polyglot-ast-slicer 1.0.0__tar.gz

polyglot_ast_slicer-1.0.0/.githooks/pre-commit

@@ -0,0 +1,3 @@
+#!/bin/sh
+set -e
+exec make gate

polyglot_ast_slicer-1.0.0/.github/ISSUE_TEMPLATE/bug_report.yml

@@ -0,0 +1,28 @@
+name: Bug report
+description: Wrong edge, missed edge, crash, or bad slice
+labels: [bug]
+body:
+  - type: textarea
+    id: what
+    attributes:
+      label: What happened
+      description: Include the exact command and the raw output.
+    validations: { required: true }
+  - type: input
+    id: version
+    attributes:
+      label: Version
+      description: Output of `polyglot-ast-slicer --version`
+    validations: { required: true }
+  - type: input
+    id: repo
+    attributes:
+      label: Repository
+      description: Public URL and commit SHA, or a minimal reproduction.
+    validations: { required: true }
+  - type: textarea
+    id: expected
+    attributes:
+      label: Expected vs actual edge
+      description: For graph bugs, include source line, target line and `explain_edge` output.
+    validations: { required: true }

polyglot_ast_slicer-1.0.0/.github/ISSUE_TEMPLATE/feature_request.yml

@@ -0,0 +1,22 @@
+name: Feature request
+description: Propose a capability
+labels: [enhancement]
+body:
+  - type: textarea
+    id: problem
+    attributes:
+      label: Problem
+      description: What can you not do today? Include a concrete repository example.
+    validations: { required: true }
+  - type: textarea
+    id: proposal
+    attributes:
+      label: Proposal
+      description: What should the tool return, and how would we measure success?
+    validations: { required: true }
+  - type: dropdown
+    id: scope
+    attributes:
+      label: Scope
+      options: [resolution, slice payload, MCP UX, packaging, language support]
+    validations: { required: true }

polyglot_ast_slicer-1.0.0/.github/workflows/ci.yml

@@ -0,0 +1,41 @@
+name: ci
+
+on:
+  push:
+    branches: [main]
+  pull_request:
+
+jobs:
+  test:
+    runs-on: ${{ matrix.os }}
+    strategy:
+      fail-fast: false
+      matrix:
+        os: [ubuntu-latest, macos-latest, windows-latest]
+        python-version: ["3.12", "3.13"]
+    steps:
+      - uses: actions/checkout@v4
+      - uses: astral-sh/setup-uv@v5
+        with:
+          python-version: ${{ matrix.python-version }}
+      - name: Install
+        run: uv sync --all-extras
+      - name: Lint
+        run: uv run ruff check .
+      - name: Type check
+        run: uv run mypy polyglot_ast_slicer
+      - name: Test
+        run: uv run pytest
+      - name: Metrics drift check (offline)
+        run: uv run python scripts/generate_metrics.py --check
+      - name: Build wheel
+        run: uv build --wheel
+      - name: Install wheel in a clean venv and smoke test
+        shell: bash
+        run: |
+          uv venv .smoke-venv
+          PYTHON=.smoke-venv/bin/python
+          if [ -f .smoke-venv/Scripts/python.exe ]; then PYTHON=.smoke-venv/Scripts/python.exe; fi
+          uv pip install --python "$PYTHON" dist/*.whl
+          "$PYTHON" -m polyglot_ast_slicer --version
+          "$PYTHON" verify/ci_smoke.py

polyglot_ast_slicer-1.0.0/.gitignore

@@ -0,0 +1,24 @@
+# Virtual environment & Python cache
+.venv/
+venv/
+__pycache__/
+*.py[cod]
+.pytest_cache/
+.ruff_cache/
+.mypy_cache/
+dist/
+build/
+*.egg-info/
+*.whl
+.coverage
+htmlcov/
+
+# Local SQLite AST cache files
+.ast_slicer_cache.db
+.ast_slicer_cache.db-shm
+.ast_slicer_cache.db-wal
+
+# OS & Editor files
+.DS_Store
+.idea/
+.vscode/

