diff --git a/docs/cudf/source/developer_guide/testing.md b/docs/cudf/source/developer_guide/testing.md index f12f809d5db..22cc1b5b8de 100644 --- a/docs/cudf/source/developer_guide/testing.md +++ b/docs/cudf/source/developer_guide/testing.md @@ -7,6 +7,23 @@ specifically the [`pytest-cov`](https://github.com/pytest-dev/pytest-cov) plugin Code coverage reports are uploaded to [Codecov](https://app.codecov.io/gh/rapidsai/cudf). Each PR also indicates whether it increases or decreases test coverage. +### Configuring pytest + +Pytest will accept configuration in [multiple different +files](https://docs.pytest.org/en/stable/reference/customize.html), +with a specified discovery and precedence order. Note in particular +that there is no automatic "include" mechanism, as soon as a matching +configuration file is found, discovery stops. + +For preference, so that all tool configuration lives in the same +place, we use `pyproject.toml`-based configuration. Test configuration +for a given package should live in that package's `pyproject.toml` +file. + +Where tests do not naturally belong to a project, for example the +`cudf.pandas` integration tests and the cuDF benchmarks, use a +`pytest.ini` file as close to the tests as possible. + ## Test organization How tests are organized depends on which of the following two groups they fall into: diff --git a/python/cudf/cudf/tests/pytest.ini b/python/cudf/cudf/tests/pytest.ini deleted file mode 100644 index 496a322ff80..00000000000 --- a/python/cudf/cudf/tests/pytest.ini +++ /dev/null @@ -1,19 +0,0 @@ -# Copyright (c) 2022, NVIDIA CORPORATION. - -[pytest] -markers = - spilling: mark benchmark a good candidate to run with `CUDF_SPILL=ON` -xfail_strict = true -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 - ignore:Jitify is performing a one-time only warm-up to populate the persistent cache:cupy._util.PerformanceWarning - # Ignore numba PEP 456 warning specific to arm machines - ignore:FNV hashing is not implemented in Numba.*:UserWarning -addopts = --tb=native diff --git a/python/cudf/cudf_pandas_tests/pytest.ini b/python/cudf/cudf_pandas_tests/pytest.ini new file mode 100644 index 00000000000..46e2448ea24 --- /dev/null +++ b/python/cudf/cudf_pandas_tests/pytest.ini @@ -0,0 +1,9 @@ +# Copyright (c) 2024, NVIDIA CORPORATION. + +# Note, this config file overrides the default "cudf" test config in +# ../pyproject.toml We do so deliberately because we have different +# treatment of markers and warnings +[pytest] +addopts = --tb=native --strict-config --strict-markers +empty_parameter_set_mark = fail_at_collect +xfail_strict = true diff --git a/python/cudf/cudf_pandas_tests/test_cudf_pandas.py b/python/cudf/cudf_pandas_tests/test_cudf_pandas.py index 2bbed40e34e..a74b7148c00 100644 --- a/python/cudf/cudf_pandas_tests/test_cudf_pandas.py +++ b/python/cudf/cudf_pandas_tests/test_cudf_pandas.py @@ -3,6 +3,7 @@ # SPDX-License-Identifier: Apache-2.0 import collections +import contextlib import copy import datetime import operator @@ -21,10 +22,15 @@ import pyarrow as pa import pytest from nbconvert.preprocessors import ExecutePreprocessor -from numba import NumbaDeprecationWarning, vectorize +from numba import ( + NumbaDeprecationWarning, + __version__ as numba_version, + vectorize, +) +from packaging import version from pytz import utc -from cudf.core._compat import PANDAS_GE_220 +from cudf.core._compat import