Cover full extent of data in Band/Range when not given min/max explic…

…itly (#3056)

* Cover full extent of data in Range when not provided with min/max variables

* Cover full extent of data in Band when not provided with min/max variables

* Add example for Band and update release notes

doc/_docstrings/objects.Band.ipynb

@@ -13,7 +13,7 @@
    "source": [
     "import seaborn.objects as so\n",
     "from seaborn import load_dataset\n",
-    "fmri = load_dataset(\"fmri\")\n",
+    "fmri = load_dataset(\"fmri\").query(\"region == 'parietal'\")\n",
     "seaice = (\n",
     "    load_dataset(\"seaice\")\n",
     "    .assign(\n",
@@ -22,7 +22,7 @@
     "    )\n",
     "    .query(\"Year >= 1980\")\n",
     "    .astype({\"Year\": str})\n",
-    "    .pivot(\"Day\", \"Year\", \"Extent\")\n",
+    "    .pivot(index=\"Day\", columns=\"Year\", values=\"Extent\")\n",
     "    .filter([\"1980\", \"2019\"])\n",
     "    .dropna()\n",
     "    .reset_index()\n",
@@ -90,8 +90,32 @@
   },
   {
    "cell_type": "raw",
-   "id": "4e817cdd-09a3-4cf6-8602-e9665607bfe1",
+   "id": "9f0c82bf-3457-4ac5-ba48-8930bac03d75",
    "metadata": {},
+   "source": [
+    "When min/max values are not explicitly assigned or added in a transform, the band will cover the full extent of the data:"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "309f578e-da3d-4dc5-b6ac-a354321334c8",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "(\n",
+    "    so.Plot(fmri, x=\"timepoint\", y=\"signal\", color=\"event\")\n",
+    "    .add(so.Line(linewidth=.5), group=\"subject\")\n",
+    "    .add(so.Band())\n",
+    ")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "4330a3cd-63fe-470a-8e83-09e9606643b5",
+   "metadata": {},
+   "outputs": [],
    "source": []
   }
  ],

doc/_docstrings/objects.Range.ipynb

@@ -57,7 +57,7 @@
     "    so.Plot(penguins, x=\"sex\", y=\"body_mass_g\", linestyle=\"species\")\n",
     "    .facet(\"species\")\n",
     "    .add(so.Line(marker=\"o\"), so.Agg())\n",
-    "    .add(so.Range(), so.Est(errorbar=\"pi\"))\n",
+    "    .add(so.Range(), so.Est(errorbar=\"sd\"))\n",
     ")"
    ]
   },
@@ -78,21 +78,39 @@
    "source": [
     "(\n",
     "    penguins\n",
-    "    .rename_axis(\"penguin\")\n",
-    "    .pipe(so.Plot, ymin=\"bill_depth_mm\", ymax=\"bill_length_mm\", x=\"penguin\")\n",
-    "    .add(so.Range(), color=\"island\", linewidth=\"body_mass_g\")\n",
-    "    .scale(x=so.Continuous().tick(count=0), linewidth=(.5, 1.5))\n",
-    "    .facet(row=\"species\", col=\"sex\")\n",
-    "    .layout(size=(8, 4))\n",
-    "    .share(x=False)\n",
-    "    .label(x=\"\", y=\"Size (mm)\")\n",
+    "    .rename_axis(index=\"penguin\")\n",
+    "    .pipe(so.Plot, x=\"penguin\", ymin=\"bill_depth_mm\", ymax=\"bill_length_mm\")\n",
+    "    .add(so.Range(), color=\"island\")\n",
+    ")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "2191bec6-a02e-48e0-b92c-69c38826049d",
+   "metadata": {},
+   "source": [
+    "When `min`/`max` variables are neither computed as part of a transform or explicitly assigned, the range will cover the full extent of the data at each unique observation on the orient axis:"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "63c6352e-4ef5-4cff-940e-35fa5804b2c7",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "(\n",
+    "    so.Plot(penguins, x=\"sex\", y=\"body_mass_g\")\n",
+    "    .facet(\"species\")\n",
+    "    .add(so.Dots(pointsize=6))\n",
+    "    .add(so.Range(linewidth=2))\n",
     ")"
    ]
   },
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "08751ee7-d0a0-4e70-92b4-c1b38ea28890",
+   "id": "c215deb1-e510-4631-b999-737f5f41cae2",
    "metadata": {},
    "outputs": [],
    "source": []

doc/whatsnew/v0.12.1.rst

@@ -4,6 +4,8 @@ v0.12.1 (Unreleased)
 
 - |Feature| Added the :class:`objects.Text` mark (:pr:`3051`).
 
+- |Feature| The :class:`Band` and :class:`Range` marks will now cover the full extent of the data if `min` / `max` variables are not explicitly assigned or added in a transform (:pr:`3056`).
+
 - |Fix| Make :class:`objects.PolyFit` robust to missing data (:pr:`3010`).
 