polyglot_ast_slicer-1.0.0/AGENTS.md

@@ -0,0 +1 @@
+Follow CLAUDE.md

polyglot_ast_slicer-1.0.0/CHANGELOG.md

@@ -0,0 +1,109 @@
+# Changelog
+
+Measured numbers are not stored here. They live in `METRICS.md`, which is generated by
+`python scripts/generate_metrics.py` from the committed evaluation scripts, and CI fails if it
+drifts. Parameters such as the 6,000-token payload cap are code constants, not measurements.
+
+## 1.0.0 — 2026-11-06
+
+### Added
+
+- Exhaustive call-site ground truth: Jedi for Python (requests, click, flask; held-out
+  werkzeug) and the TypeScript compiler API for TS/JS (zod, nest, bulletproof-react,
+  node-fetch, express; held-out zustand, moment). Per-repo, per-group and per-form
+  precision/recall with 95% Wilson intervals, wired into `METRICS.md`.
+- `verify/` evidence suite: 5k/20k/50k scale, incremental differential after random edits,
+  watcher end-to-end, network-blocked queries, MCP through the built wheel, robustness
+  inputs, dogfood slices.
+- Slice size comparison (`benchmarks/size_compare.py`) and README worked examples generated
+  from real engine output.
+- CI matrix: ubuntu/macos/windows with Python 3.12 and 3.13, offline metrics/README drift
+  check, wheel build and clean-venv Python/TS smoke test.
+- Windows `LOCALAPPDATA` cache path and simulated Windows/macOS/Linux path tests, plus
+  case-insensitive path and CRLF coverage.
+
+### Fixed
+
+- TypeScript/JavaScript: default imports; emitted specifier mapping (`.js` to `.ts`/`.tsx`,
+  `.mjs`/`.cjs`/`.jsx`); specifiers whose basename contains a dot; `super.m()` against base
+  classes; property-assigned function bodies; bare calls binding to property methods.
+- Python: incremental module reconciliation, call+type relations on one pair, nested
+  `TYPE_CHECKING`/try imports, module-scope symbols, getter/setter and overload spans,
+  `super()` and constructed receivers.
+- Removed stale test-selection references from README, CHANGELOG, CLI help and benchmarks.
+
+### Changed
+
+- `impact_of_diff` returns changed and impacted symbols only; no test selection.
+- `METRICS.md` and `README.md` are generated; `scripts/generate_metrics.py --check` verifies
+  both offline.
+
+### Removed
+
+- `benchmarks/stage_b/run.py`, `benchmarks/fault/run_click.py` and the Stage B METRICS
+  section.
+
+### Notes
+
+- The first run downloads the tree-sitter grammars, so it needs internet once; after that
+  indexing and queries run offline. No code leaves the machine.
+
+### Measured
+
+See `METRICS.md`.
+
+## 0.4.0 — 2026-10-31
+
+### Added
+
+- `impact_of_diff(base, head | diff)`: parses a unified diff, maps changed lines to symbols,
+  walks upstream call edges, and returns changed and impacted symbols by hop.
+  Verified on real git commits.
+- Slice context: each region carries `# class X:` and the referenced `# import:` lines
+  without repeating the sliced code.
+- TypeScript: `tsconfig` `baseUrl`/`paths`, `export *` barrel re-exports, and verified
+  `.tsx`/`.js` grammar loading.
+- `benchmarks/agent_eval/`: dry-run harness that measures tool payload tokens per task and
+  marks resolve rate UNMEASURED. No model adapter is shipped.
+
+### Measured
+
+See `METRICS.md`.
+
+## 0.3.0 — 2026-10-31
+
+### Changed (breaking)
+
+- Removed the 0.50 name-only edge tier; unresolved and ambiguous calls now produce no edge and
+  are tracked in a reverse index for incremental relinking.
+- `get_ast_call_slice` caps the exact JSON payload at 6,000 tokens by default, applies per-hop
+  fan-out caps (10) and a total cap (40), and treats symbols with degree >= 150 as hubs.
+- Metadata requires `include_metadata=True`; the default payload is `compiled_context` plus
+  token metrics and diagnostics.
+- Ambiguous queries return `code: "ambiguous_symbol"` with all candidates.
+- The SQLite index moved from the repository into the OS cache directory.
+- `explain_edge` exposes per-edge provenance (`resolution`, `line`, `receiver`, `name`).
+
+### Added
+
+- Self-recursion is kept as a self-edge and reported separately.
+- `POLYGLOT_AST_ALLOWED_ROOTS` allowlist; escaping symlinks, binary files and oversized files
+  are skipped and reported.
+- Differential test: incremental graph must equal a from-scratch rebuild after random edits.
+
+### Fixed
+
+- Symlinked source files outside the repository crashed indexing.
+- Read-only checkouts failed to open the cache database.
+- Class-level TS static factories made the class look self-recursive.
+- Bare calls such as `open(...)` inside a class bound to a same-named method.
+- Nested/local classes and functions now resolve within their lexical scope.
+
+### Measured
+
+See `METRICS.md`.
+
+## 0.2.0 and earlier
+
+Initial parser, cache, graph engine, MCP server and CLI. Historical numbers are superseded by
+`METRICS.md`.