PANDAS_GE_210, PANDAS_GE_220, PANDAS_VERSION from cudf.pandas import LOADED, Profiler from cudf.pandas.fast_slow_proxy import ( ProxyFallbackError, @@ -52,8 +58,6 @@ get_calendar, ) -from cudf.core._compat import PANDAS_CURRENT_SUPPORTED_VERSION, PANDAS_VERSION - # Accelerated pandas has the real pandas and cudf modules as attributes pd = xpd._fsproxy_slow cudf = xpd._fsproxy_fast @@ -622,10 +626,6 @@ def test_array_function_series_fallback(series): tm.assert_equal(expect, got) -@pytest.mark.xfail( - PANDAS_VERSION < PANDAS_CURRENT_SUPPORTED_VERSION, - reason="Fails in older versions of pandas", -) def test_timedeltaproperties(series): psr, sr = series psr, sr = psr.astype("timedelta64[ns]"), sr.astype("timedelta64[ns]") @@ -685,10 +685,6 @@ def test_maintain_container_subclasses(multiindex): assert isinstance(got, xpd.core.indexes.frozen.FrozenList) -@pytest.mark.xfail( - PANDAS_VERSION < PANDAS_CURRENT_SUPPORTED_VERSION, - reason="Fails in older versions of pandas due to unsupported boxcar window type", -) def test_rolling_win_type(): pdf = pd.DataFrame(range(5)) df = xpd.DataFrame(range(5)) @@ -697,8 +693,14 @@ def test_rolling_win_type(): tm.assert_equal(result, expected) -@pytest.mark.skip( - reason="Requires Numba 0.59 to fix segfaults on ARM. See https://github.com/numba/llvmlite/pull/1009" +@pytest.mark.skipif( + version.parse(numba_version) < version.parse("0.59"), + reason="Requires Numba 0.59 to fix segfaults on ARM. See https://github.com/numba/llvmlite/pull/1009", +) +@pytest.mark.xfail( + version.parse(numba_version) >= version.parse("0.59") + and PANDAS_VERSION < version.parse("2.1"), + reason="numba.generated_jit removed in 0.59, requires pandas >= 2.1", ) def test_rolling_apply_numba_engine(): def weighted_mean(x): @@ -709,7 +711,12 @@ def weighted_mean(x): pdf = pd.DataFrame([[1, 2, 0.6], [2, 3, 0.4], [3, 4, 0.2], [4, 5, 0.7]]) df = xpd.DataFrame([[1, 2, 0.6], [2, 3, 0.4], [3, 4, 0.2], [4, 5, 0.7]]) - with pytest.warns(NumbaDeprecationWarning): + ctx = ( + contextlib.nullcontext() + if PANDAS_GE_210 + else pytest.warns(NumbaDeprecationWarning) + ) + with ctx: expect = pdf.rolling(2, method="table", min_periods=0).apply( weighted_mean, raw=True, engine="numba" ) @@ -1305,7 +1312,7 @@ def max_times_two(self): @pytest.mark.xfail( - PANDAS_VERSION < PANDAS_CURRENT_SUPPORTED_VERSION, + PANDAS_VERSION < version.parse("2.1"), reason="DatetimeArray.__floordiv__ missing in pandas-2.0.0", ) def test_floordiv_array_vs_df(): @@ -1580,7 +1587,7 @@ def test_numpy_cupy_flatiter(series): @pytest.mark.xfail( - PANDAS_VERSION < PANDAS_CURRENT_SUPPORTED_VERSION, + PANDAS_VERSION < version.parse("2.1"), reason="pyarrow_numpy storage type was not supported in pandas-2.0.0", ) def test_arrow_string_arrays(): diff --git a/python/cudf/pyproject.toml b/python/cudf/pyproject.toml index 1b730ffd13c..c0776fd0de6 100644 --- a/python/cudf/pyproject.toml +++ b/python/cudf/pyproject.toml @@ -124,6 +124,27 @@ skip = [ "__init__.py", ] +[tool.pytest.ini_options] +addopts = "--tb=native --strict-config --strict-markers" +empty_parameter_set_mark = "fail_at_collect" +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 + "ignore:Jitify is performing a one-time only warm-up to populate the persistent cache:cupy._util.PerformanceWarning", + # Ignore numba PEP 456 warning specific to arm machines + "ignore:FNV hashing is not implemented in Numba.