fix(export): escape every "<" in JSON embedded in <script> blocks

graphify/exporters/html.py

@@ -613,9 +613,11 @@ def to_html(
         n = member_counts.get(cid, len(communities.get(cid, []))) if member_counts else len(communities.get(cid, []))
         legend_data.append({"cid": cid, "color": color, "label": lbl, "count": n})
 
-    # Escape </script> sequences so embedded JSON cannot break out of the script tag
+    # Escape every "<" so embedded JSON cannot break out of the script tag: besides
+    # "</script", an unclosed "<!--" followed by "<script" keeps the real </script>
+    # from closing the element (#4124). JSON/JS decode < back to "<".
     def _js_safe(obj) -> str:
-        return json.dumps(obj).replace("</", "<\\/")
+        return json.dumps(obj).replace("<", "\\u003c")
 
     nodes_json = _js_safe(vis_nodes)
     edges_json = _js_safe(vis_edges)

graphify/tree_html.py

@@ -568,9 +568,10 @@ def emit_html(
     svg_width: int = 6000,
     svg_height: int = 8000,
 ) -> str:
-    # Escape </script> sequences so embedded JSON cannot break out of the
-    # <script> tag, and HTML-escape values that land in <title>/<h1>.
-    data_json = json.dumps(tree, ensure_ascii=True, separators=(",", ":")).replace("</", "<\\/")
+    # Escape every "<" so embedded JSON cannot break out of the <script> tag
+    # ("</script", or "<!--" followed by "<script", #4124), and HTML-escape
+    # values that land in <title>/<h1>.
+    data_json = json.dumps(tree, ensure_ascii=True, separators=(",", ":")).replace("<", "\\u003c")
     return _HTML_TEMPLATE.format(
         title=_html.escape(title),
         header=_html.escape(header),

tests/test_export.py

@@ -1373,3 +1373,39 @@ def test_to_html_aggregated_community_nodes_runtime(tmp_path):
     assert '<div class="field">Degree: 1</div>' in info_html
     assert "Type: unknown" not in info_html
     assert "Source: -" not in info_html
+
+
+# ── Issue #4124: "<!--" + "<script" in labels swallowed the closing </script> ──
+
+def test_to_html_escapes_script_data_sequences_in_embedded_json():
+    """#4124: an unclosed `<!--` followed later by `<script` puts the HTML tokenizer
+    in the "script data double escaped" state, where the real `</script>` no longer
+    closes the data block and the whole page fails to render. Escaping only `</`
+    left both sequences intact; no `<` may reach the embedded JSON verbatim."""
+    G = build_from_json({
+        "nodes": [
+            {"id": "h1", "label": "An unclosed <!-- opener in a heading",
+             "file_type": "document", "source_file": "guide.md"},
+            {"id": "h2", "label": "Then a <script setup> heading",
+             "file_type": "document", "source_file": "guide.md"},
+        ],
+        "edges": [{"source": "h1", "target": "h2", "relation": "contains",
+                   "confidence": "EXTRACTED", "source_file": "guide.md"}],
+    })
+    G.graph["hyperedges"] = [{"id": "he1", "label": "<!-- group <script",
+                              "nodes": ["h1", "h2"]}]
+    communities = cluster(G)
+    with tempfile.TemporaryDirectory() as tmp:
+        out = Path(tmp) / "graph.html"
+        to_html(G, communities, str(out), community_labels={c: "<!-- c <script" for c in communities})
+        content = out.read_text(encoding="utf-8")
+
+    for name in ("hyperedges", "RAW_NODES", "RAW_EDGES", "LEGEND"):
+        m = re.search(rf"const {name} = (.*?);\n", content)
+        assert m, f"{name} not found in HTML"
+        assert "<" not in m.group(1), f"raw '<' in embedded {name}"
+
+    # The data must still decode to the original labels.
+    labels = {n["id"]: n["label"] for n in _vis_nodes_from_html(content)}
+    assert labels["h1"] == "An unclosed <!-- opener in a heading"
+    assert labels["h2"] == "Then a <script setup> heading"

tests/test_tree_html.py

@@ -9,6 +9,7 @@
 from __future__ import annotations
 
 import json
+import re
 import subprocess
 import sys
 from pathlib import Path
@@ -111,3 +112,19 @@ def test_tree_cli_partial_match_root_succeeds(tmp_path):
     r = _run(["tree", "--root", "pkg"], tmp_path)
     assert r.returncode == 0, r.stderr
     assert (tmp_path / "graphify-out" / "GRAPH_TREE.html").exists()
+
+
+def test_emit_html_escapes_script_data_sequences_in_embedded_json():
+    """#4124: an unclosed `<!--` followed by `<script` in the embedded JSON kept the
+    real `</script>` from closing the data block. No `<` may reach it verbatim."""
+    from graphify.tree_html import emit_html
+
+    tree = {"name": "root", "children": [
+        {"name": "An unclosed <!-- opener"},
+        {"name": "Then a <script setup> heading"},
+    ]}
+    html = emit_html(tree, title="t", header="h")
+    m = re.search(r"const initialJsonData = (.*?);\n", html)
+    assert m, "initialJsonData not found in HTML"
+    assert "<" not in m.group(1)
+    assert json.loads(m.group(1)) == tree