Move pre-commit to ruff and run fast tests on Python 3.13 and 3.14

.github/workflows/pre-commit.yml

@@ -1,6 +1,6 @@
 # Pre-commit / lint check.
 #
-# Runs `pre-commit run --all-files` (black, isort, flake8, pyupgrade, mypy,
+# Runs `pre-commit run --all-files` (ruff-check, ruff-format, mypy,
 # end-of-file-fixer, trailing-whitespace, check-merge-conflict, debug-statements).
 # ~30 s wall. Runs on every PR + push to main.
 

.github/workflows/test-fast.yml

@@ -1,6 +1,6 @@
 # Fast tests (PR gate).
 #
-# `pytest -m fast -n 2` on Python 3.11 + 3.13 matrix. ~1.5 min wall per Python.
+# `pytest -m fast -n 2` on Python 3.13 + 3.14 matrix. ~1.5 min wall per Python.
 # 676 tests in the current `fast` set (no JAX trace, < 100 ms each).
 # Runs on every PR + push to main.
 
@@ -23,8 +23,14 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        # pyproject declares ``requires-python = ">=3.13"``. Matrix matches.
-        python: ['3.13']
+        # pyproject declares ``requires-python = ">=3.13"``. Matrix covers
+        # both supported Pythons (3.13 and 3.14).
+        python: ['3.13', '3.14']
+    env:
+      # Pin every uv command to the matrix Python. Without this, `uv run`
+      # falls back to .python-version (3.13) and rebuilds the env without
+      # the bench group, so the 3.14 leg fails with "Failed to spawn: pytest".
+      UV_PYTHON: ${{ matrix.python }}
     steps:
       - name: Check out repo
         uses: actions/checkout@v4

.pre-commit-config.yaml

@@ -1,50 +1,34 @@
-# Lint stack for tuningfork. Versions bumped from sampling-book's 2022-era pins
-# so they install cleanly on Python 3.13. Tools and ignore lists kept identical
-# in spirit to sampling-book/blackjax for cross-repo consistency.
+# Lint stack for tuningfork. ruff (ruff-check + ruff-format) replaces
+# isort/pyupgrade/flake8/black/nbQA — one tool, one config, native Jupyter
+# notebook support if any .ipynb ever lands. Tools kept identical in spirit
+# to sampling-book/blackjax for cross-repo consistency.
 repos:
 - repo: https://github.com/pre-commit/pre-commit-hooks
-  rev: v5.0.0
+  rev: v6.0.0
   hooks:
     - id: check-merge-conflict
     - id: debug-statements
     - id: end-of-file-fixer
     - id: trailing-whitespace
-- repo: https://github.com/PyCQA/isort
-  rev: 5.13.2
-  hooks:
-    - id: isort
-      args: [--profile, black]
-- repo: https://github.com/asottile/pyupgrade
-  rev: v3.19.0
-  hooks:
-    - id: pyupgrade
-      args: [--py311-plus]
-- repo: https://github.com/PyCQA/flake8
-  rev: 7.1.1
-  hooks:
-    - id: flake8
-      args: ['--ignore=E501,E203,E731,W503']
-      exclude: '^scripts/'
-- repo: https://github.com/psf/black
-  rev: 24.10.0
-  hooks:
-    - id: black
+- repo: https://github.com/astral-sh/ruff-pre-commit
+  rev: v0.16.10
+  hooks:
+    - id: ruff-check
+      args: [--fix]
+    - id: ruff-format
+      # ruff-format's hook default also formats fenced code blocks in
+      # markdown (types_or includes "markdown"); the old stack
+      # (black/isort/flake8/nbQA) never touched .md prose docs, only .py
+      # and .ipynb. Keep that scope so curated doc formatting (aligned
+      # inline comments in catalog READMEs/schemas) isn't rewritten as a
+      # side effect of this migration.
+      types_or: [python, pyi, jupyter]
 - repo: https://github.com/pre-commit/mirrors-mypy
-  rev: v1.13.0
+  rev: v2.4.0
   hooks:
     - id: mypy
       args: [--ignore-missing-imports]
       exclude: '^scripts/'
-- repo: https://github.com/nbQA-dev/nbQA
-  rev: 1.9.0
-  hooks:
-    - id: nbqa-black
-    - id: nbqa-pyupgrade
-      args: [--py311-plus]
-    - id: nbqa-isort
-      args: ['--profile=black']
-    - id: nbqa-flake8
-      args: ['--ignore=E501,E203,E302,E402,E731,W503']
 - repo: https://github.com/jorisroovers/gitlint
   rev: v0.19.1
   hooks:

CONTRIBUTING.md

@@ -126,7 +126,7 @@ All `test-*` targets automatically run `make clean-orphans` first (except `test-
        assert logp == 0.0
    ```
 
-5. **Run `make lint` before committing** to pass pre-commit hooks (black, isort, flake8, mypy).
+5. **Run `make lint` before committing** to pass pre-commit hooks (ruff, mypy).
 