*:UserWarning" +] +markers = [ + "spilling: mark benchmark a good candidate to run with `CUDF_SPILL=ON`" +] +xfail_strict = true + [tool.rapids-build-backend] build-backend = "scikit_build_core.build" dependencies-file = "../../dependencies.yaml" diff --git a/python/cudf_kafka/cudf_kafka/tests/pytest.ini b/python/cudf_kafka/cudf_kafka/tests/pytest.ini deleted file mode 100644 index 7b0a9f29fb1..00000000000 --- a/python/cudf_kafka/cudf_kafka/tests/pytest.ini +++ /dev/null @@ -1,4 +0,0 @@ -# Copyright (c) 2024, NVIDIA CORPORATION. - -[pytest] -addopts = --tb=native diff --git a/python/cudf_kafka/pyproject.toml b/python/cudf_kafka/pyproject.toml index a1a3ec37842..87e19a2bccf 100644 --- a/python/cudf_kafka/pyproject.toml +++ b/python/cudf_kafka/pyproject.toml @@ -79,9 +79,12 @@ skip = [ ] [tool.pytest.ini_options] +addopts = "--tb=native --strict-config --strict-markers" +empty_parameter_set_mark = "fail_at_collect" filterwarnings = [ "error" ] +xfail_strict = true [tool.scikit-build] build-dir = "build/{wheel_tag}" diff --git a/python/cudf_polars/pyproject.toml b/python/cudf_polars/pyproject.toml index f55031e0826..5345fad41a2 100644 --- a/python/cudf_polars/pyproject.toml +++ b/python/cudf_polars/pyproject.toml @@ -50,6 +50,11 @@ license-files = ["LICENSE"] version = {file = "cudf_polars/VERSION"} [tool.pytest.ini_options] +addopts = "--tb=native --strict-config --strict-markers" +empty_parameter_set_mark = "fail_at_collect" +filterwarnings = [ + "error" +] xfail_strict = true [tool.coverage.report] diff --git a/python/cudf_polars/tests/pytest.ini b/python/cudf_polars/tests/pytest.ini deleted file mode 100644 index 7b0a9f29fb1..00000000000 --- a/python/cudf_polars/tests/pytest.ini +++ /dev/null @@ -1,4 +0,0 @@ -# Copyright (c) 2024, NVIDIA CORPORATION. - -[pytest] -addopts = --tb=native diff --git a/python/custreamz/custreamz/tests/pytest.ini b/python/custreamz/custreamz/tests/pytest.ini deleted file mode 100644 index 7b0a9f29fb1..00000000000 --- a/python/custreamz/custreamz/tests/pytest.ini +++ /dev/null @@ -1,4 +0,0 @@ -# Copyright (c) 2024, NVIDIA CORPORATION. - -[pytest] -addopts = --tb=native diff --git a/python/custreamz/custreamz/tests/test_dataframes.py b/python/custreamz/custreamz/tests/test_dataframes.py index bae4b051cae..8c0130d2818 100644 --- a/python/custreamz/custreamz/tests/test_dataframes.py +++ b/python/custreamz/custreamz/tests/test_dataframes.py @@ -377,24 +377,16 @@ def test_setitem_overwrites(stream): [ ({}, "sum"), ({}, "mean"), - pytest.param({}, "min"), + ({}, "min"), pytest.param( {}, "median", marks=pytest.mark.xfail(reason="Unavailable for rolling objects"), ), - pytest.param({}, "max"), - pytest.param( - {}, - "var", - marks=pytest.mark.xfail(reason="Unavailable for rolling objects"), - ), - pytest.param({}, "count"), - pytest.param( - {"ddof": 0}, - "std", - marks=pytest.mark.xfail(reason="Unavailable for rolling objects"), - ), + ({}, "max"), + ({}, "var"), + ({}, "count"), + ({"ddof": 0}, "std"), pytest.param( {"quantile": 0.5}, "quantile", diff --git a/python/custreamz/pyproject.toml b/python/custreamz/pyproject.toml index 85ab0024bb5..af45f49d9b4 100644 --- a/python/custreamz/pyproject.toml +++ b/python/custreamz/pyproject.toml @@ -111,6 +111,8 @@ skip = [ ] [tool.pytest.ini_options] +addopts = "--tb=native --strict-config --strict-markers" +empty_parameter_set_mark = "fail_at_collect" filterwarnings = [ "error", "ignore:unclosed