Added histograms' comparison (and updated numpy.testing asserts to pytest.approx)

README.md

@@ -126,6 +126,7 @@ histograms can be plotted via any compatible library, such as [mplhep][].
   * `/=`: Divide by a scaler (not all storages) (`hist / scalar` supported too)
   * `.kind`: Either `bh.Kind.COUNT` or `bh.Kind.MEAN`, depending on storage
   * `.sum(flow=False)`: The total count of all bins
+  * `.compare(second_hist)`: Compare the histogram with another histogram
   * `.project(ax1, ax2, ...)`: Project down to listed axis (numbers). Can also reorder axes.
   * `.to_numpy(flow=False, view=False)`: Convert to a NumPy style tuple (with or without under/overflow bins)
   * `.view(flow=False)`: Get a view on the bin contents (with or without under/overflow bins)

src/boost_histogram/_internal/hist.py

@@ -1,6 +1,7 @@
 import collections.abc
 import copy
 import logging
+import re
 import threading
 import typing
 import warnings
@@ -23,6 +24,7 @@
 )
 
 import numpy as np
+from pytest import approx
 
 import boost_histogram
 from boost_histogram import _core
@@ -297,6 +299,46 @@ def ndim(self) -> int:
         """
         return self._hist.rank()  # type: ignore[no-any-return]
 
+    def compare(self, hist2):
+        if self.view().shape == approx(hist2.view().shape):
+            if self.view() == approx(hist2.view()):
+                if self.variances() == approx(hist2.variances()):
+                    if (
+                        re.search("(?<=storage=).*", str(self.view))[0].split("(")[0]
+                        == re.search("(?<=storage=).*", str(hist2.view))[0].split("(")[
+                            0
+                        ]
+                    ):
+                        if list(map(str, [i for i in self.axes])) == list(
+                            map(str, [i for i in hist2.axes])
+                        ):
+                            return True
+                        else:
+                            assert list(map(str, [i for i in self.axes])) == list(
+                                map(str, [i for i in hist2.axes])
+                            ), "The axes are not equal."
+                    else:
+                        assert (
+                            re.search("(?<=storage=).*", str(self.view))[0].split("(")[
+                                0
+                            ]
+                            == re.search("(?<=storage=).*", str(hist2.view))[0].split(
+                                "("
+                            )[0]
+                        ), "The storage type is not equal."
+                else:
+                    assert self.variances() == approx(
+                        hist2.variances()
+                    ), "The histogram contents are not equal."
+            else:
+                assert self.view() == approx(
+                    hist2.view()
+                ), "The histogram contents are not equal."
+        else:
+            assert self.view().shape == approx(
+                hist2.view().shape
+            ), "The histogram dimensions are not equal."
+
     def view(
         self, flow: bool = False
     ) -> Union["np.typing.NDArray[Any]", WeightedSumView, WeightedMeanView, MeanView]:

tests/test_axes_object.py

@@ -1,5 +1,6 @@
 import numpy as np
 import pytest
+from pytest import approx
 
 import boost_histogram as bh
 
@@ -39,11 +40,11 @@ def test_axes_centers(h):
     full_answers = np.mgrid[0.5:10, 0.5:5, 0.5:2]
 
     for i in range(3):
-        np.testing.assert_allclose(centers.broadcast()[i], full_answers[i])
-        np.testing.assert_allclose(centers[i], answers[i])
-        np.testing.assert_allclose(centers.T[i], answers[i].T)
-        np.testing.assert_allclose(centers.flatten()[i], answers[i].flatten())
-        np.testing.assert_allclose(h.axes[i].centers, answers[i].ravel())
+        assert centers.broadcast()[i] == approx(full_answers[i])
+        assert centers[i] == approx(answers[i])
+        assert centers.T[i] == approx(answers[i].T)
+        assert centers.flatten()[i] == approx(answers[i].flatten())
+        assert h.axes[i].centers == approx(answers[i].ravel())
 
 
 def test_axes_edges(h):
@@ -52,11 +53,11 @@ def test_axes_edges(h):
     full_answers = np.mgrid[0:11, 0:6, 0:3]
 
