improve test coverage

python/cudf_polars/cudf_polars/dsl/ir.py

@@ -1666,11 +1666,15 @@ def __init__(self, schema: Schema, key: str, left: IR, right: IR):
             schema,
             key,
         )
+        assert isinstance(left, Sort)
+        assert isinstance(right, Sort)
+        assert left.order == right.order
+        assert len(left.schema.keys()) <= len(right.schema.keys())
 
     @classmethod
     def do_evaluate(cls, schema: Schema, key: str, *dfs: DataFrame) -> DataFrame:
         left, right = dfs
-        left, right = dfs
+        right = right.discard_columns(right.column_names_set - left.column_names_set)
         on_col_left = left.select_columns({key})[0]
         on_col_right = right.select_columns({key})[0]
         return DataFrame.from_table(

python/cudf_polars/tests/test_merge_sorted.py

@@ -14,21 +14,45 @@
     [
         pytest.param(
             True,
-            marks=pytest.mark.xfail(
-                reason="https://github.com/rapidsai/cudf/issues/18089"
-            ),
+            marks=pytest.mark.xfail(reason="polars/issues/21464"),
         ),
         False,
     ],
 )
-def test_merge_sorted(descending):
+def test_merge_sorted_without_nulls(descending):
+    df0 = pl.LazyFrame({"name": ["steve", "elise", "bob"], "age": [42, 44, 18]}).sort(
+        "age", descending=descending
+    )
+    df1 = pl.LazyFrame(
+        {
+            "name": ["anna", "megan", "steve", "thomas"],
+            "age": [21, 33, 42, 20],
+            "height": [5, 5, 5, 5],
+        }
+    ).sort("age", descending=descending)
+    q = df0.merge_sorted(df1, key="age")
+    assert_gpu_result_equal(q)
+
+
+@pytest.mark.parametrize(
+    "descending",
+    [
+        pytest.param(
+            True,
+            marks=pytest.mark.xfail(reason="polars/issues/21464 and cudf/issues/18089"),
+        ),
+        False,
+    ],
+)
+def test_merge_sorted_with_nulls(descending):
     df0 = pl.LazyFrame(
         {"name": ["steve", "elise", "bob", "john"], "age": [42, 44, 18, None]}
     ).sort("age", descending=descending)
     df1 = pl.LazyFrame(
         {
             "name": ["anna", "megan", "steve", "thomas", "john"],
             "age": [21, 33, 42, 20, None],
+            "height": [5, 5, 5, 5, 5],
         }
     ).sort("age", descending=descending)
     q = df0.merge_sorted(df1, key="age")