chore: update Numpy and the matching syntax (#285)

* chore: update numpy in .pre-commit-config.yaml

* fix: update syntax matching Numpy 1.21 for mypy

* chore: update Numpy to 1.14.5 in setup.cfg

* fix: change np.diag to np.diagonal

.pre-commit-config.yaml

@@ -57,7 +57,7 @@ repos:
   - id: mypy
     files: ^src
     args: []
-    additional_dependencies: ["numpy==1.20.*", "matplotlib>=3.3", "boost-histogram~=1.0.1", "uhi~=0.3.0"]
+    additional_dependencies: ["numpy==1.21.*", "matplotlib>=3.3", "boost-histogram~=1.0.1", "uhi~=0.3.0"]
 
 - repo: https://github.com/mgedmin/check-manifest
   rev: "0.46"

setup.cfg

@@ -43,7 +43,7 @@ packages = find:
 install_requires =
     boost-histogram~=1.1.0
     histoprint>=1.6
-    numpy>=1.13.3
+    numpy>=1.14.5
     typing_extensions;python_version<"3.8"
 python_requires = >=3.7
 include_package_data = True

src/hist/basehist.py

@@ -63,7 +63,7 @@ def __init__(
         *args: AxisProtocol | Storage | str | tuple[int, float, float],
         storage: Storage | str | None = None,
         metadata: Any = None,
-        data: np.ndarray | None = None,
+        data: np.typing.NDArray[Any] | None = None,
     ) -> None:
         """
         Initialize BaseHist object. Axis params can contain the names.
@@ -355,12 +355,12 @@ def profile(self: T, axis: int | str) -> T:
         retval[...] = np.stack([count, new_values, count * new_variances], axis=-1)
         return retval
 
-    def density(self) -> np.ndarray:
+    def density(self) -> np.typing.NDArray[Any]:
         """
         Density NumPy array.
         """
         total = np.sum(self.values()) * functools.reduce(operator.mul, self.axes.widths)
-        dens: np.ndarray = self.values() / np.where(total > 0, total, 1)
+        dens: np.typing.NDArray[Any] = self.values() / np.where(total > 0, total, 1)
         return dens
 
     def show(self, **kwargs: Any) -> Any:
@@ -446,7 +446,9 @@ def plot2d_full(
 
     def plot_ratio(
         self,
-        other: hist.BaseHist | Callable[[np.ndarray], np.ndarray] | str,
+        other: hist.BaseHist
+        | Callable[[np.typing.NDArray[Any]], np.typing.NDArray[Any]]
+        | str,
         *,
         ax_dict: dict[str, matplotlib.axes.Axes] | None = None,
         **kwargs: Any,
@@ -466,7 +468,7 @@ def plot_ratio(
 
     def plot_pull(
         self,
-        func: Callable[[np.ndarray], np.ndarray] | str,
+        func: Callable[[np.typing.NDArray[Any]], np.typing.NDArray[Any]] | str,
         *,
         ax_dict: dict[str, matplotlib.axes.Axes] | None = None,
         **kwargs: Any,

src/hist/intervals.py

@@ -25,10 +25,10 @@ def __dir__() -> tuple[str, ...]:
 
 
 def poisson_interval(
-    values: np.ndarray,
-    variances: np.ndarray | None = None,
+    values: np.typing.NDArray[Any],
+    variances: np.typing.NDArray[Any] | None = None,
     coverage: float | None = None,
-) -> np.ndarray:
+) -> np.typing.NDArray[Any]:
     r"""
     The Frequentist coverage interval for Poisson-distributed observations.
 
@@ -67,7 +67,7 @@ def poisson_interval(
         scale = np.ones_like(values)
         mask = np.isfinite(values) & (values != 0)
         np.divide(variances, values, out=scale, where=mask)
-        counts = values / scale
+        counts: np.typing.NDArray[Any] = values / scale
         interval_min = scale * stats.chi2.ppf((1 - coverage) / 2, 2 * counts) / 2.0
         interval_min[values == 0.0] = 0.0  # chi2.ppf produces NaN for values=0
         interval_max = (
@@ -79,8 +79,10 @@ def poisson_interval(
 
 
 def clopper_pearson_interval(
-    num: np.ndarray, denom: np.ndarray, coverage: float | None = None
-) -> np.ndarray:
+    num: np.typing.NDArray[Any],
+    denom: np.typing.NDArray[Any],
+    coverage: float | None = None,
+) -> np.typing.NDArray[Any]:
     r"""
     Compute the Clopper-Pearson coverage interval for a binomial distribution.
     c.f. http://en.wikipedia.org/wiki/Binomial_proportion_confidence_interval
@@ -112,8 +114,8 @@ def clopper_pearson_interval(
 
