Fix logscale limit updates (#383)

* test that limits get accurately updated by successive plots with logscale axes, adjusting to new data limits, see issue #381

* fix typo from PR #324

* bump version to 2.8.10

* update logscale plot limits to datalimits at the end, making use of `ax.dataLim`

README.rst

@@ -2,7 +2,7 @@
 anesthetic: nested sampling post-processing
 ===========================================
 :Authors: Will Handley and Lukas Hergt
-:Version: 2.8.9
+:Version: 2.8.10
 :Homepage: https://github.com/handley-lab/anesthetic
 :Documentation: http://anesthetic.readthedocs.io/
 

anesthetic/_version.py

@@ -1 +1 @@
-__version__ = '2.8.9'
+__version__ = '2.8.10'

anesthetic/plot.py

@@ -394,8 +394,12 @@ def _set_logticks(self):
                 if ax is not None:
                     if x in self._logx:
                         ax.xaxis.set_major_locator(LogLocator(numticks=3))
+                        if x != y:
+                            ax.set_xlim(ax.dataLim.intervalx)
                     if y in self._logy:
                         ax.yaxis.set_major_locator(LogLocator(numticks=3))
+                        if y != x:
+                            ax.set_ylim(ax.dataLim.intervaly)
 
     @staticmethod
     def _set_labels(axes, labels, **kwargs):

anesthetic/samples.py

@@ -356,7 +356,7 @@ def plot_2d(self, axes=None, *args, **kwargs):
                             if np.isinf(self[x]).any():
                                 warnings.warn(f"column {y} has inf values.")
                             selfxy = self[[x, y]]
-                            selfxy = self.replace([-np.inf, np.inf], np.nan)
+                            selfxy = selfxy.replace([-np.inf, np.inf], np.nan)
                             selfxy = selfxy.dropna(axis=0)
                             selfxy.plot(x, y, ax=ax, xlabel=xlabel,
                                         logx=x in logx, logy=y in logy,

tests/test_samples.py

@@ -526,20 +526,31 @@ def test_plot_logscale_2d(kind):
 def test_logscale_ticks():
     np.random.seed(42)
     ndim = 5
-    data = np.exp(10 * np.random.randn(200, ndim))
+    data1 = np.exp(10 * np.random.randn(200, ndim))
+    data2 = np.exp(10 * np.random.randn(200, ndim) - 50)
     params = [f'a{i}' for i in range(ndim)]
     fig, axes = make_2d_axes(params, logx=params, logy=params, upper=False)
-    samples = Samples(data, columns=params)
-    samples.plot_2d(axes)
-    for _, col in axes.iterrows():
-        for _, ax in col.items():
+    samples1 = Samples(data1, columns=params)
+    samples2 = Samples(data2, columns=params)
+    samples1.plot_2d(axes)
+    samples2.plot_2d(axes)
+    for y, col in axes.iterrows():
+        for x, ax in col.items():
             if ax is not None:
                 xlims = ax.get_xlim()
                 xticks = ax.get_xticks()
                 assert np.sum((xticks > xlims[0]) & (xticks < xlims[1])) > 1
                 ylims = ax.get_ylim()
                 yticks = ax.get_yticks()
                 assert np.sum((yticks > ylims[0]) & (yticks < ylims[1])) > 1
+                if x == y:
+                    data_min = ax.twin.dataLim.intervalx[0]
+                    data_max = ax.twin.dataLim.intervalx[1]
+                    assert xlims[0] == pytest.approx(data_min, rel=1e-14)
+                    assert xlims[1] == pytest.approx(data_max, rel=1e-14)
+                else:
+                    assert_array_equal(xlims, ax.dataLim.intervalx)
+                    assert_array_equal(ylims, ax.dataLim.intervaly)
 
 
 @pytest.mark.parametrize('k', ['hist_1d', 'hist'])