Fix Column.isin()with empty list

src/snowflake/snowpark/column.py

@@ -437,6 +437,16 @@ def validate_value(value_expr: Expression):
             for ve in value_expressions:
                 validate_value(ve)
 
+        if len(value_expressions) == 0:
+            # If `InExpression()` gets passed an empty list as `value_expressions`, it
+            # produces an invalid (sub-)query with empty brackets, like:
+            #
+            #   `SELECT ... FROM ... WHERE <column> IN ()`
+            #
+            # To avoid this we take the shortcut and return a no-op in case of an empty
+            # list.
+            return Column(Literal(None))
+
         return Column(InExpression(self._expression, value_expressions))
 
     def between(

tests/integ/test_column.py

@@ -216,3 +216,15 @@ def test_column_with_builtins_that_shadow_functions(session):
     with pytest.raises(TypeError) as ex_info:
         TestData.double1(session).select(sum(col("a"))).collect()
     assert iter_error_msg_text in str(ex_info)
+
+
+def test_return_empty_set_for_empty_filter_isin(session):
+    """
+    Filtering a dataframe with `isin()` and an empty set should yield an empty
+    dataframe.
+
+    NOTE: In SQL an `IN` expression without at least one value is not valid at all.
+    """
+    df1 = session.create_dataframe([1, 2]).to_df("a")
+    assert df1.filter(df1["a"].isin(None)).collect() == []
+    assert df1.filter(df1["a"].isin([])).collect() == []

tests/mock/test_filter.py

@@ -148,6 +148,15 @@ def test_null_nan_filter(session):
     assert res[1] == Row(1.0, 1.0)
     assert res[2] == Row(None, None)
 
+    # Check `in` expression with empty sets
+    origin_df: DataFrame = session.create_dataframe([[1], [2]], schema=["a"])
+
+    res = origin_df.filter(origin_df["a"].isin(None)).collect()
+    assert res == []
+
+    res = origin_df.filter(origin_df["a"].isin([])).collect()
+    assert res == []
+
 
 def test_chain_filter(session):
     origin_df: DataFrame = session.create_dataframe(