Remove stale python filterwarnings and skipped tests (rapidsai#22706)

* Removes and improves some `filterwarnings` used in unit tests
* Removes `test_magics_cpu` as it's been perpetually skipped without resolution

Authors:
  - Matthew Roeschke (https://github.com/mroeschke)

Approvers:
  - Lawrence Mitchell (https://github.com/wence-)
  - James Lamb (https://github.com/jameslamb)

URL: rapidsai#22706

Author: mroeschke

python/cudf/cudf/tests/input_output/test_feather.py

@@ -33,7 +33,6 @@ def pdf(request):
 
 
 @pytest.mark.filterwarnings("ignore:Using CPU")
-@pytest.mark.filterwarnings("ignore:Strings are not yet supported")
 @pytest.mark.parametrize(
     "columns",
     [["col_int8"], ["col_category"], ["col_int32", "col_float32"], None],

python/cudf/cudf/tests/input_output/test_hdf5.py

@@ -1,4 +1,4 @@
-# SPDX-FileCopyrightText: Copyright (c) 2018-2025, NVIDIA CORPORATION.
+# SPDX-FileCopyrightText: Copyright (c) 2018-2026, NVIDIA CORPORATION.
 # SPDX-License-Identifier: Apache-2.0
 
 import os
@@ -69,7 +69,6 @@ def hdf_files(request, tmp_path, pdf):
 
 
 @pytest.mark.filterwarnings("ignore:Using CPU")
-@pytest.mark.filterwarnings("ignore:Strings are not yet supported")
 @pytest.mark.parametrize(
     "columns",
     [["col_int8"], ["col_category"], ["col_int32", "col_float32"], None],

python/cudf/cudf/tests/input_output/test_json.py

@@ -82,7 +82,6 @@ def gdf_writer_types(request):
     return test_pdf
 
 
-@pytest.mark.filterwarnings("ignore:Strings are not yet supported")
 @pytest.mark.filterwarnings("ignore:Using CPU")
 @pytest.mark.parametrize("index", [True, False])
 # tests limited to compressions formats supported by pandas and cudf: bz2, gzip, zip, zstd

python/cudf/cudf/tests/series/test_binops.py

@@ -3039,7 +3039,6 @@ def test_binop_index_series(arithmetic_op):
     assert_eq(expected, actual)
 
 
-@pytest.mark.filterwarnings("ignore::DeprecationWarning")
 @pytest.mark.parametrize("name1", [None, "name1"])
 @pytest.mark.parametrize("name2", [None, "name2"])
 def test_binop_index_dt_td_series_with_names(name1, name2):

python/cudf/cudf/tests/test_doctests.py

@@ -11,7 +11,6 @@
 
 import cudf
 
-pytestmark = pytest.mark.filterwarnings("ignore::FutureWarning")
 _SKIP_DOCTESTS = frozenset(
     {
         "register_dataframe_accessor",
@@ -88,6 +87,22 @@ def _find_doctests_in_obj(obj, finder=None, criteria=None):
                     yield docstring
 
 
+marks_for_doctests = {
+    "register_dataframe_accessor": pytest.mark.filterwarnings(
+        "ignore:Attribute .* will be overridden:UserWarning"
+    ),
+    "register_index_accessor": pytest.mark.filterwarnings(
+        "ignore:Attribute .* will be overridden:UserWarning"
+    ),
+    "register_series_accessor": pytest.mark.filterwarnings(
+        "ignore:Attribute .* will be overridden:UserWarning"
+    ),
+    "StringMethods.edit_distance_matrix": pytest.mark.filterwarnings(
+        "ignore:edit_distance_matrix:FutureWarning"
+    ),
+}
+
+
 def _collect_doctests():
     """Eagerly collect all unique doctests into a list.
 
@@ -98,7 +113,11 @@ def _collect_doctests():
     results = []
     for mod in tests:
         for docstring in _find_doctests_in_obj(mod):
-            results.append(docstring)
+            if (mark := marks_for_doctests.get(docstring.name)) is not None:
+                doc = pytest.param(docstring, marks=mark)
+            else:
+                doc = docstring
+            results.append(doc)
     return results
 
 
@@ -122,8 +141,6 @@ def printoptions(cls):
         ids=lambda docstring: docstring.name,
     )
     def test_docstring(self, docstring, monkeypatch, tmp_path):
-        if docstring.name in _SKIP_DOCTESTS:
-            pytest.skip(f"{docstring.name} doctest is not runnable")
         # We ignore differences in whitespace in the doctest output, and enable
         # the use of an ellipsis "..." to match any string in the doctest
         # output. An ellipsis is useful for, e.g., memory addresses or

python/cudf/cudf/tests/test_flags.py

@@ -172,17 +172,16 @@ def test_set_flags_no_change_when_none(frame_or_series):
     _assert_flags_eq(cu_new, pd_new)
 
 
+@pytest.mark.filterwarnings(
+    "ignore:The copy keyword is deprecated:pandas.errors.Pandas4Warning"
+)
 def test_set_flags_copy_true_makes_deep_copy():
     # Match pandas' observable behaviour: mutating the copy does not
     # alter the original. ``copy=True`` is deprecated in pandas 3; we
     # still accept it for API parity.
-    import warnings
-
     pdf = pd.DataFrame({"a": [1, 2, 3]})
     gdf = cudf.DataFrame({"a": [1, 2, 3]})
-    with warnings.catch_warnings():
-        warnings.simplefilter("ignore")
-        p_new = pdf.set_flags(copy=True, allows_duplicate_labels=False)
+    p_new = pdf.set_flags(copy=True, allows_duplicate_labels=False)
     g_new = gdf.set_flags(copy=True, allows_duplicate_labels=False)
     p_new.loc[0, "a"] = 99
     g_new.loc[0, "a"] = 99

python/cudf/cudf_pandas_tests/_magics_cpu_test.py

@@ -1,7 +1,6 @@
-# SPDX-FileCopyrightText: Copyright (c) 2023-2025, NVIDIA CORPORATION & AFFILIATES.  All rights reserved.
+# SPDX-FileCopyrightText: Copyright (c) 2023-2026, NVIDIA CORPORATION & AFFILIATES.  All rights reserved.
 # SPDX-License-Identifier: Apache-2.0
 
