feat: create materialized view infers column name (#768)

Per issue #739 
Create materialized view is able infer column names. Column name is
therefore optional on the user end.

Before:
CREATE MATERIALIZED VIEW dummy_view **(id, label)** AS SELECT id,
DummyObjectDetector(data).label FROM MyVideo;

After:
CREATE MATERIALIZED VIEW dummy_view AS SELECT id,
DummyObjectDetector(data).label FROM MyVideo;

Co-authored-by: Yulai Cui <alex@Yulais-MacBook-Pro.local>

eva/parser/eva.lark

@@ -30,7 +30,7 @@ rename_table: RENAME TABLE table_name TO table_name
 create_udf: CREATE UDF if_not_exists? udf_name INPUT create_definitions OUTPUT create_definitions TYPE udf_type IMPL udf_impl udf_metadata* | CREATE UDF if_not_exists? udf_name IMPL udf_impl udf_metadata* | CREATE UDF if_not_exists? udf_name TYPE udf_type udf_metadata*
 
 // Create Materialized View
-create_materialized_view: CREATE MATERIALIZED VIEW if_not_exists? table_name ("(" uid_list ")") AS select_statement 
+create_materialized_view: CREATE MATERIALIZED VIEW if_not_exists? table_name ("(" uid_list ")") AS select_statement | CREATE MATERIALIZED VIEW if_not_exists? table_name AS select_statement
 
 // Details
 udf_name: uid

eva/parser/lark_visitor/_create_statements.py

@@ -234,6 +234,7 @@ def create_materialized_view(self, tree):
         view_info = None
         if_not_exists = False
         query = None
+        uid_list = []
 
         for child in tree.children:
             if isinstance(child, Tree):
@@ -248,9 +249,17 @@ def create_materialized_view(self, tree):
 
         # setting all other column definition attributes as None,
         # need to figure from query
-        col_list = [
-            ColumnDefinition(uid.col_name, None, None, None) for uid in uid_list
-        ]
+        if uid_list == []:
+            assert (query is not None)
+            for uid in query.target_list:
+                uid_list.append(uid)
+
+        col_list = []
+        for uid in uid_list:
+            if hasattr(uid, "col_name"):
+                col_list.append(ColumnDefinition(uid.col_name, None, None, None))
+            elif hasattr(uid, "output"):
+                col_list.append(ColumnDefinition(uid.output, None, None, None))
         return CreateMaterializedViewStatement(
             view_info, col_list, if_not_exists, query
         )

test/integration_tests/test_mat_executor.py

@@ -54,18 +54,36 @@ def tearDownClass(cls):
         execute_query_fetch_all("DROP TABLE UATRAC;")
 
     def test_should_mat_view_with_dummy(self):
-        materialized_query = """CREATE MATERIALIZED VIEW dummy_view (id, label)
+        materialized_query = """CREATE MATERIALIZED VIEW dummy_view0 (id, label)
             AS SELECT id, DummyObjectDetector(data).label FROM MyVideo;
         """
         execute_query_fetch_all(materialized_query)
 
-        select_query = "SELECT id, label FROM dummy_view;"
+        select_query = "SELECT id, label FROM dummy_view0;"
         actual_batch = execute_query_fetch_all(select_query)
         actual_batch.sort()
 
         labels = DummyObjectDetector().labels
         expected = [
-            {"dummy_view.id": i, "dummy_view.label": [labels[1 + i % 2]]}
+            {"dummy_view0.id": i, "dummy_view0.label": [labels[1 + i % 2]]}
+            for i in range(NUM_FRAMES)
+        ]
+        expected_batch = Batch(frames=pd.DataFrame(expected))
+        self.assertEqual(actual_batch, expected_batch)
+
+    def test_should_infer_mat_view_column_names_with_dummy(self):
+        materialized_query = """CREATE MATERIALIZED VIEW dummy_view1
+            AS SELECT id, DummyObjectDetector(data).label FROM MyVideo;
+        """
+        execute_query_fetch_all(materialized_query)
+
+        select_query = "SELECT id, label FROM dummy_view1;"
+        actual_batch = execute_query_fetch_all(select_query)
+        actual_batch.sort()
+
+        labels = DummyObjectDetector().labels
+        expected = [
+            {"dummy_view1.id": i, "dummy_view1.label": [labels[1 + i % 2]]}
             for i in range(NUM_FRAMES)
         ]
         expected_batch = Batch(frames=pd.DataFrame(expected))
@@ -131,18 +149,42 @@ def test_should_mat_view_with_fastrcnn_lateral_join(self):
         )
         query = (
             "CREATE MATERIALIZED VIEW IF NOT EXISTS "
-            f"uadtrac_fastRCNN_new (id, label, bbox) AS {select_query};"
+            f"uadtrac_fastRCNN_new0 (id, label, bbox) AS {select_query};"
+        )
+        execute_query_fetch_all(query)
+
+        select_view_query = "SELECT id, label, bbox FROM uadtrac_fastRCNN_new0"
+        actual_batch = execute_query_fetch_all(select_view_query)
+        actual_batch.sort()
+
+        self.assertEqual(len(actual_batch), 5)
+        # non-trivial test case
+        res = actual_batch.frames
+        for idx in res.index:
+            self.assertTrue("car" in res["uadtrac_fastrcnn_new0.label"][idx])
+
+        execute_query_fetch_all("DROP TABLE IF EXISTS uadtrac_fastRCNN;")
+
+    @pytest.mark.torchtest
+    def test_should_infer_mat_view_column_names_with_fastrcnn_lateral_join(self):
+        select_query = (
+            "SELECT id, label, bbox FROM UATRAC JOIN LATERAL "
+            "Yolo(data) AS T(label, bbox, score) WHERE id < 5;"
+        )
+        query = (
+            "CREATE MATERIALIZED VIEW IF NOT EXISTS "
+            f"uadtrac_fastRCNN_new1 AS {select_query};"
         )
         execute_query_fetch_all(query)
 
-        select_view_query = "SELECT id, label, bbox FROM uadtrac_fastRCNN_new"
+        select_view_query = "SELECT id, label, bbox FROM uadtrac_fastRCNN_new1"
         actual_batch = execute_query_fetch_all(select_view_query)
         actual_batch.sort()
 
         self.assertEqual(len(actual_batch), 5)
         # non-trivial test case
         res = actual_batch.frames
         for idx in res.index:
-            self.assertTrue("car" in res["uadtrac_fastrcnn_new.label"][idx])
+            self.assertTrue("car" in res["uadtrac_fastrcnn_new1.label"][idx])
 
         execute_query_fetch_all("DROP TABLE IF EXISTS uadtrac_fastRCNN;")