     for i in range(3):
-        np.testing.assert_allclose(edges.broadcast()[i], full_answers[i])
-        np.testing.assert_allclose(edges[i], answers[i])
-        np.testing.assert_allclose(edges.T[i], answers[i].T)
-        np.testing.assert_allclose(edges.ravel()[i], answers[i].ravel())
-        np.testing.assert_allclose(h.axes[i].edges, answers[i].ravel())
+        assert edges.broadcast()[i] == approx(full_answers[i])
+        assert edges[i] == approx(answers[i])
+        assert edges.T[i] == approx(answers[i].T)
+        assert edges.ravel()[i] == approx(answers[i].ravel())
+        assert h.axes[i].edges == approx(answers[i].ravel())
 
 
 def test_axes_widths(h):
@@ -65,11 +66,11 @@ def test_axes_widths(h):
     full_answers = np.mgrid[1:1:10j, 1:1:5j, 1:1:2j]
 
     for i in range(3):
-        np.testing.assert_allclose(widths.broadcast()[i], full_answers[i])
-        np.testing.assert_allclose(widths[i], answers[i])
-        np.testing.assert_allclose(widths.T[i], answers[i].T)
-        np.testing.assert_allclose(widths.ravel()[i], answers[i].ravel())
-        np.testing.assert_allclose(h.axes[i].widths, answers[i].ravel())
+        assert widths.broadcast()[i] == approx(full_answers[i])
+        assert widths[i] == approx(answers[i])
+        assert widths.T[i] == approx(answers[i].T)
+        assert widths.ravel()[i] == approx(answers[i].ravel())
+        assert h.axes[i].widths == approx(answers[i].ravel())
 
 
 def test_axis_misconstuct():

tests/test_numpy_interface.py

@@ -46,17 +46,17 @@ def test_histogram1d(a, opt):
     h1, e1 = np.histogram(v, **opt)
     h2, e2 = bh.numpy.histogram(v, **opt)
 
-    np.testing.assert_array_almost_equal(e1, e2)
-    np.testing.assert_array_equal(h1, h2)
+    assert e1 == approx(e2)
+    assert h1 == approx(h2)
 
     opt = copy.deepcopy(opt)
     opt["density"] = True
 
     h1, e1 = np.histogram(v, **opt)
     h2, e2 = bh.numpy.histogram(v, **opt)
 
-    np.testing.assert_array_almost_equal(e1, e2)
-    np.testing.assert_array_almost_equal(h1, h2)
+    assert e1 == approx(e2)
+    assert h1 == approx(h2)
 
 
 @pytest.mark.parametrize("a", inputs_1d)
@@ -71,11 +71,11 @@ def test_histogram1d_object(a, opt):
     bh_h2 = bh.numpy.histogram(v, **bh_opt)
     h2, e2 = bh_h2.to_numpy()
 
-    np.testing.assert_array_almost_equal(e1, e2)
-    np.testing.assert_array_equal(h1, h2)
+    assert e1 == approx(e2)
+    assert h1 == approx(h2)
 
     # Ensure reducible
-    assert bh_h2[:5].values() == pytest.approx(h1[:5])
+    assert bh_h2[:5].values() == approx(h1[:5])
 
     opt = copy.deepcopy(opt)
     opt["density"] = True
@@ -94,16 +94,16 @@ def test_histogram2d():
     h1, e1x, e1y = np.histogram2d(x, y)
     h2, e2x, e2y = bh.numpy.histogram2d(x, y)
 
-    np.testing.assert_array_almost_equal(e1x, e2x)
-    np.testing.assert_array_almost_equal(e1y, e2y)
-    np.testing.assert_array_equal(h1, h2)
+    assert e1x == approx(e2x)
+    assert e1y == approx(e2y)
+    assert h1 == approx(h2)
 
     h1, e1x, e1y = np.histogram2d(x, y, density=True)
     h2, e2x, e2y = bh.numpy.histogram2d(x, y, density=True)
 
-    np.testing.assert_array_almost_equal(e1x, e2x)
-    np.testing.assert_array_almost_equal(e1y, e2y)
-    np.testing.assert_array_almost_equal(h1, h2)
+    assert e1x == approx(e2x)
+    assert e1y == approx(e2y)
+    assert h1 == approx(h2)
 
 
 def test_histogram2d_object():
@@ -114,9 +114,9 @@ def test_histogram2d_object():
     bh_h2 = bh.numpy.histogram2d(x, y, histogram=bh.Histogram)
     h2, e2x, e2y = bh_h2.to_numpy()
 