 ## Pre-Commit and Commit Messages
 

benchmarks/_benchmark_helpers.py

@@ -15,6 +15,7 @@
 
 Used by both test_fast_recipes.py and test_e2e_recipes.py.
 """
+
 from __future__ import annotations
 
 import json

benchmarks/config.py

@@ -27,6 +27,7 @@
 The split keeps the fast suite quick enough for local smoke + targeted CI
 triggers, while the slow e2e cells run nightly only.
 """
+
 from __future__ import annotations
 
 # ---------------------------------------------------------------------------

benchmarks/test_e2e_recipes.py

@@ -24,6 +24,7 @@
 DO NOT add these to ``make benchmark-fast`` or per-PR triggers — the wall
 time makes them unsuitable for quick local checks.
 """
+
 from __future__ import annotations
 
 from typing import Any

benchmarks/test_fast_recipes.py

@@ -25,6 +25,7 @@
 Cell selection: see ``benchmarks/config.py`` (FAST_CELLS, ≤60s/cell in CI).
 Slow e2e cells (>60s): see ``benchmarks/test_e2e_recipes.py``.
 """
+
 from __future__ import annotations
 
 from typing import Any

benchmarks/test_speed_lite.py

@@ -41,6 +41,7 @@
   Trend persisted in ``actions/cache`` (not gh-pages).
   Alert at 200% with ``fail-on-alert: true`` (nightly red = regression signal).
 """
+
 from __future__ import annotations
 
 from pathlib import Path

pyproject.toml

@@ -126,8 +126,24 @@ exclude = [
 [tool.setuptools_scm]
 write_to = "tuningfork/_version.py"
 
-[tool.isort]
-profile = "black"
+[tool.ruff]
+target-version = "py313"
+line-length = 88
+
+[tool.ruff.lint]
+select = ["E", "F", "W", "I", "UP"]   # = the old flake8 (pycodestyle+pyflakes) + isort + pyupgrade
+ignore = [
+  "E501",                      # line length: ruff-format owns it (was ignored before)
+  "E731",                      # lambda assignment (was ignored before)
+  "UP040", "UP046", "UP047",   # PEP 695 `type`/generic rewrites: change runtime semantics, out of scope
+  "UP042",                     # str+Enum -> StrEnum changes str()/f-string output ("Effort.LOW" ->
+                                # "low"); test_schema.py:61 asserts the current str() form and
+                                # _resolve_execution_plan.py builds a file path from an f-strung
+                                # effort value, so this rewrite is a behavior change, not a style fix
+]
+
+[tool.ruff.lint.isort]
+known-first-party = ["tuningfork"]
 
 [tool.pytest.ini_options]
 filterwarnings = [

tests/e2e/test_nightly_regression.py

@@ -22,6 +22,7 @@
     make test-slow
     JAX_PLATFORM_NAME=cpu uv run pytest tests/e2e/test_nightly_regression.py -v
 """
+
 from __future__ import annotations
 
 import pytest
@@ -81,9 +82,9 @@ def test_lotka_dense_imm_inner_nuts_seed_20260713_passes() -> None:
         catalog_root=catalog,
     )
     z = compute_max_abs_mean_z(idata, "lotka_volterra")
-    assert (
-        z is not None
-    ), "compute_max_abs_mean_z returned None — reference/summary.json missing?"
+    assert z is not None, (
+        "compute_max_abs_mean_z returned None — reference/summary.json missing?"
+    )
     assert z < 4.0, (
         f"Seed 20260713 should pass z < 4.0 (it is a clean seed), got z={z:.3f}. "
         "If this seed now fails, the step-collapse has worsened — see issue #232."

tests/e2e/test_phase1.py

@@ -75,9 +75,9 @@ def test_mvn_cache_hit_second_run(self, tmp_path: Path) -> None:
             second_meta = json.load(fh)
         second_ts = second_meta["timestamp_utc"]
 
