test(examples): add tests for schema=None fallback and data_file override

- Add test_find_dataset_falls_back_to_schema_none for legacy dataset lookup
- Add test_get_multi_dataset_config_data_file_override for explicit data_file
- Add test_get_multi_dataset_config_data_file_missing for missing file warning

These tests cover the edge cases noted in PR review feedback.

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

Author: sadpandajoe

tests/unit_tests/examples/data_loading_test.py

@@ -228,3 +228,68 @@ def test_get_multi_dataset_config_missing_table_name(tmp_path: Path) -> None:
     # Falls back to dataset_name when table_name not in YAML
     assert result["table_name"] == "my_dataset"
     assert result["uuid"] == "test-uuid-5678"
+
+
+def test_get_multi_dataset_config_data_file_override(tmp_path: Path) -> None:
+    """Test that explicit data_file in YAML overrides the default data file."""
+    from superset.examples.data_loading import _get_multi_dataset_config
+
+    # Create datasets directory and YAML file with explicit data_file
+    datasets_dir = tmp_path / "datasets"
+    datasets_dir.mkdir()
+
+    # Create data directory and the explicit data file
+    data_dir = tmp_path / "data"
+    data_dir.mkdir()
+    explicit_file = data_dir / "custom_data.parquet"
+    explicit_file.write_bytes(b"fake parquet")
+
+    yaml_content = """
+table_name: my_dataset
+schema: public
+uuid: 14f48794-ebfa-4f60-a26a-582c49132f1b
+data_file: custom_data.parquet
+"""
+    dataset_yaml = datasets_dir / "my_dataset.yaml"
+    dataset_yaml.write_text(yaml_content)
+
+    # Default data file (would be used if no override)
+    default_data_file = data_dir / "my_dataset.parquet"
+
+    result = _get_multi_dataset_config(tmp_path, "my_dataset", default_data_file)
+
+    # Should use the explicit data_file from YAML
+    assert result["data_file"] == explicit_file
+    assert result["table_name"] == "my_dataset"
+    assert result["uuid"] == "14f48794-ebfa-4f60-a26a-582c49132f1b"
+
+
+def test_get_multi_dataset_config_data_file_missing(tmp_path: Path) -> None:
+    """Test that missing explicit data_file keeps the default data file."""
+    from superset.examples.data_loading import _get_multi_dataset_config
+
+    # Create datasets directory and YAML file with non-existent data_file
+    datasets_dir = tmp_path / "datasets"
+    datasets_dir.mkdir()
+
+    # Create data directory but NOT the explicit file
+    data_dir = tmp_path / "data"
+    data_dir.mkdir()
+
+    yaml_content = """
+table_name: my_dataset
+schema: public
+uuid: 14f48794-ebfa-4f60-a26a-582c49132f1b
+data_file: nonexistent.parquet
+"""
+    dataset_yaml = datasets_dir / "my_dataset.yaml"
+    dataset_yaml.write_text(yaml_content)
+
+    # Default data file passed to the function
+    default_data_file = data_dir / "my_dataset.parquet"
+
+    result = _get_multi_dataset_config(tmp_path, "my_dataset", default_data_file)
+
+    # Should keep the default data_file since explicit one doesn't exist
+    assert result["data_file"] == default_data_file
+    assert result["uuid"] == "14f48794-ebfa-4f60-a26a-582c49132f1b"

tests/unit_tests/examples/generic_loader_test.py

@@ -533,6 +533,49 @@ def test_find_dataset_no_uuid_no_schema(mock_db: MagicMock) -> None:
     assert found_by_uuid is False
 
 
+@patch("superset.examples.generic_loader.db")
+def test_find_dataset_falls_back_to_schema_none(mock_db: MagicMock) -> None:
+    """Test _find_dataset falls back to schema=None when exact schema match not found.
+
+    Scenario: A legacy dataset exists with schema=None (from older load-examples).
+    When looking up with schema="public", the exact match fails, but we should
+    find the legacy row with schema=None so we can backfill the schema.
+    """
+    from superset.examples.generic_loader import _find_dataset
+
+    # Legacy row with schema=None (needs backfill)
+    legacy_row = MagicMock()
+    legacy_row.uuid = None
+    legacy_row.table_name = "users"
+    legacy_row.schema = None
+
+    # Exact schema match returns None, schema=None fallback returns legacy_row
+    call_count = 0
+
+    def filter_by_side_effect(**kwargs):
+        nonlocal call_count
+        call_count += 1
+        mock_result = MagicMock()
+        if kwargs.get("schema") == "public":
+            # Exact schema match fails
+            mock_result.first.return_value = None
+        elif kwargs.get("schema") is None:
+            # Fallback to schema=None finds the legacy row
+            mock_result.first.return_value = legacy_row
+        else:
+            mock_result.first.return_value = None
+        return mock_result
+
+    mock_db.session.query.return_value.filter_by.side_effect = filter_by_side_effect
+
+    result, found_by_uuid = _find_dataset("users", 1, None, "public")
+
+    assert result is legacy_row
+    assert found_by_uuid is False
+    # Verify both lookups happened (exact match + fallback)
+    assert call_count == 2
+
+
 @patch("superset.examples.generic_loader.db")
 @patch("superset.examples.generic_loader.get_example_database")
 def test_load_parquet_table_no_backfill_when_uuid_already_set(