 - |Fix| Fixed a bug that caused an exception when more than two layers with the same mappings were added (:pr:`3055`).

seaborn/_marks/area.py

@@ -162,4 +162,9 @@ class Band(AreaBase, Mark):
     def _standardize_coordinate_parameters(self, data, orient):
         # dv = {"x": "y", "y": "x"}[orient]
         # TODO assert that all(ymax >= ymin)?
+        # TODO what if only one exist?
+        other = {"x": "y", "y": "x"}[orient]
+        if not set(data.columns) & {f"{other}min", f"{other}max"}:
+            agg = {f"{other}min": (other, "min"), f"{other}max": (other, "max")}
+            data = data.groupby(orient).agg(**agg).reset_index()
         return data

seaborn/_marks/line.py

@@ -204,8 +204,9 @@ def _plot(self, split_gen, scales, orient):
             # Handle datalim update manually
             # https://github.com/matplotlib/matplotlib/issues/23129
             ax.add_collection(lines, autolim=False)
-            xy = np.concatenate(ax_data["segments"])
-            ax.update_datalim(xy)
+            if ax_data["segments"]:
+                xy = np.concatenate(ax_data["segments"])
+                ax.update_datalim(xy)
 
     def _legend_artist(self, variables, value, scales):
 
@@ -270,9 +271,16 @@ def _setup_lines(self, split_gen, scales, orient):
                     "linestyles": [],
                 }
 
+            # TODO better checks on what variables we have
+
             vals = resolve_properties(self, keys, scales)
             vals["color"] = resolve_color(self, keys, scales=scales)
 
+            # TODO what if only one exist?
+            if not set(data.columns) & {f"{other}min", f"{other}max"}:
+                agg = {f"{other}min": (other, "min"), f"{other}max": (other, "max")}
+                data = data.groupby(orient).agg(**agg).reset_index()
+
             cols = [orient, f"{other}min", f"{other}max"]
             data = data[cols].melt(orient, value_name=other)[["x", "y"]]
             segments = [d.to_numpy() for _, d in data.groupby(orient)]

tests/_marks/test_area.py

@@ -8,7 +8,7 @@
 from seaborn._marks.area import Area, Band
 
 
-class TestAreaMarks:
+class TestArea:
 
     def test_single_defaults(self):
 
@@ -97,7 +97,10 @@ def test_unfilled(self):
         poly = ax.patches[0]
         assert poly.get_facecolor() == to_rgba(c, 0)
 
-    def test_band(self):
+
+class TestBand:
+
+    def test_range(self):
 
         x, ymin, ymax = [1, 2, 4], [2, 1, 4], [3, 3, 5]
         p = Plot(x=x, ymin=ymin, ymax=ymax).add(Band()).plot()
@@ -109,3 +112,17 @@ def test_band(self):
 
         expected_y = [2, 1, 4, 5, 3, 3, 2]
         assert_array_equal(verts[1], expected_y)
+
+    def test_auto_range(self):
+
+        x = [1, 1, 2, 2, 2]
+        y = [1, 2, 3, 4, 5]
+        p = Plot(x=x, y=y).add(Band()).plot()
+        ax = p._figure.axes[0]
+        verts = ax.patches[0].get_path().vertices.T
+
+        expected_x = [1, 2, 2, 1, 1]
+        assert_array_equal(verts[0], expected_x)
+
+        expected_y = [1, 3, 5, 2, 1]
+        assert_array_equal(verts[1], expected_y)

tests/_marks/test_line.py

@@ -246,6 +246,15 @@ def test_xy_data(self):
         assert_array_equal(verts[0], [2, 5])
         assert_array_equal(verts[1], [3, 4])
 
+    def test_single_orient_value(self):
+
+        x = [1, 1, 1]
+        y = [1, 2, 3]
+        p = Plot(x, y).add(Lines()).plot()
+        lines, = p._figure.axes[0].collections
+        paths, = lines.get_paths()
+        assert paths.vertices.shape == (0, 2)
+
 
 class TestRange:
 
@@ -263,6 +272,17 @@ def test_xy_data(self):
             assert_array_equal(verts[0], [x[i], x[i]])
             assert_array_equal(verts[1], [ymin[i], ymax[i]])
 
+    def test_auto_range(self):
+
+        x = [1, 1, 2, 2, 2]
+        y = [1, 2, 3, 4, 5]
+
+        p = Plot(x=x, y=y).add(Range()).plot()
+        lines, = p._figure.axes[0].collections
+        paths = lines.get_paths()
+        assert_array_equal(paths[0].vertices, [(1, 1), (1, 2)])
+        assert_array_equal(paths[1].vertices, [(2, 3), (2, 5)])
+
     def test_mapped_color(self):
 
         x = [1, 2, 1, 2]