-        assert (
-            first_ts == second_ts
-        ), f"Cache hit expected but timestamp changed: {first_ts!r} → {second_ts!r}"
+        assert first_ts == second_ts, (
+            f"Cache hit expected but timestamp changed: {first_ts!r} → {second_ts!r}"
+        )
 
     def test_mvn_force_regenerates(self, tmp_path: Path) -> None:
         """--force must update the timestamp (regeneration happened)."""
@@ -94,9 +94,9 @@ def test_mvn_force_regenerates(self, tmp_path: Path) -> None:
             second_meta = json.load(fh)
         second_ts = second_meta["timestamp_utc"]
 
-        assert (
-            first_ts != second_ts
-        ), f"--force expected regeneration but timestamp did not change: {first_ts!r}"
+        assert first_ts != second_ts, (
+            f"--force expected regeneration but timestamp did not change: {first_ts!r}"
+        )
 
     def test_mvn_output_contains_summary(self, tmp_path: Path) -> None:
         """CLI must print a summary table with expected fields."""

tests/e2e/test_phase3.py

@@ -47,32 +47,32 @@ class TestLeaderboardCLI:
     def test_leaderboard_mvn_10_markdown(self, tmp_path: Path) -> None:
         """tuningfork leaderboard mvn_10 must exit 0 and print markdown table."""
         result = _run_leaderboard(["mvn_10"], tmp_path)
-        assert (
-            result.returncode == 0
-        ), f"Expected exit 0.\nstdout:\n{result.stdout}\nstderr:\n{result.stderr}"
+        assert result.returncode == 0, (
+            f"Expected exit 0.\nstdout:\n{result.stdout}\nstderr:\n{result.stderr}"
+        )
         stdout = result.stdout
         # Table header and title must be present
-        assert (
-            "Leaderboard for mvn_10" in stdout
-        ), f"'Leaderboard for mvn_10' missing from stdout:\n{stdout}"
-        assert (
-            "| effort |" in stdout
-        ), f"Markdown table header missing from stdout:\n{stdout}"
+        assert "Leaderboard for mvn_10" in stdout, (
+            f"'Leaderboard for mvn_10' missing from stdout:\n{stdout}"
+        )
+        assert "| effort |" in stdout, (
+            f"Markdown table header missing from stdout:\n{stdout}"
+        )
         assert "|" in stdout, f"Table pipes missing from stdout:\n{stdout}"
         # At least a few rows expected
         lines = stdout.strip().split("\n")
         table_lines = [line for line in lines if line.startswith("|")]
         # header + separator + at least 2 data rows
-        assert (
-            len(table_lines) >= 4
-        ), f"Expected at least 2 data rows in markdown table, got {len(table_lines)} table lines:\n{stdout}"
+        assert len(table_lines) >= 4, (
+            f"Expected at least 2 data rows in markdown table, got {len(table_lines)} table lines:\n{stdout}"
+        )
 
     def test_leaderboard_mvn_10_effort_medium(self, tmp_path: Path) -> None:
         """tuningfork leaderboard mvn_10 --effort medium filters to MEDIUM rows."""
         result = _run_leaderboard(["mvn_10", "--effort", "medium"], tmp_path)
-        assert (
-            result.returncode == 0
-        ), f"Expected exit 0.\nstdout:\n{result.stdout}\nstderr:\n{result.stderr}"
+        assert result.returncode == 0, (
+            f"Expected exit 0.\nstdout:\n{result.stdout}\nstderr:\n{result.stderr}"
+        )
         stdout = result.stdout
         assert "Leaderboard for mvn_10" in stdout
         # Count MEDIUM-effort rows (ignoring header/separator)
@@ -82,29 +82,29 @@ def test_leaderboard_mvn_10_effort_medium(self, tmp_path: Path) -> None:
             for line in lines
             if line.startswith("|") and "effort" not in line and "-----" not in line
         ]
-        assert (
-            len(data_lines) >= 1
-        ), f"Expected at least 1 MEDIUM-effort row, got {len(data_lines)} rows:\n{stdout}"
+        assert len(data_lines) >= 1, (
+            f"Expected at least 1 MEDIUM-effort row, got {len(data_lines)} rows:\n{stdout}"
+        )
         # All data rows should have "medium" in the effort column.
         for line in data_lines:
             parts = line.split("|")
             effort_col = parts[1].strip()
-            assert (
-                effort_col == "medium"
-            ), f"Expected 'medium' in effort column, got '{effort_col}' in line:\n{line}"
+            assert effort_col == "medium", (
+                f"Expected 'medium' in effort column, got '{effort_col}' in line:\n{line}"
+            )
 
