Fix LeastSquares for functions with more than two arguments

.github/workflows/test.yml

@@ -46,7 +46,7 @@ jobs:
     - uses: actions/setup-python@v5
       with:
         python-version: ${{ matrix.python-version }}
-    - run: uv pip install --system -v . pytest ${{ matrix.installs }}
-    # python -m pip install .[test] is not used here to test minimum (faster),
-    # the cov workflow runs all tests.
-    - run: python -m pytest
+    # pip install .[test] is not used here to test minimum (faster)
+    # cov workflow runs all tests
+    - run: uv pip install --system . pytest pytest-xdist ${{ matrix.installs }}
+    - run: python -m pytest -n 3

.gitignore

@@ -32,12 +32,13 @@ bench/*.svg
 .project
 .pydevproject
 .settings
-.coverage
+.coverage*
 .ipynb_checkpoints
 .eggs
 .pytest_cache
 .mypy_cache
 .ruff_cache
+.nox
 
 Untitled*.ipynb
 Untitled*.py

noxfile.py

@@ -1,11 +1,14 @@
 """
-Noxfile for iMinuit.
+Noxfile for iminuit.
 
-Use `-R` to instantly reuse an existing environment and
-to avoid rebuilding the binary.
+Pass extra arguments to pytest after --
 """
 
 import nox
+import sys
+
+sys.path.append(".")
+import python_releases
 
 nox.needs_version = ">=2024.3.2"
 nox.options.default_venv_backend = "uv|virtualenv"
@@ -15,46 +18,50 @@
     "COVERAGE_CORE": "sysmon",  # faster coverage on Python 3.12
 }
 
+PYPROJECT = nox.project.load_toml("pyproject.toml")
+MINIMUM_PYTHON = PYPROJECT["project"]["requires-python"].strip(">=")
+LATEST_PYTHON = str(python_releases.latest())
+
 nox.options.sessions = ["test", "mintest", "maxtest"]
 
 
-@nox.session(reuse_venv=True)
+@nox.session()
 def test(session: nox.Session) -> None:
-    """Run the unit and regular tests."""
+    """Run all tests."""
     session.install("-e.[test]")
-    session.run("pytest", *session.posargs, env=ENV)
+    session.run("pytest", "-n=auto", *session.posargs, env=ENV)
 
 
-@nox.session(python="3.12", reuse_venv=True)
-def maxtest(session: nox.Session) -> None:
-    """Run the unit and regular tests."""
-    session.install("-e.", "scipy", "matplotlib", "pytest", "--pre")
-    session.run("pytest", *session.posargs, env=ENV)
+@nox.session(python=MINIMUM_PYTHON, venv_backend="uv")
+def mintest(session: nox.Session) -> None:
+    """Run tests on the minimum python version."""
+    session.install("-e.", "--resolution=lowest-direct")
+    session.install("pytest", "pytest-xdist")
+    session.run("pytest", "-n=auto", *session.posargs)
 
 
-@nox.session(python="3.9", venv_backend="uv")
-def mintest(session: nox.Session) -> None:
+@nox.session(python=LATEST_PYTHON)
+def maxtest(session: nox.Session) -> None:
     """Run the unit and regular tests."""
-    session.install("-e.", "--resolution=lowest-direct")
-    session.install("pytest")
-    session.run("pytest", *session.posargs)
+    session.install("-e.", "scipy", "matplotlib", "pytest", "pytest-xdist", "--pre")
+    session.run("pytest", "-n=auto", *session.posargs, env=ENV)
 
 
-@nox.session(python="pypy3.9", venv_backend="uv")
+@nox.session(python="pypy3.9")
 def pypy(session: nox.Session) -> None:
     """Run the unit and regular tests."""
     session.install("-e.")
-    session.install("pytest")
-    session.run("pytest", *session.posargs)
+    session.install("pytest", "pytest-xdist")
+    session.run("pytest", "-n=auto", *session.posargs)
 
