Assume unmerged spec behaviour

data-apis/dataframe-api#74

Author: honno

tests/conftest.py

@@ -1,3 +1,4 @@
+import re
 from typing import Union
 
 import pytest
@@ -38,17 +39,15 @@ def pytest_configure(config):
 
 
 ci_failing_ids = [
-    # dataframe objects return the interchange dataframe, not a dict, although
-    # this is the behaviour that should be in the spec soon.
-    # See https://github.com/data-apis/dataframe-api/pull/74
-    "test_dataframe_object.py::test_toplevel_dunder_dataframe[pandas]",
-    "test_dataframe_object.py::test_toplevel_dunder_dataframe[vaex]",
-    "test_dataframe_object.py::test_toplevel_dunder_dataframe[modin]",
-    "test_dataframe_object.py::test_dunder_dataframe[pandas]",
-    "test_dataframe_object.py::test_dunder_dataframe[modin]",
-    # vaex's interchange dataframe doesn't have __dataframe__()
+    # vaex's and cudf's interchange dataframe doesn't have __dataframe__()
+    # See https://github.com/data-apis/dataframe-api/issues/80
     "test_dataframe_object.py::test_dunder_dataframe[vaex]",
     "test_signatures.py::test_dataframe_method[vaex-__dataframe__]",
+    "test_dataframe_object.py::test_dunder_dataframe[cudf]",
+    "test_signatures.py::test_dataframe_method[cudf-__dataframe__]",
+    # https://github.com/rapidsai/cudf/issues/11320
+    "test_signatures.py::test_buffer_method[cudf-__dlpack__]",
+    "test_signatures.py::test_buffer_method[cudf-__dlpack_device__]",
     # https://github.com/vaexio/vaex/issues/2083
     # https://github.com/vaexio/vaex/issues/2093
     # https://github.com/vaexio/vaex/issues/2113
@@ -59,11 +58,13 @@ def pytest_configure(config):
     "test_column_object.py::test_size[vaex]",
     # https://github.com/vaexio/vaex/issues/2118
     "test_column_object.py::test_dtype[vaex]",
-    # Raises TypeError as opposed to RuntimeError, although this is the
-    # behaviour that should be in the spec soon.
+    # Raises RuntimeError, which is technically correct, but the spec will
+    # require TypeError soon.
     # See https://github.com/data-apis/dataframe-api/pull/74
-    "test_column_object.py::test_describe_categorical[pandas]",
+    "test_column_object.py::test_describe_categorical[modin]",
+    # https://github.com/vaexio/vaex/issues/2113
     "test_column_object.py::test_describe_categorical[vaex]",
+    # https://github.com/rapidsai/cudf/issues/11332
     "test_column_object.py::test_describe_categorical[cudf]",
     # https://github.com/pandas-dev/pandas/issues/47789
     "test_column_object.py::test_null_count[pandas]",
@@ -79,9 +80,13 @@ def pytest_configure(config):
     "test_column_object.py::test_get_buffers[cudf]",
 ]
 
+r_cudf_roundtrip = re.compile(r"test_from_dataframe_roundtrip\[.*cudf.*\]")
+
 
 def pytest_collection_modifyitems(config, items):
     if config.getoption("--ci"):
         for item in items:
             if any(id_ in item.nodeid for id_ in ci_failing_ids):
                 item.add_marker(pytest.mark.xfail())
+            elif r_cudf_roundtrip.search(item.nodeid):
+                item.add_marker(pytest.mark.skip("TODO"))

tests/test_column_object.py

@@ -99,7 +99,7 @@ def test_describe_categorical(libinfo: LibraryInfo, data: st.DataObject):
             if mapping is not None:
                 assert isinstance(mapping, dict)
         else:
-            with pytest.raises(RuntimeError):
+            with pytest.raises(TypeError):
                 col.describe_categorical
 
 

tests/test_dataframe_object.py

@@ -11,20 +11,14 @@
 def test_toplevel_dunder_dataframe(libinfo: LibraryInfo, data: st.DataObject):
     df = data.draw(libinfo.toplevel_dataframes(), label="df")
     assert hasattr(df, "__dataframe__")
-    out = df.__dataframe__()
-    assert isinstance(out, dict)
-    assert hasattr(out, "dataframe")
-    assert hasattr(out, "version")
+    df.__dataframe__()
 