polyglot_ast_slicer-1.0.0/CLAUDE.md

@@ -0,0 +1,8 @@
+# Guidelines for AI agents on polyglot-ast-slicer
+Bias: caution over speed. For trivial tasks, use judgment.
+1. Think before coding: state assumptions; ask only if blocked, otherwise state the assumption and proceed; list interpretations instead of picking silently; say so if a simpler approach exists.
+2. Simplicity first: minimum code for the request; no speculative features, single-use abstractions or unrequested configurability; no error handling for impossible cases.
+3. Surgical changes: every changed line traces to the request; do not restyle or refactor adjacent code; mention unrelated dead code instead of deleting it; remove only what your change orphaned.
+4. Goal-driven: bug -> failing test first, then fix; refactor -> tests green before and after; for multi-step work write a plan "step -> verify: check".
+Project rules: never invent numbers, APIs or results (unmeasured = UNMEASURED); every figure in README/METRICS comes from a committed script; prefer a missing edge to a wrong edge; never relabel data or lower a gate to pass; engine changes need a regression test plus a re-run of the labeled precision samples.
+Token discipline: don't re-read known files; don't print full test logs; keep reports short.

polyglot_ast_slicer-1.0.0/CONTRIBUTING.md

@@ -0,0 +1,43 @@
+# Contributing
+
+## Setup
+
+```bash
+uv sync --all-extras
+make gate          # offline: ruff, mypy, pytest, lock, metrics-check
+make install-hooks # routes git commit through .githooks/pre-commit
+make metrics       # network: clones pinned repos and rewrites artifacts + METRICS.md
+```
+
+## Gates (must pass before a PR is reviewed)
+
+1. `uv run pytest` — no failures. Skips are only allowed for a grammar that cannot be
+   downloaded; the skip must name the grammar.
+2. `uv run ruff check .`
+3. `uv run mypy polyglot_ast_slicer`
+4. `uv lock --check`
+5. `uv run python scripts/generate_metrics.py --check` — offline; renders `METRICS.md` from
+   the committed `metrics/*.json` artifacts, never clones. `make metrics` is the networked
+   path: it clones the pinned repos (verifying their SHAs), reruns the harnesses, and rewrites
+   both the artifacts and `METRICS.md`. Never hand-edit either.
+6. A wheel install smoke test: `uv build --wheel`, install into a fresh venv, run
+   `polyglot-ast-slicer --version`.
+
+## Resolution changes
+
+- Prefer a missing edge to a wrong edge. Every edge must carry `relation`, `confidence`,
+  `resolution` and `line`; do not add a name-only fallback.
+- Run `benchmarks/edge_eval/precision.py` on `requests` and `click` before and after. Precision
+  must not drop below 98%; if it does, revert the rule.
+- New inference rules need a regression fixture plus the precision delta in the PR description.
+
+## Languages
+
+Do not add a language without a manifest (tier, grammar source, limitations), golden fixtures,
+two pinned repositories and a labelled 100-edge sample. Anything below its tier's precision
+gate ships as experimental and hidden by default.
+
+## Commits
+
+Conventional commits (`feat:`, `fix:`, `docs:`, `test:`). One logical change per commit; gates
+must pass in every commit you push.