 
 # Python-3.12 provides coverage info faster
-@nox.session(python="3.12", reuse_venv=True)
+@nox.session(python="3.12", venv_backend="uv")
 def cov(session: nox.Session) -> None:
     """Run covage and place in 'htmlcov' directory."""
     session.install("-e.[test,doc]")
     session.run("coverage", "run", "-m", "pytest", env=ENV)
-    session.run("coverage", "html", "-d", "htmlcov")
+    session.run("coverage", "html", "-d", "build/htmlcov")
     session.run("coverage", "report", "-m")
 
 

pyproject.toml

@@ -57,6 +57,7 @@ test = [
     "numba; platform_python_implementation=='CPython'",
     "numba-stats; platform_python_implementation=='CPython'",
     "pytest",
+    "pytest-xdist",
     "scipy",
     "tabulate",
     "boost_histogram",

python_releases.py

@@ -0,0 +1,63 @@
+"""Get the latest Python release which is online."""
+
+import urllib.request
+import re
+from html.parser import HTMLParser
+import gzip
+from packaging.version import Version
+
+
+class PythonVersionParser(HTMLParser):
+    """Specialized HTMLParser to get Python version number."""
+
+    def __init__(self):
+        super().__init__()
+        self.versions = set()
+        self.found_version = False
+
+    def handle_starttag(self, tag, attrs):
+        """Look for the right tag and store result in an attribute."""
+        if tag == "a":
+            for attr in attrs:
+                if attr[0] == "href" and "/downloads/release/python-" in attr[1]:
+                    self.found_version = True
+                    return
+
+    def handle_data(self, data):
+        """Extract Python version from entry."""
+        if self.found_version:
+            self.found_version = False
+            match = re.search(r"Python (\d+\.\d+)", data)
+            if match:
+                self.versions.add(Version(match.group(1)))
+
+
+def versions():
+    """Get all Python release versions."""
+    req = urllib.request.Request("https://www.python.org/downloads/")
+    req.add_header("Accept-Encoding", "gzip")
+
+    with urllib.request.urlopen(req) as response:
+        raw = response.read()
+        if response.info().get("Content-Encoding") == "gzip":
+            raw = gzip.decompress(raw)
+        html = raw.decode("utf-8")
+
+    parser = PythonVersionParser()
+    parser.feed(html)
+
+    return parser.versions
+
+
+def latest():
+    """Return version of latest Python release."""
+    return max(versions())
+
+
+def main():
+    """Print all discovered release versions."""
+    print(" ".join(str(x) for x in sorted(versions())))
+
+
+if __name__ == "__main__":
+    main()

src/iminuit/cost.py

@@ -2195,7 +2195,7 @@ def __init__(
         y = _norm(y)
         assert x.ndim >= 1  # guaranteed by _norm
 
-        self._ndim = x.ndim
+        self._ndim = x.shape[0] if x.ndim > 1 else 1
         self._model = model
         self._model_grad = grad
         self.loss = loss

src/iminuit/typing.py

@@ -4,16 +4,18 @@
 These are used by mypy and similar tools.
 """
 
-from typing import Protocol, Optional, List, Union, runtime_checkable, NamedTuple
+from typing import (
+    Protocol,
+    Optional,
+    List,
+    Union,
+    runtime_checkable,
+    NamedTuple,
+    Annotated,
+)
 from numpy.typing import NDArray
 import numpy as np
 import dataclasses
-import sys
-
-if sys.version_info < (3, 9):
-    from typing_extensions import Annotated  # noqa pragma: no cover
-else:
-    from typing import Annotated  # noqa pragma: no cover
 
 
 # Key for ValueView, ErrorView, etc.

src/iminuit/util.py

@@ -27,17 +27,15 @@
     Collection,
     Sequence,
     TypeVar,
+    Annotated,
+    get_args,
+    get_origin,
 )
 import abc
 from time import monotonic
 import warnings
 import sys
 
-if sys.version_info < (3, 9):
-    from typing_extensions import Annotated, get_args, get_origin  # pragma: no cover
-else:
-    from typing import Annotated, get_args, get_origin  # pragma: no cover
-
 T = TypeVar("T")
 
 __all__ = (

tests/test_cost.py

@@ -1066,34 +1066,69 @@ def model(x, a, b):
 
 
 def test_LeastSquares_2D():
-    x = np.array([1.0, 2.0, 3.0])
-    y = np.array([4.0, 5.0, 6.0])
-    z = 1.5 * x + 0.2 * y
-    ze = 1.5
-
     def model(xy, a, b):
         x, y = xy
         return a * x + b * y
 
