Chore: improve test stability and address some warnings

.circleci/continue_config.yml

@@ -310,10 +310,10 @@ workflows:
                 - athena
                 - fabric
                 - gcp-postgres
-          filters:
-            branches:
-              only:
-                - main
+          #filters:
+          #  branches:
+          #    only:
+          #      - main
       - ui_style
       - ui_test
       - vscode_test

.circleci/install-prerequisites.sh

@@ -34,4 +34,9 @@ echo "Installing OS-level dependencies: $ALL_DEPENDENCIES"
 
 sudo apt-get clean && sudo apt-get -y update && sudo ACCEPT_EULA='Y' apt-get -y install $ALL_DEPENDENCIES
 
+if [ "$ENGINE" == "spark" ]; then
+    echo "Using Java version for spark:"
+    java -version
+fi
+
 echo "All done"

Makefile

@@ -188,7 +188,7 @@ postgres-test: engine-postgres-up
 	pytest -n auto -m "postgres" --retries 3 --junitxml=test-results/junit-postgres.xml
 
 spark-test: engine-spark-up
-	pytest -n auto -m "spark" --retries 3 --junitxml=test-results/junit-spark.xml
+	pytest -n auto -m "spark" --retries 3 --junitxml=test-results/junit-spark.xml && pytest -n auto -m "pyspark" --retries 3 --junitxml=test-results/junit-pyspark.xml
 
 trino-test: engine-trino-up
 	pytest -n auto -m "trino" --retries 3 --junitxml=test-results/junit-trino.xml

pyproject.toml

@@ -264,8 +264,13 @@ markers = [
     "redshift: test for Redshift",
     "snowflake: test for Snowflake",
     "spark: test for Spark",
+    "pyspark: test for PySpark that need to run separately from the other spark tests",
     "trino: test for Trino (all connectors)",
-    "risingwave: test for Risingwave"
+    "risingwave: test for Risingwave",
+
+    # Other
+    "set_default_connection",
+    "registry_isolation"
 ]
 addopts = "-n 0 --dist=loadgroup"
 asyncio_default_fixture_loop_scope = "session"

sqlmesh/core/test/context.py

@@ -18,6 +18,8 @@ class TestExecutionContext(ExecutionContext):
         models: All upstream models to use for expansion and mapping of physical locations.
     """
 
+    __test__ = False  # prevent pytest trying to collect this as a test class
+
     def __init__(
         self,
         engine_adapter: EngineAdapter,

sqlmesh/dbt/test.py

@@ -61,6 +61,10 @@ class TestConfig(GeneralConfig):
         error_if: Conditional expression (default "!=0") to detect if error condition met (Not supported).
     """
 
+    __test__ = (
+        False  # prevent pytest trying to collect this as a test class when it's imported in a test
+    )
+
     # SQLMesh fields
     path: Path = Path()
     name: str

tests/conftest.py

@@ -222,27 +222,19 @@ def pytest_runtest_makereport(item: pytest.Item, call: pytest.CallInfo):
     # note: the hook always has to yield
     outcome = yield
 
-    # we only care about tests that used the tmp_path fixture
-    if "tmp_path" not in getattr(item, "fixturenames", []):
-        return
-
     result: pytest.TestReport = outcome.get_result()
 
     if result.when != "teardown":
-        return
+        return result
 
     # If we specifically failed with a StashKey error in teardown, mark the test as passed
-    if result.failed:
-        exception = call.excinfo
-        if (
-            exception
-            and isinstance(exception.value, KeyError)
-            and "_pytest.stash.StashKey" in repr(exception)
-        ):
-            result.outcome = "passed"
-            item.add_report_section(
-                "teardown", "stderr", f"Ignored tmp_path teardown error: {exception}"
-            )
+    if (exception := call.excinfo) and "_pytest.stash.StashKey" in repr(exception):
+        call.excinfo = None
+        result.outcome = "passed"
+        item.add_report_section(
+            "teardown", "stderr", f"Ignored tmp_path teardown error: {exception}"
+        )
+        return result
 
 
 def pytest_configure(config: pytest.Config):

tests/dbt/test_integration.py

@@ -27,6 +27,8 @@
 
 
 class TestType(str, Enum):
+    __test__ = False  # prevent pytest trying to collect this as a test class
+
     DBT_RUNTIME = "dbt_runtime"
     DBT_ADAPTER = "dbt_adapter"
     SQLMESH = "sqlmesh"
@@ -53,6 +55,8 @@ def is_sqlmesh_runtime(self) -> bool:
 
 
 class TestStrategy(str, Enum):
+    __test__ = False  # prevent pytest trying to collect this as a test class
+
     CHECK = "check"
     TIMESTAMP = "timestamp"
 

tests/engines/spark/test_db_api.py

@@ -4,10 +4,10 @@
 from sqlmesh.engines.spark.db_api import errors
 from sqlmesh.engines.spark.db_api import spark_session as spark_session_db
 
-pytestmark = [
-    pytest.mark.slow,
-    pytest.mark.spark_pyspark,
-]
+# note: this is deliberately not marked with 'spark' so that it
+# can run separately from the spark integration tests.
+# running them at the same time mutates some global state in the SparkSession which breaks these tests
+pytestmark = [pytest.mark.slow, pytest.mark.pyspark]
 
 
 def test_spark_session_cursor(spark_session: SparkSession):

tests/integrations/github/cicd/test_github_commands.py

@@ -5,6 +5,8 @@
 from unittest import TestCase, mock
 from unittest.result import TestResult
 
+TestResult.__test__ = False  # prevent pytest trying to collect this as a test class
+
 import pytest
 from pytest_mock.plugin import MockerFixture