polyglot_ast_slicer-1.0.0/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Polyglot AST Slicer contributors
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

polyglot_ast_slicer-1.0.0/METRICS.md

@@ -0,0 +1,150 @@
+# Generated metrics
+
+<!-- Generated by `make metrics`. Do not edit by hand. -->
+
+Pinned inputs: requests `3c7baaa`, click `2247b35`. Intervals are 95% Wilson score intervals.
+Test suite (generated): 107 tests collected.
+
+## Definitions
+
+- **Confirmed by Jedi (strict)** — Jedi's definition lands inside the claimed target's
+  file and line span.
+- **Not contradicted by Jedi (canonical)** — strict, or Jedi lands on a file that
+  re-exports the claimed symbol. This is not precision and is not the headline.
+- **Hand-labeled precision (headline)** — human verdict on the blind 100-edge packet.
+  requests and click are labeled; zod is still unlabeled.
+- **Conditional recall** — found edges divided by call sites Jedi resolves to a project
+  symbol. **Unconditional recall is UNMEASURED.**
+## Hand-labeled precision (headline)
+
+| Repo | Labeled | Correct | Unsure | Precision (unsure=incorrect) | Precision (unsure excluded) |
+|---|---:|---:|---:|---|---|
+| `requests` | 100 | 100 | 0 | 100.0% (95% CI 96.3%–100.0%) | 100.0% (95% CI 96.3%–100.0%) |
+| `click` | 100 | 100 | 0 | 100.0% (95% CI 96.3%–100.0%) | 100.0% (95% CI 96.3%–100.0%) |
+| `zod` | 0 | — | — | UNMEASURED | UNMEASURED |
+
+## Canary check (seeded wrong edges mixed into the blind set)
+
+| Repo | Seeded | Labeled correct | False-accept rate | False rejects | Unsure | Agreement with prior pass | Cohen's kappa |
+|---|---:|---:|---|---:|---:|---|---|
+| `requests` | 20 | 0 | 0.0% (95% CI 0.0%–16.1%) | 0/100 | 0 | 100/100 | n/a |
+| `click` | 15 | 0 | 0.0% (95% CI 0.0%–20.4%) | 0/100 | 0 | n/a (no prior pass) | n/a |
+| `zod` | — | — | UNMEASURED | — | — | — | — |
+
+## Hard-kind precision (v2 packets)
+
+| Repo | Kind | n | Correct | Unsure | Precision (unsure=incorrect) |
+|---|---|---:|---:|---:|---|
+| `requests` | all | — | — | — | UNMEASURED |
+| `click` | all | — | — | — | UNMEASURED |
+| `zod` | all | — | — | — | UNMEASURED |
+## Edge agreement with Jedi (100 random call edges, seed 42)
+
+| Repo | Sampled | Confirmed (strict) | Not contradicted (canonical) |
+|---|---:|---:|---:|
+| `requests` | 100 | 39/100 | 100/100 (100.0%; 95% CI 96.3%–100.0%) |
+| `click` | 100 | 33/100 | 100/100 (100.0%; 95% CI 96.3%–100.0%) |
+
+The canonical column also accepts re-export hops; it is not hand-labeled precision.
+
+## Conditional recall (150 call sites, seed 7; Jedi-resolvable only)
+
+| Repo | Found / resolvable | Recall | Unresolved by Jedi |
+|---|---:|---:|---:|
+| `requests` | 45/58 | 77.6% (95% CI 65.3%–86.4%) | 92 |
+| `click` | 73/77 | 94.8% (95% CI 87.4%–98.0%) | 73 |
+
+Unconditional recall (hand-labelled true targets) is UNMEASURED.
+
+## Python call-site ground truth (Jedi, `src/` only)
+
+Pinned: requests `3c7baaa`, click `2247b35`, flask `49943b5`. Every raw tree-sitter call site is resolved with Jedi (`follow_imports=True`).
+Precision = correct/(correct+wrong); recall = correct/project sites. High-confidence
