Restrict log plots to 4 magnitude levels by default

sarracen/render.py

@@ -313,15 +313,19 @@ def render(data: 'SarracenDataFrame',  # noqa: F821
     kwargs.setdefault("origin", 'lower')
     kwargs.setdefault("extent", [xlim[0], xlim[1], ylim[0], ylim[1]])
     if log_scale:
+        # By default, a log scale plot will only cover 4 levels of magnitude.
+        vminref = 10 ** (np.log10(kwargs.get("vmax", img.max())) - 4)
+        vmin = kwargs.get('vmin', max(vminref, img.min()))
+
         if symlog_scale:
             kwargs.setdefault("norm",
                               SymLogNorm(kwargs.pop("linthresh", 1e-9),
                                          linscale=kwargs.pop("linscale", 1.),
-                                         vmin=kwargs.get('vmin'),
+                                         vmin=vmin,
                                          vmax=kwargs.get('vmax')))
         else:
             kwargs.setdefault("norm", LogNorm(clip=True,
-                                              vmin=kwargs.get('vmin'),
+                                              vmin=vmin,
                                               vmax=kwargs.get('vmax')))
         kwargs.pop("vmin", None)
         kwargs.pop("vmax", None)

sarracen/tests/test_render.py

@@ -4,6 +4,8 @@
 from numpy.testing import assert_array_equal
 from pytest import mark
 
+import numpy as np
+
 from sarracen import SarracenDataFrame
 from sarracen import interpolate_2d, interpolate_2d_line
 from sarracen import interpolate_3d_proj, interpolate_3d_cross
@@ -342,3 +344,43 @@ def test_plot_bounds(backend):
         assert ax.get_xlim() == (3, 6)
         assert ax.get_ylim() == (1, 5)
         plt.close(fig)
+
+
+@mark.parametrize("backend", backends)
+def test_log_vmin(backend):
+    """
+    Verify that the vmin value of log scales are set properly.
+    """
+    data = {'x': [6, 3, 2], 'y': [5, 1, 2], 'z': [3, 4, 2], 'P': [1, 1, 1],
+            'h': [1, 1, 1], 'rho': [1e10, 1e7, 1e4], 'm': [1, 1, 1]}
+    sdf = SarracenDataFrame(data)
+    sdf.backend = backend
+
+    # With no bounds defined, the log scale should cover 4 orders of magnitude.
+    interpolate = interpolate_3d_proj(sdf, 'rho')
+    fig, ax = plt.subplots()
+    render(sdf, "rho", ax=ax, log_scale=True, x_pixels=20)
+
+    imgs = ax.get_images()
+    vmin, vmax = imgs[0].get_clim()
+
+    assert vmin == 10 ** (np.log10(interpolate.max()) - 4)
+    assert vmax == interpolate.max()
+
+    # With only vmax defined, the vmin value is scaled accordingly.
+    fig, ax = plt.subplots()
+    render(sdf, "rho", vmax=10, log_scale=True, x_pixels=20)
+    imgs = ax.get_images()
+    vmin, vmax = imgs[0].get_clim()
+
+    assert vmin == 1e-3
+    assert vmax == 10
+
+    # With both bounds defined, no adjustment is made.
+    fig, ax = plt.subplots()
+    render(sdf, "rho", vmin=5, vmax=10, log_scale=True, x_pixels=20)
+    imgs = ax.get_images()
+    vmin, vmax = imgs[0].get_clim()
+
+    assert vmin == 5
+    assert vmax == 10