 
 def ratio_uncertainty(
-    num: np.ndarray,
-    denom: np.ndarray,
+    num: np.typing.NDArray[Any],
+    denom: np.typing.NDArray[Any],
     uncertainty_type: Literal["poisson", "poisson-ratio", "efficiency"] = "poisson",
 ) -> Any:
     r"""
@@ -155,7 +157,7 @@ def ratio_uncertainty(
         # Details: see https://github.com/scikit-hep/hist/issues/279
         p_lim = clopper_pearson_interval(num, num + denom)
         with np.errstate(divide="ignore", invalid="ignore"):
-            r_lim = p_lim / (1 - p_lim)
+            r_lim: np.typing.NDArray[Any] = p_lim / (1 - p_lim)
             ratio_uncert = np.abs(r_lim - ratio)
     elif uncertainty_type == "efficiency":
         ratio_uncert = np.abs(clopper_pearson_interval(num, denom) - ratio)

src/hist/plot.py

@@ -125,11 +125,11 @@ def _expr_to_lambda(expr: str) -> Callable[..., Any]:
 
 def _curve_fit_wrapper(
     func: Callable[..., Any],
-    xdata: np.ndarray,
-    ydata: np.ndarray,
-    yerr: np.ndarray,
+    xdata: np.typing.NDArray[Any],
+    ydata: np.typing.NDArray[Any],
+    yerr: np.typing.NDArray[Any],
     likelihood: bool = False,
-) -> tuple[tuple[float, ...], np.ndarray]:
+) -> tuple[tuple[float, ...], np.typing.NDArray[Any]]:
     """
     Wrapper around `scipy.optimize.curve_fit`. Initial parameters (`p0`)
     can be set in the function definition with defaults for kwargs
@@ -156,7 +156,7 @@ def _curve_fit_wrapper(
         from iminuit import Minuit
         from scipy.special import gammaln
 
-        def fnll(v: Iterable[np.ndarray]) -> float:
+        def fnll(v: Iterable[np.typing.NDArray[Any]]) -> float:
             ypred = func(xdata, *v)
             if (ypred <= 0.0).any():
                 return 1e6
@@ -261,7 +261,7 @@ def plot2d_full(
 
 def _construct_gaussian_callable(
     __hist: hist.BaseHist,
-) -> Callable[[np.ndarray], np.ndarray]:
+) -> Callable[[np.typing.NDArray[Any]], np.typing.NDArray[Any]]:
     x_values = __hist.axes[0].centers
     hist_values = __hist.values()
 
@@ -272,13 +272,13 @@ def _construct_gaussian_callable(
 
     # gauss is a closure that will get evaluated in _fit_callable_to_hist
     def gauss(
-        x: np.ndarray,
+        x: np.typing.NDArray[Any],
         constant: float = constant,
         mean: float = mean,
         sigma: float = sigma,
-    ) -> np.ndarray:
-        # Note: Force np.ndarray type as numpy ufuncs have type "Any"
-        ret: np.ndarray = constant * np.exp(
+    ) -> np.typing.NDArray[Any]:
+        # Note: Force np.typing.NDArray[Any] type as numpy ufuncs have type "Any"
+        ret: np.typing.NDArray[Any] = constant * np.exp(
             -np.square(x - mean) / (2 * np.square(sigma))
         )
         return ret
@@ -287,10 +287,15 @@ def gauss(
 
 
 def _fit_callable_to_hist(
-    model: Callable[[np.ndarray], np.ndarray],
+    model: Callable[[np.typing.NDArray[Any]], np.typing.NDArray[Any]],
     histogram: hist.BaseHist,
     likelihood: bool = False,
-) -> tuple[np.ndarray, np.ndarray, np.ndarray, tuple[tuple[float, ...], np.ndarray]]:
+) -> tuple[
+    np.typing.NDArray[Any],
+    np.typing.NDArray[Any],
+    np.typing.NDArray[Any],
+    tuple[tuple[float, ...], np.typing.NDArray[Any]],
+]:
     """
     Fit a model, a callable function, to the histogram values.
     """
@@ -312,7 +317,7 @@ def _fit_callable_to_hist(
         n_samples = 100
         vopts = np.random.multivariate_normal(popt, pcov, n_samples)
         sampled_ydata = np.vstack([model(xdata, *vopt).T for vopt in vopts])
-        model_uncert = np.nanstd(sampled_ydata, axis=0)
+        model_uncert = np.nanstd(sampled_ydata, axis=0)  # type: ignore
     else:
         model_uncert = np.zeros_like(hist_uncert)
 