+means the tool's reported edge confidence is at least 0.95.
+
+| Repo | Sites | Correct | Wrong | Missed | Ambiguous (reported) | Precision | High-conf precision | Recall |
+|---|---:|---:|---:|---:|---:|---|---|---|
+| `requests` | 307 | 276 | 0 | 31 | 0 (0) | 100.0% (95% CI 98.6%–100.0%) | 100.0% (257/257) | 89.9% (95% CI 86.0%–92.8%) |
+| `click` | 740 | 552 | 1 | 187 | 8 (4) | 99.8% (95% CI 99.0%–100.0%) | 99.8% (460/461) | 74.6% (95% CI 71.3%–77.6%) |
+| `flask` | 348 | 210 | 0 | 138 | 2 (0) | 100.0% (95% CI 98.2%–100.0%) | 100.0% (193/193) | 60.3% (95% CI 55.1%–65.3%) |
+
+Combined: 1038 correct, 1 wrong, 356 missed of 1395 project sites; precision 99.9%, recall 74.4%, high-confidence precision 99.9% (910/911).
+
+| Form | Sites | Correct | Wrong | Missed | Ambiguous reported | Precision | Recall |
+|---|---:|---:|---:|---:|---:|---|---|
+| `bare` | 517 | 398 | 1 | 118 | 4 | 99.7% | 77.0% |
+| `module_attr` | 6 | 6 | 0 | 0 | 0 | 100.0% | 100.0% |
+| `self_cls` | 442 | 366 | 0 | 76 | 0 | 100.0% | 82.8% |
+| `super` | 43 | 34 | 0 | 9 | 0 | 100.0% | 79.1% |
+| `constructor` | 175 | 158 | 0 | 17 | 0 | 100.0% | 90.3% |
+| `aliased_import` | 13 | 13 | 0 | 0 | 0 | 100.0% | 100.0% |
+| `chained` | 1 | 0 | 0 | 1 | 0 | 0.0% | 0.0% |
+| `callback` | 13 | 0 | 0 | 13 | 0 | 0.0% | 0.0% |
+| `other` | 185 | 63 | 0 | 122 | 0 | 100.0% | 34.1% |
+
+Independent hand audit: 30/30 sampled unjudged edges correct, with file:line citations in `metrics/python_ground_truth.json`.
+
+## TypeScript/JavaScript call-site ground truth (TypeScript compiler)
+
+Pinned: zod `ca42965` (src; excludes __tests__,benchmarks), nest `416830c` (packages; excludes test,.spec.ts,.e2e-spec.ts), bulletproof-react `9506629` (src; excludes __tests__,test,testing,.test.ts,.test.tsx,.stories.ts,.stories.tsx), node-fetch `8b3320d` (src), express `1faf228` (lib).
+Every call, `new`, and JSX tag in production source is resolved with `checker.getSymbolAtLocation` (aliases followed); test and story files are excluded per repo.
+Precision = correct/(correct+wrong); recall = correct/project sites. High-confidence means the tool's reported confidence is at least 0.95.
+
+| Repo | Group | Sites | Correct | Wrong | Missed | Ambiguous (reported) | Precision | High-conf precision | Recall |
+|---|---|---:|---:|---:|---:|---:|---|---|---|
+| `zod` | TS+TSX | 557 | 452 | 0 | 105 | 0 (0) | 100.0% (95% CI 99.2%–100.0%) | 100.0% (417/417) | 81.1% (95% CI 77.7%–84.2%) |
+| `nest` | TS+TSX | 1998 | 1783 | 0 | 215 | 0 (0) | 100.0% (95% CI 99.8%–100.0%) | 100.0% (1442/1442) | 89.2% (95% CI 87.8%–90.5%) |
+| `bulletproof-react` | TS+TSX | 213 | 159 | 0 | 54 | 0 (0) | 100.0% (95% CI 97.6%–100.0%) | 100.0% (159/159) | 74.6% (95% CI 68.4%–80.0%) |
+| `node-fetch` | JS+JSX | 89 | 84 | 0 | 5 | 0 (0) | 100.0% (95% CI 95.6%–100.0%) | 100.0% (84/84) | 94.4% (95% CI 87.5%–97.6%) |
+| `express` | JS+JSX | 67 | 21 | 0 | 46 | 0 (0) | 100.0% (95% CI 84.5%–100.0%) | 100.0% (20/20) | 31.3% (95% CI 21.5%–43.2%) |
+
+| Group | Sites | Correct | Wrong | Missed | Ambiguous (reported) | Precision | High-conf precision | Recall |
+|---|---:|---:|---:|---:|---:|---|---|---|
