SNOW-1625830: Fix circular import when calling to_snowpark_pandas without initializing Snowpark pandas

CHANGELOG.md

@@ -26,6 +26,7 @@
 - Fixed a bug in `session.read.csv` that caused an error when setting `PARSE_HEADER = True` in an externally defined file format.
 - Fixed a bug in query generation from set operations that allowed generation of duplicate queries when children have common subqueries.
 - Fixed a bug in `session.get_session_stage` that referenced a non-existing stage after switching database or schema.
+- Fixed a bug where calling `DataFrame.to_snowpark_pandas_dataframe` without explicitly initializing the Snowpark pandas plugin caused an error.
 
 ### Snowpark Local Testing Updates
 

src/snowflake/snowpark/dataframe.py

@@ -1007,8 +1007,12 @@ def to_snowpark_pandas(
             B A
             2 1  1  3  1
         """
-        import snowflake.snowpark.modin.pandas as pd  # pragma: no cover
-
+        # black and isort disagree on how to format this section with isort: skip
+        # fmt: off
+        import snowflake.snowpark.modin.plugin  # isort: skip  # noqa: F401
+        # If snowflake.snowpark.modin.plugin was successfully imported, then modin.pandas is available
+        import modin.pandas as pd  # isort: skip
+        # fmt: on
         # create a temporary table out of the current snowpark dataframe
         temporary_table_name = random_name_for_temp_object(
             TempObjectType.TABLE

tests/integ/test_df_to_snowpark_pandas.py

@@ -0,0 +1,56 @@
+#!/usr/bin/env python3
+#
+# Copyright (c) 2012-2024 Snowflake Computing Inc. All rights reserved.
+#
+
+# Tests behavior of to_snowpark_pandas() without explicitly initializing Snowpark pandas.
+
+import pytest
+
+from snowflake.snowpark._internal.utils import TempObjectType
+from tests.utils import Utils
+
+pytestmark = [
+    pytest.mark.xfail(
+        "config.getoption('local_testing_mode', default=False)",
+        reason="This is testing Snowpark pandas installation",
+        run=False,
+    )
+]
+
+
+@pytest.fixture(scope="module")
+def tmp_table_basic(session):
+    table_name = Utils.random_name_for_temp_object(TempObjectType.TABLE)
+    Utils.create_table(
+        session, table_name, "id integer, foot_size float, shoe_model varchar"
+    )
+    session.sql(f"insert into {table_name} values (1, 32.0, 'medium')").collect()
+    session.sql(f"insert into {table_name} values (2, 27.0, 'small')").collect()
+    session.sql(f"insert into {table_name} values (3, 40.0, 'large')").collect()
+
+    try:
+        yield table_name
+    finally:
+        Utils.drop_table(session, table_name)
+
+
+def test_to_snowpark_pandas_no_modin(session, tmp_table_basic):
+    snowpark_df = session.table(tmp_table_basic)
+    # Check if modin is installed (if so, we're running in Snowpark pandas; if not, we're just in Snowpark Python)
+    try:
+        import modin  # noqa: F401
+    except ModuleNotFoundError:
+        # Current Snowpark Python installs pandas==2.2.2, but Snowpark pandas depends on modin
+        # 0.28.1, which needs pandas==2.2.1. The pandas version check is currently performed
+        # before Snowpark pandas checks whether modin is installed.
+        # TODO: SNOW-1552497: after upgrading to modin 0.30.1, Snowpark pandas will support
+        # all pandas 2.2.x, and this function call will raise a ModuleNotFoundError since
+        # modin is not installed.
+        with pytest.raises(
+            RuntimeError,
+            match="does not match the supported pandas version in Snowpark pandas",
+        ):
+            snowpark_df.to_snowpark_pandas()
+    else:
+        snowpark_df.to_snowpark_pandas()  # should have no errors

tox.ini

@@ -96,11 +96,11 @@ commands =
     local: {env:SNOWFLAKE_PYTEST_CMD} --local_testing_mode -m "integ or unit or mock" {posargs:} tests
     dailynotdoctest: {env:SNOWFLAKE_PYTEST_DAILY_CMD} -m "{env:SNOWFLAKE_TEST_TYPE} or udf" {posargs:} tests
     # Snowpark pandas commands:
-    snowparkpandasnotdoctest: {env:MODIN_PYTEST_CMD} --durations=20 -m "{env:SNOWFLAKE_TEST_TYPE}" {posargs:} {env:SNOW_1314507_WORKAROUND_RERUN_FLAGS} tests/unit/modin tests/integ/modin
+    snowparkpandasnotdoctest: {env:MODIN_PYTEST_CMD} --durations=20 -m "{env:SNOWFLAKE_TEST_TYPE}" {posargs:} {env:SNOW_1314507_WORKAROUND_RERUN_FLAGS} tests/unit/modin tests/integ/modin tests/integ/test_df_to_snowpark_pandas.py
     # This one only run doctest but we still need to include the tests folder to let tests/conftest.py to mark the doctest files for us
     snowparkpandasdoctest: {env:MODIN_PYTEST_CMD} --durations=20 -m "{env:SNOWFLAKE_TEST_TYPE}" {posargs:} src/snowflake/snowpark/modin/ tests/unit/modin
     # This one is used by daily_modin_precommit.yml
-    snowparkpandasdailynotdoctest: {env:MODIN_PYTEST_DAILY_CMD} --durations=20 -m "{env:SNOWFLAKE_TEST_TYPE}" {posargs:} {env:SNOW_1314507_WORKAROUND_RERUN_FLAGS} tests/unit/modin tests/integ/modin
+    snowparkpandasdailynotdoctest: {env:MODIN_PYTEST_DAILY_CMD} --durations=20 -m "{env:SNOWFLAKE_TEST_TYPE}" {posargs:} {env:SNOW_1314507_WORKAROUND_RERUN_FLAGS} tests/unit/modin tests/integ/modin tests/integ/test_df_to_snowpark_pandas.py
     # This one is only called by jenkins job and the only difference from `snowparkpandasnotdoctest` is that it uses
     # MODIN_PYTEST_NO_COV_CMD instead of MODIN_PYTEST_CMD
     snowparkpandasjenkins: {env:MODIN_PYTEST_NO_COV_CMD} --durations=20 -m "{env:SNOWFLAKE_TEST_TYPE}" {posargs:} {env:SNOW_1314507_WORKAROUND_RERUN_FLAGS} tests/unit/modin tests/integ/modin