Delegate hue in PairGrid to plotting functions (#2234)

* Add provisional support for delegating hue in PairGrid * Use histplot on pairplot diagonal * Update Pairgrid tests * Return self from map_diag * Improve test coverage * Convert PairGrid docstring to notebook and update * Fix test * Improve support for legends and markers with new plots * Make color/label injection optional * Update markers test * More flexibility in PairGrid * Convert pairplot API examples to notebook * Add public access to Grid legend object * Fix iterative plot_bivariate and improve test coverage * Don't cast diagonal data to array (fixes #1663)
mwaskom · Aug 31, 2020 · d40c0b0 · d40c0b0
1 parent 954c018
commit d40c0b0
Show file tree

Hide file tree

Showing 6 changed files with 747 additions and 253 deletions.
diff --git a/doc/docstrings/PairGrid.ipynb b/doc/docstrings/PairGrid.ipynb
@@ -0,0 +1,271 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "tags": [
+     "hide"
+    ]
+   },
+   "outputs": [],
+   "source": [
+    "import seaborn as sns; sns.set()\n",
+    "import matplotlib.pyplot as plt"
+   ]
+  },
+  {
+   "cell_type": "raw",
+   "metadata": {},
+   "source": [
+    "Calling the constructor sets up a blank grid of subplots with each row and one column corresponding to a numeric variable in the dataset:"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "penguins = sns.load_dataset(\"penguins\")\n",
+    "g = sns.PairGrid(penguins)"
+   ]
+  },
+  {
+   "cell_type": "raw",
+   "metadata": {},
+   "source": [
+    "Passing a bivariate function to :meth:`PairGrid.map` will draw a bivariate plot on every axes:"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "g = sns.PairGrid(penguins)\n",
+    "g.map(sns.scatterplot)"
+   ]
+  },
+  {
+   "cell_type": "raw",
+   "metadata": {},
+   "source": [
+    "Passing separate functions to :meth:`PairGrid.map_diag` and :meth:`PairGrid.map_offdiag` will show each variable's marginal distribution on the diagonal:"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "g = sns.PairGrid(penguins)\n",
+    "g.map_diag(sns.histplot)\n",
+    "g.map_offdiag(sns.scatterplot)"
+   ]
+  },
+  {
+   "cell_type": "raw",
+   "metadata": {},
+   "source": [
+    "It's also possible to use different functions on the upper and lower triangles of the plot (which are otherwise redundant):"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "g = sns.PairGrid(penguins, diag_sharey=False)\n",
+    "g.map_upper(sns.scatterplot)\n",
+    "g.map_lower(sns.kdeplot)\n",
+    "g.map_diag(sns.kdeplot)"
+   ]
+  },
+  {
+   "cell_type": "raw",
+   "metadata": {},
+   "source": [
+    "Or to avoid the redundancy altogether:"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "g = sns.PairGrid(penguins, diag_sharey=False, corner=True)\n",
+    "g.map_lower(sns.scatterplot)\n",
+    "g.map_diag(sns.kdeplot)"
+   ]
+  },
+  {
+   "cell_type": "raw",
+   "metadata": {},
+   "source": [
+    "The :class:`PairGrid` constructor accepts a ``hue`` variable. This variable is passed directly to functions that understand it:"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "g = sns.PairGrid(penguins, hue=\"species\")\n",
+    "g.map_diag(sns.histplot)\n",
+    "g.map_offdiag(sns.scatterplot)\n",
+    "g.add_legend()"
+   ]
+  },
+  {
+   "cell_type": "raw",
+   "metadata": {},
+   "source": [
+    "But you can also pass matplotlib functions, in which case a groupby is performed internally and a separate plot is drawn for each level:"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "g = sns.PairGrid(penguins, hue=\"species\")\n",
+    "g.map_diag(plt.hist)\n",
+    "g.map_offdiag(plt.scatter)\n",
+    "g.add_legend()"
+   ]
+  },
+  {
+   "cell_type": "raw",
+   "metadata": {},
+   "source": [
+    "Additional semantic variables can be assigned by passing data vectors directly while mapping the function:"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "g = sns.PairGrid(penguins, hue=\"species\")\n",
+    "g.map_diag(sns.histplot)\n",
+    "g.map_offdiag(sns.scatterplot, size=penguins[\"sex\"])\n",
+    "g.add_legend(title=\"\", adjust_subtitles=True)"
+   ]
+  },
+  {
+   "cell_type": "raw",
+   "metadata": {},
+   "source": [
+    "When using seaborn functions that can implement a numeric hue mapping, you will want to disable mapping of the variable on the diagonal axes. Note that the ``hue`` variable is excluded from the list of variables shown by default:"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "g = sns.PairGrid(penguins, hue=\"body_mass_g\")\n",
+    "g.map_diag(sns.histplot, hue=None, color=\".3\")\n",
+    "g.map_offdiag(sns.scatterplot)\n",
+    "g.add_legend()"
+   ]
+  },
+  {
+   "cell_type": "raw",
+   "metadata": {},
+   "source": [
+    "The ``vars`` parameter can be used to control exactly which variables are used:"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "variables = [\"body_mass_g\", \"bill_length_mm\", \"flipper_length_mm\"]\n",
+    "g = sns.PairGrid(penguins, hue=\"body_mass_g\", vars=variables)\n",
+    "g.map_diag(sns.histplot, hue=None, color=\".3\")\n",
+    "g.map_offdiag(sns.scatterplot)\n",
+    "g.add_legend()"
+   ]
+  },
+  {
+   "cell_type": "raw",
+   "metadata": {},
+   "source": [
+    "The plot need not be square: separate variables can be used to define the rows and columns:"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "x_vars = [\"body_mass_g\", \"bill_length_mm\", \"bill_depth_mm\", \"flipper_length_mm\"]\n",
+    "y_vars = [\"body_mass_g\"]\n",
+    "g = sns.PairGrid(penguins, hue=\"species\", x_vars=x_vars, y_vars=y_vars)\n",
+    "g.map_diag(sns.histplot, color=\".3\")\n",
+    "g.map_offdiag(sns.scatterplot)\n",
+    "g.add_legend()"
+   ]
+  },
+  {
+   "cell_type": "raw",
+   "metadata": {},
+   "source": [
+    "It can be useful to explore different approaches to resolving multiple distributions on the diagonal axes:"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "g = sns.PairGrid(penguins, hue=\"species\")\n",
+    "g.map_diag(sns.histplot, multiple=\"stack\", element=\"step\")\n",
+    "g.map_offdiag(sns.scatterplot)\n",
+    "g.add_legend()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": []
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "seaborn-py38-latest",
+   "language": "python",
+   "name": "seaborn-py38-latest"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.8.5"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 4
+}