-    np.testing.assert_array_almost_equal(e1x, e2x)
-    np.testing.assert_array_almost_equal(e1y, e2y)
-    np.testing.assert_array_equal(h1, h2)
+    assert e1x == approx(e2x)
+    assert e1y == approx(e2y)
+    assert h1 == approx(h2)
 
     with pytest.raises(KeyError):
         bh.numpy.histogram2d(x, y, density=True, histogram=bh.Histogram)
@@ -130,18 +130,18 @@ def test_histogramdd():
     h1, (e1x, e1y, e1z) = np.histogramdd([x, y, z])
     h2, (e2x, e2y, e2z) = bh.numpy.histogramdd([x, y, z])
 
-    np.testing.assert_array_almost_equal(e1x, e2x)
-    np.testing.assert_array_almost_equal(e1y, e2y)
-    np.testing.assert_array_almost_equal(e1z, e2z)
-    np.testing.assert_array_equal(h1, h2)
+    assert e1x == approx(e2x)
+    assert e1y == approx(e2y)
+    assert e1z == approx(e2z)
+    assert h1 == approx(h2)
 
     h1, (e1x, e1y, e1z) = np.histogramdd([x, y, z], density=True)
     h2, (e2x, e2y, e2z) = bh.numpy.histogramdd([x, y, z], density=True)
 
-    np.testing.assert_array_almost_equal(e1x, e2x)
-    np.testing.assert_array_almost_equal(e1y, e2y)
-    np.testing.assert_array_almost_equal(e1z, e2z)
-    np.testing.assert_array_almost_equal(h1, h2)
+    assert e1x == approx(e2x)
+    assert e1y == approx(e2y)
+    assert e1z == approx(e2z)
+    assert h1 == approx(h2)
 
 
 def test_histogramdd_object():
@@ -153,10 +153,10 @@ def test_histogramdd_object():
     bh_h2 = bh.numpy.histogramdd([x, y, z], histogram=bh.Histogram)
     h2, (e2x, e2y, e2z) = bh_h2.to_numpy(dd=True)
 
-    np.testing.assert_array_almost_equal(e1x, e2x)
-    np.testing.assert_array_almost_equal(e1y, e2y)
-    np.testing.assert_array_almost_equal(e1z, e2z)
-    np.testing.assert_array_equal(h1, h2)
+    assert e1x == approx(e2x)
+    assert e1y == approx(e2y)
+    assert e1z == approx(e2z)
+    assert h1 == approx(h2)
 
     with pytest.raises(KeyError):
         bh.numpy.histogramdd([x, y, z], density=True, histogram=bh.Histogram)

tests/test_pickle.py

@@ -6,7 +6,7 @@
 
 import numpy as np
 import pytest
-from numpy.testing import assert_almost_equal, assert_array_equal
+from pytest import approx
 
 import boost_histogram as bh
 
@@ -86,7 +86,7 @@ def test_axes(axis, args, opts, copy_fn):
     orig = axis(*args, **opts)
     new = copy_fn(orig)
     assert new == orig
-    np.testing.assert_array_equal(new.centers, orig.centers)
+    assert new.centers == approx(orig.centers)
 
 
 @pytest.mark.parametrize("axis,args,opts", axes_creations)
@@ -97,7 +97,7 @@ def test_metadata_str(axis, args, opts, copy_fn):
     assert new.metadata == orig.metadata
     new.metadata = orig.metadata
     assert new == orig
-    np.testing.assert_array_equal(new.centers, orig.centers)
+    assert new.centers == approx(orig.centers)
 
 
 # Special test: Deepcopy should change metadata id, copy should not
@@ -162,7 +162,7 @@ def test_storage(benchmark, copy_fn, storage, extra):
         hist.fill(x, weight=np.arange(2 * n + 4) + 1, sample=np.arange(2 * n + 4) + 1)
 
     new = benchmark(copy_fn, hist)
-    assert_array_equal(hist.view(True), new.view(True))
+    assert hist.view(True) == approx(new.view(True))
     assert new == hist
 
 
@@ -213,8 +213,8 @@ def test_pickle_transforms(mod, copy_fn):
     ax3 = bh.axis.Regular(100, 1, 100, transform=bh.axis.transform.log)
 
     assert ax1 == ax2
-    assert_array_equal(ax1.centers, ax2.centers)
-    assert_almost_equal(ax2.centers, ax3.centers, decimal=10)
+    assert ax1.centers == approx(ax2.centers)
+    assert ax2.centers == approx(ax3.centers)
 
 
 def test_hist_axes_reference(copy_fn):