Add material about escape hatches

intermediate/hierarchical_computation.ipynb

@@ -137,26 +137,28 @@
    "source": [
     "## Applying functions designed for `Dataset` with `map_over_datasets`\n",
     "\n",
-    "What if we wanted to convert the data to log-space? For a `Dataset` or `DataArray`, we could just use {py:func}`xarray.ufuncs.log`, but that does not support `DataTree` objects, yet:"
+    "What if we wanted to apply a element-wise function, for example to convert the data to log-space? For a `DataArray` we could just use {py:func}`numpy.log`, but this is not supported for `DataTree` objects:"
    ]
   },
   {
    "cell_type": "code",
    "execution_count": null,
    "id": "12",
-   "metadata": {},
+   "metadata": {
+    "tags": [
+     "raises-exception"
+    ]
+   },
    "outputs": [],
    "source": [
-    "xr.ufuncs.log(tree)"
+    "np.log(tree)"
    ]
   },
   {
    "cell_type": "markdown",
    "id": "13",
    "metadata": {},
    "source": [
-    "Note how the result is a empty `Dataset`?\n",
-    "\n",
     "To map a function to all nodes, we can use {py:func}`xarray.map_over_datasets` and {py:meth}`xarray.DataTree.map_over_datasets`: "
    ]
   },
@@ -203,8 +205,7 @@
    "id": "18",
    "metadata": {
     "tags": [
-     "raises-exception",
-     "hide-output"
+     "raises-exception"
     ]
    },
    "outputs": [],
@@ -235,6 +236,78 @@
     "\n",
     "tree.map_over_datasets(demean)"
    ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "21",
+   "metadata": {},
+   "source": [
+    "## Escape hatches\n",
+    "\n",
+    "For some more complex operations, it might make sense to work on {py:class}`xarray.Dataset` or {py:class}`xarray.DataArray` objects and reassemble the tree afterwards.\n",
+    "\n",
+    "Let's look at a new dataset:"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "22",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "precipitation = xr.tutorial.open_datatree(\"precipitation.nc4\").load()\n",
+    "precipitation"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "23",
+   "metadata": {},
+   "source": [
+    "Suppose we wanted to interpolate the observed precipitation to the modelled precipitation. We could use `map_over_datasets` for this, but we can also have a bit more control:"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "24",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "interpolated = xr.DataTree.from_dict(\n",
+    "    {\n",
+    "        \"/\": precipitation.ds,\n",
+    "        \"/observed\": precipitation[\"/observed\"].ds.interp(\n",
+    "            lat=precipitation[\"/reanalysis/lat\"],\n",
+    "            lon=precipitation[\"/reanalysis/lon\"],\n",
+    "        ),\n",
+    "        \"/reanalysis\": precipitation[\"/reanalysis\"],\n",
+    "    }\n",
+    ")\n",
+    "interpolated"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "25",
+   "metadata": {},
+   "source": [
+    "::::{admonition} Exercise\n",
+    ":class: tip\n",
+    "Compute the difference between total observed and modelled precipitation, and plot the result.\n",
+    "\n",
+    ":::{admonition} Solution\n",
+    ":class: dropdown\n",
+    "\n",
+    "```python\n",
+    "total = precipitation.sum(dim=[\"lon\", \"lat\"])\n",
+    "difference = total[\"/observed/precipitation\"] - total[\"/reanalysis/precipitation\"]\n",
+    "difference.plot()\n",
+    "```\n",
+    ":::\n",
+    "::::\n"
+   ]
   }
  ],
  "metadata": {