@@ -321,8 +326,8 @@ def _fit_callable_to_hist(
 
 def _plot_fit_result(
     __hist: hist.BaseHist,
-    model_values: np.ndarray,
-    model_uncert: np.ndarray,
+    model_values: np.typing.NDArray[Any],
+    model_uncert: np.typing.NDArray[Any],
     ax: matplotlib.axes.Axes,
     eb_kwargs: dict[str, Any],
     fp_kwargs: dict[str, Any],
@@ -361,8 +366,8 @@ def _plot_fit_result(
 
 def plot_ratio_array(
     __hist: hist.BaseHist,
-    ratio: np.ndarray,
-    ratio_uncert: np.ndarray,
+    ratio: np.typing.NDArray[Any],
+    ratio_uncert: np.typing.NDArray[Any],
     ax: matplotlib.axes.Axes,
     **kwargs: Any,
 ) -> RatioArtists:
@@ -435,7 +440,7 @@ def plot_ratio_array(
                 valid_ratios + ratio_uncert[1][valid_ratios_idx],
             ]
         )
-        max_delta = np.max(np.abs(extrema - central_value))
+        max_delta = np.amax(np.abs(extrema - central_value))
         ratio_extrema = np.abs(max_delta + central_value)
 
         _alpha = 2.0
@@ -453,7 +458,7 @@ def plot_ratio_array(
 
 def plot_pull_array(
     __hist: hist.BaseHist,
-    pulls: np.ndarray,
+    pulls: np.typing.NDArray[Any],
     ax: matplotlib.axes.Axes,
     bar_kwargs: dict[str, Any],
     pp_kwargs: dict[str, Any],
@@ -502,7 +507,9 @@ def plot_pull_array(
 
 def _plot_ratiolike(
     self: hist.BaseHist,
-    other: hist.BaseHist | Callable[[np.ndarray], np.ndarray] | str,
+    other: hist.BaseHist
+    | Callable[[np.typing.NDArray[Any]], np.typing.NDArray[Any]]
+    | str,
     likelihood: bool = False,
     *,
     ax_dict: dict[str, matplotlib.axes.Axes] | None = None,
@@ -608,7 +615,7 @@ def _plot_ratiolike(
         if fit_fmt is not None:
             parnames = list(inspect.signature(other).parameters)[1:]
             popt, pcov = bestfit_result
-            perr = np.sqrt(np.diag(pcov))
+            perr = np.sqrt(np.diagonal(pcov))
 
             fp_label = "Fit"
             for name, value, error in zip(parnames, popt, perr):
@@ -651,7 +658,9 @@ def _plot_ratiolike(
             )
 
         elif view == "pull":
-            pulls = (hist_values - compare_values) / hist_values_uncert
+            pulls: np.typing.NDArray[Any] = (
+                hist_values - compare_values
+            ) / hist_values_uncert
 
             pulls[np.isnan(pulls) | np.isinf(pulls)] = 0
 

src/hist/svgplots.py

@@ -1,6 +1,6 @@
 from __future__ import annotations
 
-from typing import Callable
+from typing import Any, Callable
 
 import numpy as np
 from boost_histogram.axis import Axis
@@ -70,8 +70,8 @@ def svg_hist_1d(h: hist.BaseHist) -> svg:
     (edges,) = h.axes.edges
     norm_edges = (edges - edges[0]) / (edges[-1] - edges[0])
     density = h.density()
-    max_dens = np.max(density) or 1
-    norm_vals = density / max_dens
+    max_dens = np.amax(density) or 1
+    norm_vals: np.typing.NDArray[Any] = density / max_dens
 
     arr = np.empty((2, len(norm_vals) * 2 + 2), dtype=float)
     arr[0, 0:-1:2] = arr[0, 1::2] = width * norm_edges
@@ -119,8 +119,8 @@ def svg_hist_1d_c(h: hist.BaseHist) -> svg:
     (edges,) = h.axes.edges
     norm_edges = (edges - edges[0]) / (edges[-1] - edges[0]) * np.pi * 2
     density = h.density()
-    max_dens = np.max(density) or 1
-    norm_vals = density / max_dens
+    max_dens = np.amax(density) or 1
+    norm_vals: np.typing.NDArray[Any] = density / max_dens
 
     arr = np.empty((2, len(norm_vals) * 2), dtype=float)
     arr[0, :-1:2] = arr[0, 1::2] = norm_edges[:-1]
@@ -153,8 +153,8 @@ def svg_hist_2d(h: hist.BaseHist) -> svg:
     ey = -(e1 - e1[0]) / (e1[-1] - e1[0]) * height
 
     density = h.density()
-    max_dens = np.max(density) or 1
-    norm_vals = density / max_dens
+    max_dens = np.amax(density) or 1
+    norm_vals: np.typing.NDArray[Any] = density / max_dens
 
     boxes = []
     for r, (up_edge, bottom_edge) in enumerate(zip(ey[:-1], ey[1:])):