 
 @given(data=st.data())
 def test_dunder_dataframe(libinfo: LibraryInfo, data: st.DataObject):
     df = data.draw(libinfo.interchange_dataframes(), label="df")
     assert hasattr(df, "__dataframe__")
-    out = df.__dataframe__()
-    assert isinstance(out, dict)
-    assert hasattr(out, "dataframe")
-    assert hasattr(out, "version")
+    df.__dataframe__()
 
 
 @given(data=st.data())

tests/wrappers.py

@@ -19,15 +19,13 @@ class LibraryInfo(NamedTuple):
     mock_to_toplevel: Callable[[MockDataFrame], TopLevelDataFrame]
     from_dataframe: Callable[[TopLevelDataFrame], DataFrame]
     frame_equal: Callable[[TopLevelDataFrame, DataFrame], bool]
-    toplevel_to_interchange: Callable[[TopLevelDataFrame], DataFrame] = lambda df: (
-        df.__dataframe__()["dataframe"]
-    )
     exclude_dtypes: List[NominalDtype] = []
     allow_zero_cols: bool = True
     allow_zero_rows: bool = True
 
     def mock_to_interchange(self, mock_dataframe: MockDataFrame) -> DataFrame:
-        return self.toplevel_to_interchange(self.mock_to_toplevel(mock_dataframe))
+        toplevel_df = self.mock_to_toplevel(mock_dataframe)
+        return toplevel_df.__dataframe__()
 
     @property
     def mock_dataframes_kwargs(self) -> Dict[str, Any]:
@@ -44,7 +42,7 @@ def toplevel_dataframes(self) -> st.SearchStrategy[TopLevelDataFrame]:
         return self.mock_dataframes().map(self.mock_to_toplevel)
 
     def interchange_dataframes(self) -> st.SearchStrategy[TopLevelDataFrame]:
-        return self.toplevel_dataframes().map(self.toplevel_to_interchange)
+        return self.toplevel_dataframes().map(lambda df: df.__dataframe__())
 
     def __repr__(self) -> str:
         return f"LibraryInfo(<{self.name}>)"
@@ -82,7 +80,6 @@ def pandas_mock_to_toplevel(mock_df: MockDataFrame) -> pd.DataFrame:
         mock_to_toplevel=pandas_mock_to_toplevel,
         from_dataframe=pandas_from_dataframe,
         frame_equal=lambda df1, df2: df1.equals(df2),
-        toplevel_to_interchange=lambda df: df.__dataframe__(),
         exclude_dtypes=[NominalDtype.DATETIME64NS],
     )
     libinfo_params.append(pytest.param(pandas_libinfo, id=pandas_libinfo.name))
@@ -140,7 +137,6 @@ def vaex_frame_equal(df1, df2) -> bool:
         mock_to_toplevel=vaex_mock_to_toplevel,
         from_dataframe=vaex_from_dataframe,
         frame_equal=vaex_frame_equal,
-        toplevel_to_interchange=lambda df: df.__dataframe__(),
         exclude_dtypes=[NominalDtype.DATETIME64NS],
         # https://github.com/vaexio/vaex/issues/2094
         allow_zero_cols=False,
@@ -220,7 +216,6 @@ def modin_frame_equal(df1: mpd.DataFrame, df2: mpd.DataFrame) -> bool:
         mock_to_toplevel=modin_mock_to_toplevel,
         from_dataframe=modin_from_dataframe,
         frame_equal=modin_frame_equal,
-        toplevel_to_interchange=lambda df: df.__dataframe__(),
         # https://github.com/modin-project/modin/issues/4654
         # https://github.com/modin-project/modin/issues/4652
         exclude_dtypes=[
@@ -297,7 +292,6 @@ def cudf_mock_to_toplevel(mock_df: MockDataFrame) -> cudf.DataFrame:
         mock_to_toplevel=cudf_mock_to_toplevel,
         from_dataframe=cudf_from_dataframe,
         frame_equal=lambda df1, df2: df1.equals(df2),  # NaNs considered equal
-        toplevel_to_interchange=lambda df: df.__dataframe__(),
         exclude_dtypes=[NominalDtype.DATETIME64NS],
     )
     libinfo_params.append(pytest.param(cupy_libinfo, id=cupy_libinfo.name))