review

Author: dilpath

doc/examples/other_model_types.ipynb

@@ -100,17 +100,18 @@
     "## Model selection\n",
     "\n",
     "To perform model selection, we define three methods to:\n",
+    "\n",
     "1. calibrate a single model\n",
     "\n",
-    "This is where we convert a PEtab Select model into a statsmodels model, fit the model with statsmodels, then save the log-likelihood value in the PEtab Select model.\n",
+    "    This is where we convert a PEtab Select model into a statsmodels model, fit the model with statsmodels, then save the log-likelihood value in the PEtab Select model.\n",
     "\n",
     "2. perform a single iteration of a model selection method involving calibration of multiple models\n",
     "\n",
-    "This is generic code that executes the required PEtab Select commands.\n",
+    "    This is generic code that executes the required PEtab Select commands.\n",
     "\n",
     "3. perform a full model selection run, involving all iterations of a model selection method\n",
     "\n",
-    "This loads the PEtab Select problem from disk and performs all iterations of a model selection method.\n",
+    "    This loads the PEtab Select problem from disk and performs all iterations of a model selection method.\n",
     "\n",
     "In this case, the PEtab Select problem is setup to perform backward selection. See the files in `other_model_types_problem/petab_select`."
    ]
@@ -259,15 +260,18 @@
    "outputs": [],
    "source": [
     "# Plot the history of model selection iterations.\n",
+    "import matplotlib.pyplot as plt\n",
     "\n",
     "draw_networkx_kwargs = {\n",
     "    \"arrowstyle\": \"-|>\",\n",
     "    \"node_shape\": \"s\",\n",
-    "    \"node_size\": 1000,\n",
     "    \"edgecolors\": \"k\",\n",
     "}\n",
+    "fig, ax = plt.subplots(figsize=(20, 20))\n",
     "petab_select.plot.graph_iteration_layers(\n",
-    "    plot_data=plot_data, draw_networkx_kwargs=draw_networkx_kwargs\n",
+    "    plot_data=plot_data,\n",
+    "    draw_networkx_kwargs=draw_networkx_kwargs,\n",
+    "    ax=ax,\n",
     ");"
    ]
   }

petab_select/plot.py

@@ -28,6 +28,10 @@
 """The font size of axis tick labels."""
 DEFAULT_NODE_COLOR = "darkgrey"
 """The default color of nodes in graph plots."""
+FONT_HEIGHT_WIDTH_RATIO = 2
+"""The ratio of the font height to font width. Used for graph node sizes."""
+NODE_SIZE_LABEL_SIZE_RATIO = 500
+"""The ratio of node size to label size. Used for graph node sizes."""
 
 
 __all__ = [
@@ -510,6 +514,22 @@ def scatter_criterion_vs_n_estimated(
     return ax
 
 
+def labels_to_node_sizes(G: nx.Graph) -> list[float]:
+    """Compute reasonable node sizes to scale with node label sizes."""
+    node_sizes = []
+    for label in G.nodes:
+        height = 1
+        width = len(label)
+        if "\n" in label:
+            height = len(label.split("\n"))
+            width = len(sorted(label.split("\n"), key=lambda x: len(x))[-1])
+        label_size = (
+            width if width * FONT_HEIGHT_WIDTH_RATIO > height else height
+        )
+        node_sizes.append(label_size * NODE_SIZE_LABEL_SIZE_RATIO)
+    return node_sizes
+
+
 def graph_iteration_layers(
     plot_data: PlotData,
     ax: plt.Axes = None,
@@ -541,7 +561,6 @@ def graph_iteration_layers(
     default_draw_networkx_kwargs = {
         "arrowstyle": "-|>",
         "node_shape": "s",
-        "node_size": 250,
         "edgecolors": "k",
     }
     if draw_networkx_kwargs is None:
@@ -610,6 +629,11 @@ def graph_iteration_layers(
     # Apply custom labels. Need `copy=True` to preserve node ordering.
     G = nx.relabel_nodes(G, mapping=plot_data.labels, copy=True)
 
+    draw_networkx_kwargs["node_size"] = draw_networkx_kwargs.get(
+        "node_size",
+        labels_to_node_sizes(G=G),
+    )
+
     nx.draw_networkx(
         G, pos, ax=ax, node_color=node_colors, **draw_networkx_kwargs
     )