pyspi 3.0.0

.github/workflows/run_dataset_generation.yaml

@@ -1,35 +1,35 @@
 name: Generate benchmarking dataset tables
 
-on: 
+on:
   workflow_dispatch:
 
 jobs:
-  test-ubuntu:
+  generate:
     runs-on: ubuntu-latest
-    strategy:
-      matrix:
-        python-version: ["3.9"]
+    # Regenerating the baselines runs the full calculator many times over.
+    timeout-minutes: 360
     steps:
       - uses: actions/checkout@v4
-      - name: Setup python ${{ matrix.python-version }}
+
+      - name: Setup python 3.12
         uses: actions/setup-python@v5
         with:
-          python-version: ${{ matrix.python-version }}
+          python-version: "3.12"
           cache: 'pip'
-      - name: Install octave
-        run: |
-          sudo apt-get update
-          sudo apt-get install -y build-essential octave
-      - name: Install pyspi dependencies
+
+      - name: Install pyspi
         run: |
           python -m pip install --upgrade pip
-          pip install -r requirements.txt
           pip install .
+
       - name: Run data generation
-        run: |
-          python tests/generate_benchmark_tables.py
+        run: python tests/tools/generate_benchmark_tables.py --dataset all
+
       - name: Upload artifact
         uses: actions/upload-artifact@v4
         with:
           name: benchmark-tables
-          path: tests/CML7_benchmark_tables_new.pkl
+          # Glob rather than a fixed filename so this keeps working regardless of
+          # which baselines the script writes and what it names them.
+          path: tests/data/baselines/*.npz
+          if-no-files-found: error

.github/workflows/run_package_tests.yaml

@@ -0,0 +1,108 @@
+name: Packaging Pipeline
+
+# The unit-test job installs pyspi as an editable checkout, so it exercises the
+# repository, not the artefact users receive. Everything that can only break in
+# a built wheel -- a data file left out of package-data, a config or dataset
+# resolved relative to the source tree, a console entry point that is not
+# wired up -- is invisible to it. This job builds the wheel, installs it into a
+# clean environment, and runs from a directory that contains no pyspi source.
+
+on: [push, pull_request]
+
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: true
+
+jobs:
+  wheel:
+    name: wheel / python ${{ matrix.python-version }}
+    runs-on: ubuntu-latest
+    timeout-minutes: 30
+    strategy:
+      fail-fast: false
+      matrix:
+        python-version: ["3.10", "3.12"]
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Setup python ${{ matrix.python-version }}
+        uses: actions/setup-python@v5
+        with:
+          python-version: ${{ matrix.python-version }}
+
+      - name: Build the wheel
+        run: |
+          python -m pip install --upgrade pip build
+          python -m build --wheel --outdir dist/
+
+      - name: Install it into a clean environment
+        run: |
+          python -m venv /tmp/clean
+          /tmp/clean/bin/pip install --upgrade pip
+          /tmp/clean/bin/pip install dist/*.whl
+
+      # Run from an empty directory: from the repository root, `import pyspi`
+      # finds the source tree and the installed package is never touched.
+      - name: Import, resolve every bundled config and dataset, run fabfour
+        run: |
+          mkdir -p /tmp/run && cd /tmp/run
+          /tmp/clean/bin/python - <<'PY'
+          import os, tempfile
+          import numpy as np
+          import pyspi
+          from pyspi.calculator import (Calculator, bundled_configs, load_table,
+                                        load_spis_from_yaml, resolve_config)
+          from pyspi.data import available_datasets, load_dataset
+
+          for name in bundled_configs():
+              path = resolve_config(name)
+              assert os.path.exists(path), f"{name} -> {path}"
+              assert load_spis_from_yaml(path, quiet=True), name
+          for name in available_datasets():
+              load_dataset(name)
+
+          rng = np.random.default_rng(0)
+          calc = Calculator(dataset=rng.standard_normal((3, 100)), config="fabfour")
+          calc.compute()
+          assert not calc.errors, calc.errors
+
+          out = os.path.join(tempfile.mkdtemp(), "results.npz")
+          calc.save(out)
+          table = load_table(out)
+          assert np.allclose(table.to_numpy(), calc.table.to_numpy(), equal_nan=True)
+          assert table.attrs["run_digest"] == calc.run_digest
+          print("package smoke test OK")
+          PY
+
+      - name: Exercise the CLI
+        run: |
+          mkdir -p /tmp/cli && cd /tmp/cli
+          /tmp/clean/bin/python -c "import numpy as np; np.save('d.npy', np.random.default_rng(0).standard_normal((3, 100)))"
+          /tmp/clean/bin/python -m pyspi compute --data d.npy --config fabfour --output r.npz --quiet
+          /tmp/clean/bin/python -m pyspi --help
+          /tmp/clean/bin/python -c "
+          from pyspi.calculator import load_table
+          t = load_table('r.npz')
+          assert t.shape[0] == 3, t.shape
+          print('CLI round-trip OK')
+          "
+
+  lock:
+    name: locked resolution (uv.lock)
+    runs-on: ubuntu-latest
+    timeout-minutes: 30
+    steps:
+      - uses: actions/checkout@v4
+      - uses: astral-sh/setup-uv@v5
+
+      # The matrix above floats to the latest compatible dependencies, which is
+      # what catches upstream breakage. This pins to the committed lock, which
+      # is what makes a failure there attributable: if both jobs fail the change
+      # is ours, if only the floating one does it is a dependency's.
+      - name: Check the lock is current
+        run: uv lock --check
+
+      - name: Run the fast suite against the locked resolution
+        run: |
+          uv sync --locked --extra testing
+          uv run pytest -q