     def test_leaderboard_mvn_10_json_format(self, tmp_path: Path) -> None:
         """tuningfork leaderboard mvn_10 --format json must output JSON list."""
         result = _run_leaderboard(["mvn_10", "--format", "json"], tmp_path)
-        assert (
-            result.returncode == 0
-        ), f"Expected exit 0.\nstdout:\n{result.stdout}\nstderr:\n{result.stderr}"
+        assert result.returncode == 0, (
+            f"Expected exit 0.\nstdout:\n{result.stdout}\nstderr:\n{result.stderr}"
+        )
         # Parse JSON output
         data = json.loads(result.stdout)
         assert isinstance(data, list), f"Expected JSON list, got {type(data)}"
-        assert (
-            len(data) >= 1
-        ), f"Expected at least 1 element in JSON list, got {len(data)}"
+        assert len(data) >= 1, (
+            f"Expected at least 1 element in JSON list, got {len(data)}"
+        )
         # Verify schema
         for item in data:
             assert "effort" in item
@@ -116,11 +116,11 @@ def test_leaderboard_mvn_10_json_format(self, tmp_path: Path) -> None:
     def test_leaderboard_bad_model_exits_2(self, tmp_path: Path) -> None:
         """tuningfork leaderboard does_not_exist must exit 2 and mention model in stderr."""
         result = _run_leaderboard(["does_not_exist"], tmp_path)
-        assert (
-            result.returncode == 2
-        ), f"Expected exit 2 for bad model.\nstdout:\n{result.stdout}\nstderr:\n{result.stderr}"
+        assert result.returncode == 2, (
+            f"Expected exit 2 for bad model.\nstdout:\n{result.stdout}\nstderr:\n{result.stderr}"
+        )
         # Stderr should mention model or list valid models
         stderr_lower = result.stderr.lower()
-        assert (
-            "does_not_exist" in result.stderr or "model" in stderr_lower
-        ), f"Expected bad-model name or 'model' in stderr:\n{result.stderr}"
+        assert "does_not_exist" in result.stderr or "model" in stderr_lower, (
+            f"Expected bad-model name or 'model' in stderr:\n{result.stderr}"
+        )

tests/groundtruth/test_analytic_iid.py

@@ -84,12 +84,12 @@ def test_analytic_iid_smoke(model_name: str, tmp_path: Path) -> None:
     with np.load(str(draws_path), allow_pickle=True) as draws:
         for site in draws.files:
             shape = draws[site].shape
-            assert (
-                shape[0] == _SMOKE_N_CHAINS
-            ), f"{model_name}.{site}: expected n_chains={_SMOKE_N_CHAINS}, got {shape[0]}"
-            assert (
-                shape[1] == _SMOKE_N_DRAWS
-            ), f"{model_name}.{site}: expected n_draws={_SMOKE_N_DRAWS}, got {shape[1]}"
+            assert shape[0] == _SMOKE_N_CHAINS, (
+                f"{model_name}.{site}: expected n_chains={_SMOKE_N_CHAINS}, got {shape[0]}"
+            )
+            assert shape[1] == _SMOKE_N_DRAWS, (
+                f"{model_name}.{site}: expected n_draws={_SMOKE_N_DRAWS}, got {shape[1]}"
+            )
 
     # --- crude mean coherence vs committed GT ---
     # At smoke scale (2 chains × 50 draws = 100 samples), the new mean can
@@ -123,9 +123,9 @@ def test_analytic_iid_smoke(model_name: str, tmp_path: Path) -> None:
                 "wrong distribution"
             )
             # Also check means are finite (catches NaN/Inf in sampler output)
-            assert np.all(
-                np.isfinite(new_mean)
-            ), f"{model_name}.{site}: new mean contains non-finite values"
+            assert np.all(np.isfinite(new_mean)), (
+                f"{model_name}.{site}: new mean contains non-finite values"
+            )
 
 
 def test_analytic_iid_verify_integration(tmp_path: Path) -> None:
@@ -197,6 +197,6 @@ def test_analytic_iid_custom_seed_differs(tmp_path: Path) -> None:
         np.load(str(out2 / "draws.npz"), allow_pickle=True) as d2,
     ):
         site = d1.files[0]
-        assert not np.allclose(
-            d1[site], d2[site]
-        ), "Different seeds produced identical draws — RNG not working correctly"
+        assert not np.allclose(d1[site], d2[site]), (
+            "Different seeds produced identical draws — RNG not working correctly"
+        )