-    c = LeastSquares((x, y), z, ze, model, grad=numerical_model_gradient(model))
+    x = np.array([1.0, 2.0, 3.0])
+    y = np.array([4.0, 5.0, 6.0])
+    f = model((x, y), 1.5, 0.2)
+    fe = 1.5
+
+    c = LeastSquares((x, y), f, fe, model, grad=numerical_model_gradient(model))
     assert c.ndata == 3
 
     ref = numerical_cost_gradient(c)
     assert_allclose(c.grad(1, 2), ref(1, 2))
 
     assert_equal(c.x, (x, y))
-    assert_equal(c.y, z)
-    assert_equal(c.yerror, ze)
+    assert_equal(c.y, f)
+    assert_equal(c.yerror, fe)
     assert_allclose(c(1.5, 0.2), 0.0)
-    assert_allclose(c(2.5, 0.2), np.sum(((z - 2.5 * x - 0.2 * y) / ze) ** 2))
-    assert_allclose(c(1.5, 1.2), np.sum(((z - 1.5 * x - 1.2 * y) / ze) ** 2))
+    assert_allclose(c(2.5, 0.2), np.sum(((f - 2.5 * x - 0.2 * y) / fe) ** 2))
+    assert_allclose(c(1.5, 1.2), np.sum(((f - 1.5 * x - 1.2 * y) / fe) ** 2))
 
-    c.y = 2 * z
-    assert_equal(c.y, 2 * z)
+    c.y = 2 * f
+    assert_equal(c.y, 2 * f)
     c.x = (y, x)
     assert_equal(c.x, (y, x))
 
 
+def test_LeastSquares_3D():
+    def model(xyz, a, b):
+        x, y, z = xyz
+        return a * x + b * y + a * b * z
+
+    x = np.array([1.0, 2.0, 3.0, 4.0])
+    y = np.array([4.0, 5.0, 6.0, 7.0])
+    z = np.array([7.0, 8.0, 9.0, 10.0])
+
+    f = model((x, y, z), 1.5, 0.2)
+    fe = 1.5
+
+    c = LeastSquares((x, y, z), f, fe, model, grad=numerical_model_gradient(model))
+    assert c.ndata == 4
+
+    ref = numerical_cost_gradient(c)
+    assert_allclose(c.grad(1, 2), ref(1, 2))
+
+    assert_equal(c.x, (x, y, z))
+    assert_equal(c.y, f)
+    assert_equal(c.yerror, fe)
+    assert_allclose(c(1.5, 0.2), 0.0)
+    assert_allclose(
+        c(2.5, 0.2), np.sum(((f - 2.5 * x - 0.2 * y - 2.5 * 0.2 * z) / fe) ** 2)
+    )
+    assert_allclose(
+        c(1.5, 1.2), np.sum(((f - 1.5 * x - 1.2 * y - 1.5 * 1.2 * z) / fe) ** 2)
+    )
+
+    c.y = 2 * z
+    assert_equal(c.y, 2 * z)
+    c.x = (y, x, z)
+    assert_equal(c.x, (y, x, z))
+
+
 def test_LeastSquares_bad_input():
     with pytest.raises(ValueError, match="shape mismatch"):
         LeastSquares([1, 2], [], [1], lambda x, a: 0)
@@ -1208,7 +1243,7 @@ def line(x, par):
 def test_LeastSquares_visualize_2D():
     pytest.importorskip("matplotlib")
 
-    c = LeastSquares([[1, 2]], [[2, 3]], 0.1, line)
+    c = LeastSquares([[1, 2], [2, 3]], [1, 2], 0.1, line)
 
     with pytest.raises(ValueError, match="not implemented for multi-dimensional"):
         c.visualize((1, 2))

tests/test_without_ipywidgets.py

@@ -6,6 +6,7 @@
 
 
 def test_interactive():
+    pytest.importorskip("matplotlib")
     import iminuit
 
     cost = LeastSquares([1.1, 2.2], [3.3, 4.4], 1, lambda x, a: a * x)