ignore type

posthog/temporal/data_imports/pipelines/sql_database/arrow_helpers.py

@@ -123,10 +123,12 @@ def row_tuples_to_arrow(rows: Sequence[RowAny], columns: TTableSchemaColumns, tz
 
         if issubclass(py_type, bytes) or issubclass(py_type, str):
             # For bytes/str columns, ensure any dict values are serialized to JSON strings
-            columnar_known_types[field.name] = [
+            # Convert to numpy array after processing
+            processed_values = [
                 None if x is None else json_dumps(x) if isinstance(x, dict | list) else x
                 for x in columnar_known_types[field.name]
             ]
+            columnar_known_types[field.name] = np.array(processed_values, dtype=object)
 
     # If there are unknown type columns, first create a table to infer their types
     if columnar_unknown_types:

posthog/temporal/data_imports/pipelines/sql_database/test/test_arrow_helpers.py

@@ -29,7 +29,7 @@ def test_row_tuples_to_arrow_string_column_with_dict():
     columns = {"string_col": {"name": "string_col", "data_type": "text", "nullable": True}}
 
     # This should now succeed and serialize the dictionary to JSON
-    table = row_tuples_to_arrow(rows, columns, "UTC")
+    table = row_tuples_to_arrow(rows, columns, "UTC")  # type: ignore
 
     # Verify the results
     assert table.column("string_col")[0].as_py() == ""