Merge pull request #1664 from AnswerDotAI/strip-outputs

Support stripping specified output representations

Author: jph00

nbdev/clean.py

@@ -62,12 +62,13 @@ def _clean_cell_output_id(lines):
 
 # %% ../nbs/api/11_clean.ipynb #b4cde615
 def _clean_cell_output(cell, clean_ids, allowed_out_meta_keys):
-    "Remove `cell` output execution count and optionally ids from text reprs"
+    "Remove `cell` output execution count, optionally ids from text reprs, and representations listed under `nbdev.strip`"
     outputs = cell.get('outputs', [])
     for o in outputs:
         if 'execution_count' in o: o['execution_count'] = None
         data = o.get('data', {})
         data.pop("application/vnd.google.colaboratory.intrinsic+json", None)
+        for k in nested_idx(o, 'metadata', 'nbdev', 'strip') or []: data.pop(k, None)
         for k in data:
             if k.startswith('text') and clean_ids: data[k] = _clean_cell_output_id(data[k])
             if k.startswith('image') and "svg" not in k: data[k] = data[k].rstrip()

nbs/api/11_clean.ipynb

@@ -166,12 +166,13 @@
    "source": [
     "#| export\n",
     "def _clean_cell_output(cell, clean_ids, allowed_out_meta_keys):\n",
-    "    \"Remove `cell` output execution count and optionally ids from text reprs\"\n",
+    "    \"Remove `cell` output execution count, optionally ids from text reprs, and representations listed under `nbdev.strip`\"\n",
     "    outputs = cell.get('outputs', [])\n",
     "    for o in outputs:\n",
     "        if 'execution_count' in o: o['execution_count'] = None\n",
     "        data = o.get('data', {})\n",
     "        data.pop(\"application/vnd.google.colaboratory.intrinsic+json\", None)\n",
+    "        for k in nested_idx(o, 'metadata', 'nbdev', 'strip') or []: data.pop(k, None)\n",
     "        for k in data:\n",
     "            if k.startswith('text') and clean_ids: data[k] = _clean_cell_output_id(data[k])\n",
     "            if k.startswith('image') and \"svg\" not in k: data[k] = data[k].rstrip()\n",
@@ -379,6 +380,28 @@
     "```"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "id": "063fbebb",
+   "metadata": {},
+   "source": [
+    "A display object with a large, regenerable representation beside a small one can ask for the large one to be removed on clean. It lists the mimetypes to drop under `nbdev.strip` in the output's metadata, which IPython takes from the second element of a `_repr_mimebundle_` return value. `clean_nb` removes those representations and keeps the others:"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "6148ce9b",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "out = dict(output_type='display_data', metadata={'nbdev': {'strip': ['text/html']}},\n",
+    "    data={'text/html': '<iframe srcdoc=\"...\"></iframe>', 'text/markdown': '```html\\n<p>Hi</p>\\n```'})\n",
+    "_nb = dict2nb(dict(cells=[mk_cell('preview()', outputs=[out])], metadata={}))\n",
+    "clean_nb(_nb)\n",
+    "test_eq(list(_nb.cells[0].outputs[0].data), ['text/markdown'])"
+   ]
+  },
   {
    "cell_type": "markdown",
    "id": "89d0a249",