.github/workflows/run_slow_tests.yaml

@@ -0,0 +1,37 @@
+name: Slow Regression Suite
+
+# The slow suite compares every SPI against frozen baseline tables and takes
+# a couple of minutes plus three full-config computations, so it is not run
+# per-push. Pull requests, weekly, and on demand.
+on:
+  pull_request:
+  schedule:
+    # 04:00 UTC every Monday.
+    - cron: '0 4 * * 1'
+  workflow_dispatch:
+
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: true
+
+jobs:
+  slow-test:
+    name: regression (ubuntu / python 3.12)
+    runs-on: ubuntu-latest
+    timeout-minutes: 90
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Setup python 3.12
+        uses: actions/setup-python@v5
+        with:
+          python-version: "3.12"
+          cache: 'pip'
+
+      - name: Install pyspi
+        run: |
+          python -m pip install --upgrade pip
+          pip install -e '.[testing]'
+
+      - name: Run regression tests
+        run: pytest -m slow

.github/workflows/run_unit_tests.yaml

@@ -1,35 +1,37 @@
 name: Unit Testing Pipeline
 
-on: 
-  push:
+on: [push, pull_request]
+
+# Supersede in-flight runs for the same ref; keep runs on different refs independent.
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: true
 
 jobs:
-  test-ubuntu:
-    runs-on: ubuntu-latest
+  test:
+    name: ${{ matrix.os }} / python ${{ matrix.python-version }}
+    runs-on: ${{ matrix.os }}
+    timeout-minutes: 30
     strategy:
+      fail-fast: false
       matrix:
-        python-version: ["3.8", "3.9", "3.10", "3.11", "3.12"]
+        os: [ubuntu-latest, macos-latest]
+        python-version: ["3.10", "3.11", "3.12"]
     steps:
       - uses: actions/checkout@v4
+
       - name: Setup python ${{ matrix.python-version }}
         uses: actions/setup-python@v5
         with:
           python-version: ${{ matrix.python-version }}
           cache: 'pip'
-      - name: Install octave
-        run: |
-          sudo apt-get update
-          sudo apt-get install -y build-essential octave
-      - name: Install pyspi dependencies
+
+      - name: Install pyspi
         run: |
           python -m pip install --upgrade pip
-          pip install setuptools
-          pip install -r requirements.txt
-          pip install .
-      - name: Run pyspi calculator/utils unit tests
-        run: |
-          pytest -v ./tests/test_calc.py
-          pytest -v ./tests/test_utils.py
-      - name: Run pyspi SPI unit tests
-        run: |
-          pytest -v ./tests/test_SPIs.py
+          pip install -e '.[testing]'
+
+      - name: Run unit tests
+        # pyproject.toml sets addopts = "-m 'not slow'", so the baseline-drift
+        # suite is excluded here by design. See run_slow_tests.yaml for that.
+        run: pytest -q

.gitignore

@@ -1,13 +1,35 @@
-pyspi.egg-info
-pyspi/__pycache__
+# Build artefacts
+build/
+dist/
+*.egg-info/
 **/__pycache__/**
-dist
+
+# Environments and tool caches
+.venv/
+.pytest_cache/
+.ruff_cache/
+
+# Data / logs. Regression baselines under tests/ are tracked deliberately.
 *.pkl
 !tests/*.pkl
 *.log
 *.bkp
+
+# Editor / OS
 .DS_Store
-.vscode
-build
-octave-workspace
-readme.html
+.vscode/
+
+# Exploratory notebooks: regenerable from bench/results/cells + analyse_cells,
+# and they carry large non-rendering outputs. report.md is the tracked summary.
+bench/results/analysis/*.ipynb
+
+# Raw benchmark cells are large, machine-specific inputs to the tracked summaries.
+bench/results/cells/physics_config*.json
+
+# Local assistant instructions
+CLAUDE.md
+AGENTS.md
+.claude/
+
+# Review/critique artifacts (not part of the package)
+critiques.md