-import os
 import pathlib
 import subprocess
 import sys
@@ -31,19 +30,3 @@ def test_magics_gpu():
         [sys.executable, LOCATION / "_magics_gpu_test.py"], capture_output=True
     )
     assert sp_completed.stderr.decode() == ""
-
-
-@pytest.mark.skip(
-    "This test was viable when cudf.pandas was separate from cudf, but now "
-    "that it is a subpackage we always require a GPU to be present and cannot "
-    "run this test."
-)
-def test_magics_cpu():
-    env = os.environ.copy()
-    env["CUDA_VISIBLE_DEVICES"] = ""
-    sp_completed = subprocess.run(
-        [sys.executable, LOCATION / "_magics_cpu_test.py"],
-        capture_output=True,
-        env=env,
-    )
-    assert sp_completed.stderr.decode() == ""

python/cudf/cudf_pandas_tests/test_magics.py

@@ -100,8 +100,6 @@ filterwarnings = [
     "error",
     "ignore:::.*xdist.*",
     "ignore:::.*pytest.*",
-    # some third-party dependencies (e.g. 'boto3') still using datetime.datetime.utcnow()
-    "ignore:.*datetime.*utcnow.*scheduled for removal.*:DeprecationWarning:botocore",
     # Deprecation warning from Pyarrow Table.to_pandas() with pandas-2.2+
     "ignore:Passing a BlockManager to DataFrame is deprecated:DeprecationWarning",
     # PerformanceWarning from cupy warming up the JIT cache
@@ -114,7 +112,6 @@ filterwarnings = [
 markers = [
     "spilling: mark benchmark a good candidate to run with `CUDF_SPILL=ON`",
     "serial: Mark a test as not being parallel-safe.",
-    "no_copy_on_write: A test that is not valid when copy-on-write is enabled.",
 ]
 xfail_strict = true
 

python/cudf/pyproject.toml

@@ -82,23 +82,22 @@ def test_merge():
 @pytest.mark.skipif(
     not at_least_n_gpus(2), reason="Machine does not have two GPUs"
 )
+@pytest.mark.filterwarnings("ignore:Port")
 def test_ucx_seriesgroupby():
     pytest.importorskip("distributed_ucxx")
 
-    with warnings.catch_warnings():
-        warnings.filterwarnings("ignore", category=Warning, message="Port")
-        with (
-            dask_cuda.LocalCUDACluster(
-                n_workers=3, dashboard_address=None
-            ) as cluster,
-            cluster.get_client(),
-        ):
-            # Repro Issue#3913
-            df = cudf.DataFrame({"a": [1, 2, 3, 4], "b": [5, 1, 2, 5]})
-            dask_df = dask_cudf.from_cudf(df, npartitions=2)
-            dask_df_g = dask_df.groupby(["a"]).b.sum().compute()
+    with (
+        dask_cuda.LocalCUDACluster(
+            n_workers=3, dashboard_address=None
+        ) as cluster,
+        cluster.get_client(),
+    ):
+        # Repro Issue#3913
+        df = cudf.DataFrame({"a": [1, 2, 3, 4], "b": [5, 1, 2, 5]})
+        dask_df = dask_cudf.from_cudf(df, npartitions=2)
+        dask_df_g = dask_df.groupby(["a"]).b.sum().compute()
 
-            assert dask_df_g.name == "b"
+        assert dask_df_g.name == "b"
 
 
 @pytest.mark.usefixtures("dask_client")
@@ -141,24 +140,22 @@ def test_p2p_shuffle():
     reason="Machine does not have three GPUs",
 )
 @pytest.mark.filterwarnings("ignore::ResourceWarning")
+@pytest.mark.filterwarnings("ignore:Port")
 def test_unique():
     # Using `"p2p"` can produce dispatching problems
     # TODO: Test "p2p" after dask > 2024.4.1 is required
     # See: https://github.com/dask/dask/pull/11040
-    with warnings.catch_warnings():
-        warnings.filterwarnings("ignore", category=Warning, message="Port")
-
-        with dask_cuda.LocalCUDACluster(
-            n_workers=3, dashboard_address=None
-        ) as cluster:
-            with Client(cluster):
-                df = cudf.DataFrame({"x": ["a", "b", "c", "a", "a"]})
-                ddf = dask_cudf.from_cudf(df, npartitions=2)
-                dd.assert_eq(
-                    df.x.unique(),
-                    ddf.x.unique().compute(),
-                    check_index=False,
-                )
+    with dask_cuda.LocalCUDACluster(
+        n_workers=3, dashboard_address=None
+    ) as cluster:
+        with Client(cluster):
+            df = cudf.DataFrame({"x": ["a", "b", "c", "a", "a"]})
+            ddf = dask_cudf.from_cudf(df, npartitions=2)
+            dd.assert_eq(
+                df.x.unique(),
+                ddf.x.unique().compute(),
+                check_index=False,
+            )
 
 
 @pytest.mark.usefixtures("dask_client")