feat: Enhance sv support

code_review_graph/changes.py

@@ -309,10 +309,15 @@ def analyze_changes(
         for fp in changed_files:
             changed_nodes.extend(store.get_nodes_by_file(fp))
 
-    # Filter to functions/tests for risk scoring (skip File nodes).
+    # Filter to functions/tests for risk scoring (skip File nodes). Verilog/
+    # SystemVerilog signal-level nodes (ports/nets/params, modeled as Function
+    # nodes) are declarations, not callable functions: they have no TESTED_BY
+    # edges, so including them would flood the output with bogus "test gap"
+    # entries on any .sv edit.
     changed_funcs = [
         n for n in changed_nodes
         if n.kind in ("Function", "Test", "Class")
+        and not n.extra.get("verilog_kind")
     ]
 
     # Cap to prevent O(N*M) query explosion on large PRs.

code_review_graph/flows.py

@@ -177,6 +177,12 @@ def detect_entry_points(
         if not include_tests and (node.is_test or _is_test_file(node.file_path)):
             continue
 
+        # Skip Verilog/SystemVerilog signal-level nodes (ports/nets/params,
+        # modeled as Function nodes): they have no incoming CALLS edges and
+        # would otherwise all register as bogus entry-point roots.
+        if node.extra.get("verilog_kind"):
+            continue
+
         is_entry = False
 
         # True root: no one calls this function.

code_review_graph/graph.py

@@ -719,6 +719,14 @@ def get_impact_radius_sql(
         # Batch-fetch nodes
         changed_nodes = self._batch_get_nodes(seeds)
         impacted_nodes = self._batch_get_nodes(impacted_qns)
+        # Drop Verilog/SystemVerilog signal-level nodes (ports/nets/params,
+        # modeled as Function nodes). They are leaves reached via CONTAINS/
+        # REFERENCES edges from a touched module, so excluding them removes no
+        # real downstream impact but stops every signal of that module from
+        # inflating impacted_files and the impact count.
+        impacted_nodes = [
+            n for n in impacted_nodes if not n.extra.get("verilog_kind")
+        ]
 
         total_impacted = len(impacted_nodes)
         truncated = total_impacted > max_nodes
@@ -830,9 +838,19 @@ def get_stats(self) -> GraphStats:
         total_nodes = self._conn.execute("SELECT COUNT(*) FROM nodes").fetchone()[0]
         total_edges = self._conn.execute("SELECT COUNT(*) FROM edges").fetchone()[0]
 
+        # Verilog/SystemVerilog signal-level nodes (ports/nets/params/etc.) are
+        # stored as kind='Function' with a ``verilog_kind`` discriminator in
+        # ``extra``. Re-bucket them under 'Signal' so they do not inflate the
+        # user-facing Function count. (LIKE on the JSON text avoids a json1
+        # dependency; the key is a fixed literal, not user input.)
         nodes_by_kind: dict[str, int] = {}
-        for row in self._conn.execute("SELECT kind, COUNT(*) as cnt FROM nodes GROUP BY kind"):
-            nodes_by_kind[row["kind"]] = row["cnt"]
+        for row in self._conn.execute(
+            "SELECT CASE WHEN extra LIKE '%\"verilog_kind\"%' THEN 'Signal'"
+            " ELSE kind END AS k, COUNT(*) as cnt FROM nodes"
+            " GROUP BY CASE WHEN extra LIKE '%\"verilog_kind\"%' THEN 'Signal'"
+            " ELSE kind END"
+        ):
+            nodes_by_kind[row["k"]] = row["cnt"]
 
         edges_by_kind: dict[str, int] = {}
         for row in self._conn.execute("SELECT kind, COUNT(*) as cnt FROM edges GROUP BY kind"):
@@ -884,6 +902,10 @@ def get_nodes_by_size(
             "line_start IS NOT NULL",
             "line_end IS NOT NULL",
             "(line_end - line_start + 1) >= ?",
+            # Exclude Verilog/SystemVerilog signal-level nodes (ports/nets/params,
+            # modeled as Function nodes): a multi-line covergroup/typedef is not
+            # an oversized function to decompose.
+            "extra NOT LIKE '%\"verilog_kind\"%'",
         ]
         params: list = [min_lines]