Address Copilot review on #576

- tests: identify rationale nodes by file_type metadata, not id substring
  (decouples regression from current `<stem>_rationale_<line>` naming)
- tests: assert contact_form -> SQLiteQueue.enqueue calls edge directly
  (previously only inversions were guarded; a dropped enqueue edge would
  have passed silently)
- extract.py: clarify cross-file import indexing comment — the filter
  intentionally indexes only class-level entities (function/method labels
  end in "()" and are excluded by design)

Author: opnsrcntrbtr

graphify/extract.py

@@ -2612,8 +2612,11 @@ def _resolve_cross_file_imports(
             stem = Path(src).stem
             label = node.get("label", "")
             nid = node.get("id", "")
-            # Only index real classes/functions (not file nodes, not method stubs,
-            # and not rationale nodes whose labels are free-form docstring text — #563)
+            # Index class-level entities only. Function/method labels end in "()"
+            # so are excluded by the `endswith(")")` filter; file nodes end in ".py";
+            # private/internal labels start with "_"; rationale nodes carry
+            # file_type=="rationale" and must never participate in cross-file
+            # import resolution (#563).
             if (
                 label
                 and not label.endswith((")", ".py"))

tests/test_issue_563.py

@@ -125,13 +125,25 @@ def repro_graph(tmp_path: Path) -> dict:
 # ─────────────────────────────────────────────────────────────────────────────
 
 
+def _file_type_map(graph: dict) -> dict[str, str]:
+    """Map node id -> file_type from the serialized graph.
+
+    Identifying rationale nodes via node metadata (file_type == "rationale")
+    is more durable than substring-matching on the id, which couples to the
+    current `<stem>_rationale_<line>` naming scheme.
+    """
+    return {n["id"]: n.get("file_type", "") for n in graph["nodes"]}
+
+
 def test_no_rationale_nodes_in_uses_or_calls_edges(repro_graph):
     """Rationale nodes describe code; they cannot `use`, `call`, or `reference`
     anything. The only relation a rationale node may participate in is
     `rationale_for` (extractor emits `rationale --rationale_for--> parent`)."""
+    ftype = _file_type_map(repro_graph)
     bad = [
         e for e in repro_graph["links"]
-        if ("rationale" in e["source"] or "rationale" in e["target"])
+        if (ftype.get(e["source"]) == "rationale"
+            or ftype.get(e["target"]) == "rationale")
         and e["relation"] != "rationale_for"
     ]
     assert bad == [], (
@@ -146,9 +158,10 @@ def test_rationale_for_direction_is_rationale_to_parent(repro_graph):
     for this code entity": source is the rationale node, target is the parent.
     The export-side direction restore (#563) was the only thing that made
     this hold under undirected NetworkX storage."""
+    ftype = _file_type_map(repro_graph)
     rats = [e for e in repro_graph["links"] if e["relation"] == "rationale_for"]
     assert rats, "expected some rationale_for edges in the repro"
-    bad = [e for e in rats if "rationale" not in e["source"]]
+    bad = [e for e in rats if ftype.get(e["source"]) != "rationale"]
     assert bad == [], (
         f"rationale_for edges must have a rationale node as source. "
         f"Found {len(bad)} flipped: {bad[:3]}"
@@ -177,6 +190,10 @@ def test_calls_direction_is_caller_to_callee(repro_graph):
         "contact_form_contact_form",
         "jobq_sqlitequeue_check_idempotency",
     ) in calls, f"expected caller->callee edge missing. got: {calls}"
+    assert (
+        "contact_form_contact_form",
+        "jobq_sqlitequeue_enqueue",
+    ) in calls, f"contact_form->enqueue edge missing. got: {calls}"
 
     # The test function calls contact_form()
     assert (