+| **TS+TSX** | 2768 | 2394 | 0 | 374 | 0 (0) | 100.0% (95% CI 99.8%–100.0%) | 100.0% (2018/2018) | 86.5% (95% CI 85.2%–87.7%) |
+| **JS+JSX** | 156 | 105 | 0 | 51 | 0 (0) | 100.0% (95% CI 96.5%–100.0%) | 100.0% (104/104) | 67.3% (95% CI 59.6%–74.2%) |
+
+| Group | Form | Sites | Correct | Wrong | Missed | Ambiguous reported | Precision | Recall |
+|---|---|---:|---:|---:|---:|---:|---|---|
+| TS+TSX | `bare` | 581 | 480 | 0 | 101 | 0 | 100.0% | 82.6% |
+| TS+TSX | `self_cls` | 1431 | 1321 | 0 | 110 | 0 | 100.0% | 92.3% |
+| TS+TSX | `constructor` | 333 | 326 | 0 | 7 | 0 | 100.0% | 97.9% |
+| TS+TSX | `chained` | 7 | 0 | 0 | 7 | 0 | 0.0% | 0.0% |
+| TS+TSX | `other` | 313 | 171 | 0 | 142 | 0 | 100.0% | 54.6% |
+| TS+TSX | `super` | 7 | 7 | 0 | 0 | 0 | 100.0% | 100.0% |
+| TS+TSX | `jsx` | 96 | 89 | 0 | 7 | 0 | 100.0% | 92.7% |
+| JS+JSX | `bare` | 86 | 61 | 0 | 25 | 0 | 100.0% | 70.9% |
+| JS+JSX | `self_cls` | 25 | 7 | 0 | 18 | 0 | 100.0% | 28.0% |
+| JS+JSX | `constructor` | 36 | 36 | 0 | 0 | 0 | 100.0% | 100.0% |
+| JS+JSX | `other` | 8 | 0 | 0 | 8 | 0 | 0.0% | 0.0% |
+| JS+JSX | `module_attr` | 1 | 1 | 0 | 0 | 0 | 100.0% | 100.0% |
+
+Gate (high-confidence precision >= 95%, overall precision >= 90%, recall >= 50%): TS+TSX PASS; JS+JSX PASS.
+
+Independent hand audit: 30/30 sampled edges the compiler cannot judge were correct, with file:line citations in `metrics/ts_ground_truth.json`.
+
+## Slice size comparison (30 random symbols per language, seed 42)
+
+Default slice payload tokens vs the entry symbol's whole file vs every unique file containing its callers and callees. Sizes only; no sufficiency claim.
+
+| Language | Symbols | Payload tokens p10 / median / p90 | Entry file p10 / median / p90 | Caller+callee files p10 / median / p90 |
+|---|---:|---|---|---|
+| python | 30 | 259 / 522 / 5422 | 1254 / 4906 / 14296 | 1254 / 5430 / 31961 |
+| typescript | 30 | 256 / 2323 / 5506 | 182 / 1032 / 39169 | 539 / 2404 / 39169 |
+
+## Synthetic token benchmark (hops=2)
+
+| Target | Touched files | Repo | Slice | Saved vs files |
+|---|---:|---:|---:|---:|
+| `process_payment` | 10540 | 12214 | 539 | 94.89% |
+| `handle_checkout_webhook` | 10540 | 12214 | 539 | 94.89% |
+| `submitOrderCheckout` | 1674 | 12214 | 168 | 89.96% |
+
+Corpus: 4 files, 252 symbols, 16 edges. Latency is machine-dependent and omitted here; run `benchmarks/scale/scale_bench.py` for timings.
+
+## Scale (5000 generated files, seed 42)
+
+| Files | Symbols | Edges |
+|---:|---:|---:|
+| 5000 | 5000 | 5000 |
+
+Larger runs: `python benchmarks/scale/scale_bench.py --files 50000 --json`.

polyglot_ast_slicer-1.0.0/Makefile

@@ -0,0 +1,27 @@
+PY ?= uv run --extra dev
+
+.PHONY: gate lint types test lock metrics metrics-check install-hooks
+
+gate: lint types test lock metrics-check
+
+lint:
+	$(PY) ruff check .
+
+types:
+	$(PY) mypy polyglot_ast_slicer
+
+test:
+	$(PY) pytest
+
+lock:
+	uv lock --check
+
+metrics:
+	$(PY) python scripts/generate_metrics.py
+
+metrics-check:
+	$(PY) python scripts/generate_metrics.py --check
+
+install-hooks:
+	git config core.hooksPath .githooks
+	@echo "pre-commit hook installed"