pr create --title "Fix #<number>: <title>" --body "
 
 ## Summary
 <Brief description from brainstorming>
 
 Closes #<number>"
 ```
 
-**CRITICAL:** The PR body MUST start with `[action]` on the first line. This triggers automated plan execution.
 
 ## Example
 
@@ -181,17 +175,15 @@ All required info is present. I'll create the plan...
 
 [Writes docs/plans/2026-02-09-independentset-to-qubo.md]
 [Creates branch, commits, pushes]
-[Creates PR with body starting with "[action]"]
+[Creates PR]
 
-Created PR #45: Fix #42: Add IndependentSet → QUBO reduction
-The [action] trigger will automatically execute the plan.
+Created PR #45: Fix #42: Add IndependentSet → QUBO reduction, description: ...
 ```
 
 ## Common Mistakes
 
 | Mistake | Fix |
 |---------|-----|
 | Issue template incomplete | Ask contributor to fill in missing sections before proceeding |
-| `[action]` not at start | PR body must BEGIN with `[action]` |
 | Including implementation code in initial PR | First PR: plan only |
 | Generic plan | Use specifics from the issue |

.gitignore

@@ -77,3 +77,4 @@ docs/paper/examples/
 
 # Claude Code logs
 claude-output.log
+.worktrees/

Makefile

@@ -165,9 +165,9 @@ compare: rust-export
 INSTRUCTIONS ?=
 OUTPUT ?= claude-output.log
 AGENT_TYPE ?= claude
+PLAN_FILE ?= $(shell ls -t docs/plans/*.md 2>/dev/null | head -1)
 
 run-plan:
-	PLAN_FILE ?= $(shell ls -t docs/plans/*.md 2>/dev/null | head -1)
 	@NL=$$'\n'; \
 	BRANCH=$$(git branch --show-current); \
 	if [ "$(AGENT_TYPE)" = "claude" ]; then \
@@ -182,6 +182,7 @@ run-plan:
 	PROMPT="$${PROMPT}$${NL}$${NL}## Process$${NL}$${PROCESS}$${NL}$${NL}## Rules$${NL}- Tests should be strong enough to catch regressions.$${NL}- Do not modify tests to make them pass.$${NL}- Test failure must be reported."; \
 	echo "=== Prompt ===" && echo "$$PROMPT" && echo "===" ; \
 	claude --dangerously-skip-permissions \
-		--model claude-opus-4-6 \
+		--model opus \
+		--verbose \
 		--max-turns 500 \
 		-p "$$PROMPT" 2>&1 | tee "$(OUTPUT)"

docs/paper/reductions.typ

@@ -806,6 +806,25 @@ The following reductions to Integer Linear Programming are straightforward formu
   _Construction._ Variables: $x_v in {0, 1}$ for each $v in V$. Constraints: $x_u + x_v <= 1$ for each $(u, v) in.not E$ (non-edges). Objective: maximize $sum_v x_v$. Equivalently, IS on the complement graph. _Solution extraction:_ $K = {v : x_v = 1}$.
 ]
 
+#reduction-rule("TravelingSalesman", "ILP",
+  example: true,
+  example-caption: [Weighted $K_4$: the optimal tour $0 arrow 1 arrow 3 arrow 2 arrow 0$ with cost 80 is found by position-based ILP.],
+)[
+  The traveling salesman problem reduces to binary ILP with $n^2 + 2 m n$ variables via position-based encoding with McCormick linearization.
+][
+  _Construction._ For graph $G = (V, E)$ with $n = |V|$ and $m = |E|$:
+
+  _Variables:_ Binary $x_(v,k) in {0, 1}$ for each vertex $v in V$ and position $k in {0, ..., n-1}$. Interpretation: $x_(v,k) = 1$ iff vertex $v$ is at position $k$ in the tour.
+
+  _Auxiliary variables:_ For each edge $(u,v) in E$ and position $k$, introduce $y_(u,v,k)$ and $y_(v,u,k)$ to linearize the products $x_(u,k) dot x_(v,(k+1) mod n)$ and $x_(v,k) dot x_(u,(k+1) mod n)$ respectively.
+
+  _Constraints:_ (1) Each vertex has exactly one position: $sum_(k=0)^(n-1) x_(v,k) = 1$ for all $v in V$. (2) Each position has exactly one vertex: $sum_(v in V) x_(v,k) = 1$ for all $k$. (3) Non-edge consecutive prohibition: if ${v,w} in.not E$, then $x_(v,k) + x_(w,(k+1) mod n) <= 1$ for all $k$. (4) McCormick: $y <= x_(v,k)$, $y <= x_(w,(k+1) mod n)$, $y >= x_(v,k) + x_(w,(k+1) mod n) - 1$.
+
+  _Objective:_ Minimize $sum_((u,v) in E) w(u,v) dot sum_k (y_(u,v,k) + y_(v,u,k))$.
+
+  _Solution extraction._ For each position $k$, find vertex $v$ with $x_(v,k) = 1$ to recover the tour permutation; then select edges between consecutive positions.
+]
+
 == Unit Disk Mapping
 
 #reduction-rule("MaximumIndependentSet", "GridGraph")[
@@ -918,6 +937,7 @@ The following table shows concrete variable overhead for example instances, gene
   "kcoloring_to_ilp", "factoring_to_ilp",
   "maximumsetpacking_to_ilp", "minimumsetcovering_to_ilp",
   "minimumdominatingset_to_ilp", "maximumclique_to_ilp",
+  "travelingsalesman_to_ilp",
 )
 
 #let examples = example-files.map(n => {