Fix histplot auto line width with log scale or categorical y axis (#2522

)

* Fix histplot auto linewidth on log scaled axis

Fixes #2513

* Fix auto linewidth for categorical histograms on the y axis

Fixes #2523

(cherry picked from commit 0f538fb)

doc/releases/v0.12.0.txt

@@ -2,6 +2,7 @@ v0.12.0 (Unreleased)
 --------------------
 
 - |Fix| |Enhancement| Improved robustness to missing data, including additional support for the `pd.NA` type (:pr:`2417).
+
 - |API| |Feature| |Enhancement| TODO (Flesh this out further). Increased flexibility of what can be shown by the internally-calculated errorbars (:pr:2407).
 
 - |Fix| |Enhancement| Improved robustness to missing data, including additional support for the `pd.NA` type (:pr:`2417`).
@@ -29,8 +30,11 @@ v0.12.0 (Unreleased)
 - |Fix| In :func:`histplot` and :func:`kdeplot`, fixed a bug where the `multiple` was ignored when `hue` was provided as a vector without a name (:pr:`2462`).
 
  - |Defaults| In :func:`displot`, the default alpha value now adjusts to a provided `multiple` parameter even when `hue` is not assigned (:pr:`2462`).
+
 - |Fix| In :func:`histplot`, fixed a bug where using `shrink` with non-discrete bins shifted bar positions inaccurately (:pr:`2477`).
 
+- |Fix| In :func:`histplot`, fixed two bugs where automatically computed edge widths were too thick for log-scaled histograms and categorical histograms on the y axis (:pr:2522`).
+
 - |Fix| In :func:`displot`, fixed a bug where `common_norm` was ignored when `kind="hist"` and faceting was used without assigning `hue` (:pr:`2468`).
 
 - |Defaults| In :func:`displot`, the default alpha value now adjusts to a provided `multiple` parameter even when `hue` is not assigned (:pr:`2462`).

seaborn/distributions.py

@@ -658,15 +658,18 @@ def plot_univariate_histogram(
                 ax.autoscale_view()
 
                 # We will base everything on the minimum bin width
-                hist_metadata = [h.index.to_frame() for _, h in histograms.items()]
-                binwidth = min([
-                    h["widths"].min() for h in hist_metadata
-                ])
-
-                # Convert binwidtj from data coordinates to pixels
-                pts_x, pts_y = 72 / ax.figure.dpi * (
-                    ax.transData.transform([binwidth, binwidth])
-                    - ax.transData.transform([0, 0])
+                hist_metadata = pd.concat([
+                    # Use .items for generality over dict or df
+                    h.index.to_frame() for _, h in histograms.items()
+                ]).reset_index(drop=True)
+                thin_bar_idx = hist_metadata["widths"].idxmin()
+                binwidth = hist_metadata.loc[thin_bar_idx, "widths"]
+                left_edge = hist_metadata.loc[thin_bar_idx, "edges"]
+
+                # Convert binwidth from data coordinates to pixels
+                pts_x, pts_y = 72 / ax.figure.dpi * abs(
+                    ax.transData.transform([left_edge + binwidth] * 2)
+                    - ax.transData.transform([left_edge] * 2)
                 )
                 if self.data_variable == "x":
                     binwidth_points = pts_x

seaborn/tests/test_distributions.py

@@ -886,7 +886,7 @@ def test_log_scale(self, rng):
         for c1, c2 in zip(ax1.collections, ax2.collections):
             assert_array_equal(c1.get_segments(), c2.get_segments())
 
-    def test_bandwiddth(self, rng):
+    def test_bandwidth(self, rng):
 
         n = 100
         x, y = rng.multivariate_normal([0, 0], [(.2, .5), (.5, 2)], n).T
@@ -1591,6 +1591,18 @@ def test_auto_linewidth(self, flat_series, fill):
         histplot(flat_series, **kws, bins=30, ax=ax2)
         assert get_lw(ax1) > get_lw(ax2)
 
+        f, ax1 = plt.subplots(figsize=(4, 5))
+        f, ax2 = plt.subplots(figsize=(4, 5))
+        histplot(flat_series, **kws, bins=30, ax=ax1)
+        histplot(10 ** flat_series, **kws, bins=30, log_scale=True, ax=ax2)
+        assert get_lw(ax1) == pytest.approx(get_lw(ax2))
+
+        f, ax1 = plt.subplots(figsize=(4, 5))
+        f, ax2 = plt.subplots(figsize=(4, 5))
+        histplot(y=[0, 1, 1], **kws, discrete=True, ax=ax1)
+        histplot(y=["a", "b", "b"], **kws, ax=ax2)
+        assert get_lw(ax1) == pytest.approx(get_lw(ax2))
+
     def test_bar_kwargs(self, flat_series):
 
         lw = 2