fix(sql): emit index nodes for CREATE [UNIQUE] INDEX statements

CHANGELOG.md

@@ -4,6 +4,7 @@ Full release notes with details on each version: [GitHub Releases](https://githu
 
 ## 0.9.57 (unreleased)
 
+- Fix: `CREATE INDEX` / `CREATE UNIQUE INDEX` statements now produce index nodes linked to their table by an `indexes` edge — the grammar parsed them, but the SQL walk never dispatched on `create_index`, so every index was silently dropped (#3467, thanks @lrafasouza).
 - Fix: an incremental rebuild no longer wipes cross-file project AST nodes — re-extracting one `.csproj`/`.sln` was dropping package/framework nodes of a *referenced* project (whose stub carried the referenced file's `source_file`); the AST-replacement set is now derived from the files actually extracted (#3411, thanks @hopstreax).
 - Fix: when duplicate nodes merge, the richer (more complete) node is now kept as the survivor and the losers' non-empty fields are folded in, instead of a shorter-id passing mention winning and dropping content (#3372, thanks @abhay-codes07).
 - Fix: a C# generic call site with explicit type arguments — `Get<int>(...)`, unqualified or through `this` — now resolves to the method definition instead of capturing `Get<int>` as the callee and failing to match (#3406, thanks @abhay-codes07).

graphify/extractors/sql.py

@@ -486,6 +486,32 @@ def walk(node) -> None:
                     tbl_nid = table_nids.get(_norm_ident(tbl_name)) or _ref_stub(tbl_name)
                     _add_edge(trig_nid, tbl_nid, "triggers", line)
 
+        elif t == "create_index":
+            # CREATE [UNIQUE] INDEX [CONCURRENTLY] [IF NOT EXISTS] <name>
+            # ON <table> (...). Unlike CREATE POLICY (#3401) the grammar
+            # parses this statement fine; the walk simply never dispatched on
+            # it, so every index was silently dropped (#3467). The name is the
+            # identifier (or a quoted literal) before ON; the table is the
+            # object_reference after it. An unnamed index (`CREATE INDEX ON
+            # t (c)`) has nothing to name a node after and is skipped.
+            index_name: str | None = None
+            index_table: str | None = None
+            after_on = False
+            for c in node.children:
+                if c.type == "keyword_on":
+                    after_on = True
+                elif not after_on and index_name is None and c.type in ("identifier", "literal"):
+                    index_name = _read(c).strip('"`')
+                elif after_on and index_table is None and c.type == "object_reference":
+                    index_table = _read(c)
+            if index_name:
+                index_nid = _make_id(stem, index_name)
+                _add_node(index_nid, index_name, line)
+                if index_table:
+                    index_tbl_nid = (table_nids.get(_norm_ident(index_table))
+                                     or _ref_stub(index_table))
+                    _add_edge(index_nid, index_tbl_nid, "indexes", line)
+
         # NOTE: there is deliberately NO recovery scan on individual ERROR
         # nodes. Any ERROR node anywhere makes root.has_error true, so the
         # whole-file masked scan below this walk already recovers everything

tests/test_multilang.py

@@ -531,6 +531,41 @@ def test_sql_no_dangling_edges():
     for e in r["edges"]:
         assert e["source"] in node_ids, f"dangling source: {e['source']}"
 
+def test_sql_create_index_emits_index_node_linked_to_its_table(tmp_path):
+    """#3467: CREATE [UNIQUE] INDEX parsed fine but the walk never dispatched
+    on create_index, so every index was silently dropped."""
+    pytest.importorskip("tree_sitter_sql")
+    p = tmp_path / "schema.sql"
+    p.write_text(
+        "CREATE TABLE public.profiles (id uuid PRIMARY KEY, owner uuid NOT NULL);\n"
+        "CREATE INDEX profiles_owner_idx ON public.profiles (owner);\n"
+        "CREATE UNIQUE INDEX IF NOT EXISTS profiles_id_uniq ON public.profiles (id);\n"
+        "CREATE INDEX CONCURRENTLY orders_customer_idx ON public.orders (customer_id);\n"
+        'CREATE INDEX "quoted idx" ON public.profiles (owner);\n'
+        "CREATE INDEX ON public.profiles (owner);\n",
+        encoding="utf-8",
+    )
+    r = extract_sql(p)
+    by_label = {n["label"]: n for n in r["nodes"]}
+    for name in ("profiles_owner_idx", "profiles_id_uniq", "orders_customer_idx", "quoted idx"):
+        assert name in by_label, name
+        assert by_label[name]["source_file"] == str(p)
+    edges = {(e["source"], e["relation"], e["target"]) for e in r["edges"]}
+    profiles = by_label["public.profiles"]["id"]
+    assert (by_label["profiles_owner_idx"]["id"], "indexes", profiles) in edges
+    assert (by_label["profiles_id_uniq"]["id"], "indexes", profiles) in edges
+    assert (by_label["quoted idx"]["id"], "indexes", profiles) in edges
+    # An index on a table defined in another file links to a sourceless stub,
+    # the same way a trigger does (#2324).
+    orders = by_label["public.orders"]
+    assert orders["source_file"] == ""
+    assert (by_label["orders_customer_idx"]["id"], "indexes", orders["id"]) in edges
+    # The unnamed index is skipped and nothing dangles.
+    node_ids = {n["id"] for n in r["nodes"]}
+    assert all(e["source"] in node_ids and e["target"] in node_ids for e in r["edges"])
+    assert sum(1 for e in r["edges"] if e["relation"] == "indexes") == 4
+
+
 def test_sql_tsql_bracketed_procedure_is_recovered(tmp_path):
     """T-SQL CREATE PROCEDURE [Schema].[Name] ... AS BEGIN...END.