Split up test_dataflow_to_sql_plan module into more manageable components

metricflow/test/fixtures/sql_fixtures.py

@@ -2,9 +2,22 @@
 
 import pytest
 
+from metricflow.model.semantic_manifest_lookup import SemanticManifestLookup
+from metricflow.plan_conversion.column_resolver import DunderColumnAssociationResolver
+from metricflow.plan_conversion.dataflow_to_sql import DataflowToSqlQueryPlanConverter
 from metricflow.sql.render.sql_plan_renderer import DefaultSqlQueryPlanRenderer, SqlQueryPlanRenderer
 
 
 @pytest.fixture
 def default_sql_plan_renderer() -> SqlQueryPlanRenderer:  # noqa: D
     return DefaultSqlQueryPlanRenderer()
+
+
+@pytest.fixture(scope="session")
+def dataflow_to_sql_converter(  # noqa: D
+    simple_semantic_manifest_lookup: SemanticManifestLookup,
+) -> DataflowToSqlQueryPlanConverter:
+    return DataflowToSqlQueryPlanConverter(
+        column_association_resolver=DunderColumnAssociationResolver(simple_semantic_manifest_lookup),
+        semantic_manifest_lookup=simple_semantic_manifest_lookup,
+    )

metricflow/test/generate_snapshots.py

@@ -121,8 +121,8 @@ def run_tests(test_configuration: MetricFlowTestConfiguration) -> None:  # noqa:
         os.environ["MF_SQL_ENGINE_PASSWORD"] = test_configuration.credential_set.engine_password
 
     if test_configuration.engine is SqlEngine.DUCKDB:
-        # Can't use --use-persistent-source-schema with duckdb since it's in memory.
-        run_command(f"pytest -x -vv -n 4 --overwrite-snapshots -m '{SQL_ENGINE_SNAPSHOT_MARKER_NAME}' {TEST_DIRECTORY}")
+        # DuckDB is fast, so generate all snapshots, including the engine-agnostic ones
+        run_command(f"pytest -x -vv -n 4 --overwrite-snapshots -k 'not itest' {TEST_DIRECTORY}")
     elif (
         test_configuration.engine is SqlEngine.REDSHIFT
         or test_configuration.engine is SqlEngine.SNOWFLAKE