SNOW-2676993: Add support for to_snowpark in faster pandas (#3989)

Author: sfc-gh-helmeleegy

CHANGELOG.md

@@ -158,6 +158,7 @@
   - `groupby.all`
   - `groupby.unique`
   - `to_snowflake`
+  - `to_snowpark`
 - Make faster pandas disabled by default (opt-in instead of opt-out).
 - Improve performance of `drop_duplicates` by avoiding joins when `keep!=False` in faster pandas.
 

src/snowflake/snowpark/modin/plugin/compiler/snowflake_query_compiler.py

@@ -2332,6 +2332,22 @@ def _to_snowflake_internal(
 
     def to_snowpark(
         self, index: bool = True, index_label: Optional[IndexLabel] = None
+    ) -> SnowparkDataFrame:
+        """
+        Wrapper around _to_snowpark_internal to be supported in faster pandas.
+        """
+        if self._relaxed_query_compiler is not None and not index:
+            return self._relaxed_query_compiler._to_snowpark_internal(
+                index=index,
+                index_label=index_label,
+            )
+        return self._to_snowpark_internal(
+            index=index,
+            index_label=index_label,
+        )
+
+    def _to_snowpark_internal(
+        self, index: bool = True, index_label: Optional[IndexLabel] = None
     ) -> SnowparkDataFrame:
         """
         Convert the Snowpark pandas Dataframe to Snowpark Dataframe. The Snowpark Dataframe is created by selecting

tests/integ/modin/test_faster_pandas.py

@@ -1786,6 +1786,40 @@ def test_to_snowflake(session):
         assert_frame_equal(snow_result, native_result)
 
 
+@sql_count_checker(query_count=3)
+def test_to_snowpark(session):
+    with session_parameter_override(
+        session, "dummy_row_pos_optimization_enabled", True
+    ):
+        # create tables
+        table_name = Utils.random_name_for_temp_object(TempObjectType.TABLE)
+        session.create_dataframe(
+            native_pd.DataFrame([[2, 12], [1, 11], [3, 13]], columns=["A", "B"])
+        ).write.save_as_table(table_name, table_type="temp")
+
+        # create snow dataframes
+        df = pd.read_snowflake(table_name)
+        sdf = df.to_snowpark(index=False)
+        snow_result = sdf.to_snowpark_pandas().sort_values(by="A")
+
+        # verify that the input dataframe has a populated relaxed query compiler
+        assert df._query_compiler._relaxed_query_compiler is not None
+        assert df._query_compiler._relaxed_query_compiler._dummy_row_pos_mode is True
+        # verify that the output dataframe also has a populated relaxed query compiler
+        assert snow_result._query_compiler._relaxed_query_compiler is not None
+        assert (
+            snow_result._query_compiler._relaxed_query_compiler._dummy_row_pos_mode
+            is True
+        )
+
+        # create pandas dataframes
+        native_df = df.to_pandas()
+        native_result = native_df.sort_values(by="A")
+
+        # compare results
+        assert_frame_equal(snow_result, native_result)
+
+
 @sql_count_checker(query_count=0)
 def test_dummy_row_pos_optimization_enabled_on_session(db_parameters):
     with Session.builder.configs(db_parameters).create() as new_session: