CI: Test against old versions of key dependencies (rapidsai#16570)

This adds explicit tests with old versions of key dependencies. Specifically:
- `numba==0.57`
- `numpy==1.23`
- `pandas==2.0`
- ~`fsspec==0.6.0`~ excluded it. `transformers==4.39.3` requires `huggingface_hub` which requires `fsspec>=2023.5.0`.  In principle one could include it e.g. only for conda which doesn't pull in `transformers`, but that seemed not worth the trouble?
- `cupy==12.0.0`
- `pyarrow==16.1.0`

See also rapidsai/build-planning#81

(Marking as draft until I see that things work.)

Authors:
  - Sebastian Berg (https://github.com/seberg)
  - Matthew Roeschke (https://github.com/mroeschke)
  - GALI PREM SAGAR (https://github.com/galipremsagar)

Approvers:
  - Matthew Roeschke (https://github.com/mroeschke)
  - Bradley Dice (https://github.com/bdice)
  - Vyas Ramasubramani (https://github.com/vyasr)
  - Charles Blackmon-Luca (https://github.com/charlesbluca)

URL: rapidsai#16570

ci/cudf_pandas_scripts/run_tests.sh

@@ -54,8 +54,19 @@ else
     RAPIDS_PY_WHEEL_NAME="libcudf_${RAPIDS_PY_CUDA_SUFFIX}" rapids-download-wheels-from-s3 cpp ./dist
     RAPIDS_PY_WHEEL_NAME="pylibcudf_${RAPIDS_PY_CUDA_SUFFIX}" rapids-download-wheels-from-s3 python ./dist
 
-    # echo to expand wildcard before adding `[extra]` requires for pip
+    echo "" > ./constraints.txt
+    if [[ $RAPIDS_DEPENDENCIES == "oldest" ]]; then
+        # `test_python` constraints are for `[test]` not `[cudf-pandas-tests]`
+        rapids-dependency-file-generator \
+            --output requirements \
+            --file-key test_python \
+            --matrix "cuda=${RAPIDS_CUDA_VERSION%.*};arch=$(arch);py=${RAPIDS_PY_VERSION};dependencies=${RAPIDS_DEPENDENCIES}" \
+        | tee ./constraints.txt
+    fi
+
     python -m pip install \
+        -v \
+        --constraint ./constraints.txt \
         "$(echo ./dist/cudf_${RAPIDS_PY_CUDA_SUFFIX}*.whl)[test,cudf-pandas-tests]" \
         "$(echo ./dist/libcudf_${RAPIDS_PY_CUDA_SUFFIX}*.whl)" \
         "$(echo ./dist/pylibcudf_${RAPIDS_PY_CUDA_SUFFIX}*.whl)"

ci/test_python_common.sh

@@ -14,7 +14,8 @@ ENV_YAML_DIR="$(mktemp -d)"
 rapids-dependency-file-generator \
   --output conda \
   --file-key test_python \
-  --matrix "cuda=${RAPIDS_CUDA_VERSION%.*};arch=$(arch);py=${RAPIDS_PY_VERSION}" | tee "${ENV_YAML_DIR}/env.yaml"
+  --matrix "cuda=${RAPIDS_CUDA_VERSION%.*};arch=$(arch);py=${RAPIDS_PY_VERSION};dependencies=${RAPIDS_DEPENDENCIES}" \
+    | tee "${ENV_YAML_DIR}/env.yaml"
 
 rapids-mamba-retry env create --yes -f "${ENV_YAML_DIR}/env.yaml" -n test
 

ci/test_wheel_cudf.sh

@@ -10,8 +10,22 @@ RAPIDS_PY_WHEEL_NAME="cudf_${RAPIDS_PY_CUDA_SUFFIX}" rapids-download-wheels-from
 RAPIDS_PY_WHEEL_NAME="libcudf_${RAPIDS_PY_CUDA_SUFFIX}" rapids-download-wheels-from-s3 cpp ./dist
 RAPIDS_PY_WHEEL_NAME="pylibcudf_${RAPIDS_PY_CUDA_SUFFIX}" rapids-download-wheels-from-s3 python ./dist
 
+rapids-logger "Install cudf, pylibcudf, and test requirements"
+
+# Constrain to minimum dependency versions if job is set up as "oldest"
+echo "" > ./constraints.txt
+if [[ $RAPIDS_DEPENDENCIES == "oldest" ]]; then
+    rapids-dependency-file-generator \
+        --output requirements \
+        --file-key py_test_cudf \
+        --matrix "cuda=${RAPIDS_CUDA_VERSION%.*};arch=$(arch);py=${RAPIDS_PY_VERSION};dependencies=${RAPIDS_DEPENDENCIES}" \
+      | tee ./constraints.txt
+fi
+
 # echo to expand wildcard before adding `[extra]` requires for pip
 python -m pip install \
+    -v \
+    --constraint ./constraints.txt \
   "$(echo ./dist/cudf_${RAPIDS_PY_CUDA_SUFFIX}*.whl)[test]" \
   "$(echo ./dist/libcudf_${RAPIDS_PY_CUDA_SUFFIX}*.whl)" \
   "$(echo ./dist/pylibcudf_${RAPIDS_PY_CUDA_SUFFIX}*.whl)[test]"

ci/test_wheel_cudf_polars.sh

@@ -25,9 +25,20 @@ RAPIDS_PY_WHEEL_NAME="libcudf_${RAPIDS_PY_CUDA_SUFFIX}" rapids-download-wheels-f
 RAPIDS_PY_WHEEL_NAME="pylibcudf_${RAPIDS_PY_CUDA_SUFFIX}" rapids-download-wheels-from-s3 python ./dist
 
 rapids-logger "Installing cudf_polars and its dependencies"
+# Constraint to minimum dependency versions if job is set up as "oldest"
+echo "" > ./constraints.txt
+if [[ $RAPIDS_DEPENDENCIES == "oldest" ]]; then
+    rapids-dependency-file-generator \
+        --output requirements \
+        --file-key py_test_cudf_polars \
+        --matrix "cuda=${RAPIDS_CUDA_VERSION%.*};arch=$(arch);py=${RAPIDS_PY_VERSION};dependencies=${RAPIDS_DEPENDENCIES}" \
+      | tee ./constraints.txt
+fi
 
 # echo to expand wildcard before adding `[extra]` requires for pip
 python -m pip install \
+    -v \
+    --constraint ./constraints.txt \
     "$(echo ./dist/cudf_polars_${RAPIDS_PY_CUDA_SUFFIX}*.whl)[test]" \
     "$(echo ./dist/libcudf_${RAPIDS_PY_CUDA_SUFFIX}*.whl)" \
     "$(echo ./dist/pylibcudf_${RAPIDS_PY_CUDA_SUFFIX}*.whl)"

ci/test_wheel_dask_cudf.sh

@@ -11,8 +11,21 @@ RAPIDS_PY_WHEEL_NAME="cudf_${RAPIDS_PY_CUDA_SUFFIX}" rapids-download-wheels-from
 RAPIDS_PY_WHEEL_NAME="libcudf_${RAPIDS_PY_CUDA_SUFFIX}" rapids-download-wheels-from-s3 cpp ./dist
 RAPIDS_PY_WHEEL_NAME="pylibcudf_${RAPIDS_PY_CUDA_SUFFIX}" rapids-download-wheels-from-s3 python ./dist
 
+rapids-logger "Install dask_cudf, cudf, pylibcudf, and test requirements"
+# Constraint to minimum dependency versions if job is set up as "oldest"
+echo "" > ./constraints.txt
+if [[ $RAPIDS_DEPENDENCIES == "oldest" ]]; then
+    rapids-dependency-file-generator \
+        --output requirements \
+        --file-key py_test_dask_cudf \
+        --matrix "cuda=${RAPIDS_CUDA_VERSION%.*};arch=$(arch);py=${RAPIDS_PY_VERSION};dependencies=${RAPIDS_DEPENDENCIES}" \
+      | tee ./constraints.txt
+fi
+
 # echo to expand wildcard before adding `[extra]` requires for pip
 python -m pip install \
+  -v \
+  --constraint ./constraints.txt \
   "$(echo ./dist/cudf_${RAPIDS_PY_CUDA_SUFFIX}*.whl)" \
   "$(echo ./dist/dask_cudf_${RAPIDS_PY_CUDA_SUFFIX}*.whl)[test]" \
   "$(echo ./dist/libcudf_${RAPIDS_PY_CUDA_SUFFIX}*.whl)" \

dependencies.yaml

@@ -696,6 +696,28 @@ dependencies:
           - pytest<8
           - pytest-cov
           - pytest-xdist
+    specific:
+      # Define additional constraints for testing with oldest dependencies.
+      - output_types: [conda, requirements]
+        matrices:
+          - matrix: {dependencies: "oldest"}
+            packages:
+              - numba==0.57.*
+              - numpy==1.23.*
+              - pandas==2.0.*
+              - pyarrow==14.0.0
+              - cupy==12.0.0  # ignored as pip constraint
+          - matrix:
+            packages:
+      - output_types: requirements
+        # Using --constraints for pip install, so we list cupy multiple times
+        matrices:
+          - matrix: {dependencies: "oldest"}
+            packages:
+              - cupy-cuda11x==12.0.0
+              - cupy-cuda12x==12.0.0
+          - matrix:
+            packages:
   test_python_pylibcudf:
     common:
       - output_types: [conda, requirements, pyproject]

python/cudf/cudf/tests/indexes/test_interval.py

@@ -149,6 +149,10 @@ def test_interval_range_periods_basic_dtype(start_t, end_t, periods_t):
     assert_eq(pindex, gindex)
 
 
+@pytest.mark.skipif(
+    PANDAS_VERSION < PANDAS_CURRENT_SUPPORTED_VERSION,
+    reason="Does not warn on older versions of pandas",
+)
 def test_interval_range_periods_warnings():
     start_val, end_val, periods_val = 0, 4, 1.0
 

python/cudf/cudf/tests/test_avro_reader_fastavro_integration.py

@@ -23,6 +23,7 @@
 import pytest
 
 import cudf
+from cudf.core._compat import PANDAS_CURRENT_SUPPORTED_VERSION, PANDAS_VERSION
 from cudf.testing import assert_eq
 from cudf.testing.dataset_generator import rand_dataframe
 
@@ -302,6 +303,10 @@ def get_days_from_epoch(date: datetime.date | None) -> int | None:
 @pytest.mark.parametrize("namespace", [None, "root_ns"])
 @pytest.mark.parametrize("nullable", [True, False])
 @pytest.mark.parametrize("prepend_null", [True, False])
+@pytest.mark.skipif(
+    PANDAS_VERSION < PANDAS_CURRENT_SUPPORTED_VERSION,
+    reason="Fails in older versions of pandas (datetime(9999, ...) too large)",
+)
 def test_can_parse_avro_date_logical_type(namespace, nullable, prepend_null):
     avro_type = {"logicalType": "date", "type": "int"}
     if nullable:

python/cudf/cudf/tests/test_binops.py

@@ -13,7 +13,11 @@
 
 import cudf
 from cudf import Index, Series
-from cudf.core._compat import PANDAS_CURRENT_SUPPORTED_VERSION, PANDAS_VERSION
+from cudf.core._compat import (
+    PANDAS_CURRENT_SUPPORTED_VERSION,
+    PANDAS_GE_220,
+    PANDAS_VERSION,
+)
 from cudf.core.buffer.spill_manager import get_global_manager
 from cudf.testing import _utils as utils, assert_eq
 from cudf.utils.dtypes import (
@@ -1781,6 +1785,20 @@ def test_datetime_dateoffset_binaryop(
             reason="https://github.com/pandas-dev/pandas/issues/57448",
         )
     )
+    if (
+        not PANDAS_GE_220
+        and dtype in {"datetime64[ms]", "datetime64[s]"}
+        and frequency in ("microseconds", "nanoseconds")
+        and n_periods != 0
+    ):
+        pytest.skip(reason="https://github.com/pandas-dev/pandas/pull/55595")
+    if (
+        not PANDAS_GE_220
+        and dtype == "datetime64[us]"
+        and frequency == "nanoseconds"
+        and n_periods != 0
+    ):
+        pytest.skip(reason="https://github.com/pandas-dev/pandas/pull/55595")
 
     date_col = [
         f"2000-01-01 00:00:{components}",
@@ -1834,7 +1852,11 @@ def test_datetime_dateoffset_binaryop(
     "ignore:Discarding nonzero nanoseconds:UserWarning"
 )
 @pytest.mark.parametrize("op", [operator.add, operator.sub])
-def test_datetime_dateoffset_binaryop_multiple(date_col, kwargs, op):
+@pytest.mark.skipif(
+    PANDAS_VERSION < PANDAS_CURRENT_SUPPORTED_VERSION,
+    reason="Fails in older versions of pandas",
+)
+def test_datetime_dateoffset_binaryop_multiple(request, date_col, kwargs, op):
     gsr = cudf.Series(date_col, dtype="datetime64[ns]")
     psr = gsr.to_pandas()
 
@@ -1873,6 +1895,21 @@ def test_datetime_dateoffset_binaryop_multiple(date_col, kwargs, op):
 def test_datetime_dateoffset_binaryop_reflected(
     n_periods, frequency, dtype, components
 ):
+    if (
+        not PANDAS_GE_220
+        and dtype in {"datetime64[ms]", "datetime64[s]"}
+        and frequency in ("microseconds", "nanoseconds")
+        and n_periods != 0
+    ):
+        pytest.skip(reason="https://github.com/pandas-dev/pandas/pull/55595")
+    if (
+        not PANDAS_GE_220
+        and dtype == "datetime64[us]"
+        and frequency == "nanoseconds"
+        and n_periods != 0
+    ):
+        pytest.skip(reason="https://github.com/pandas-dev/pandas/pull/55595")
+
     date_col = [
         f"2000-01-01 00:00:{components}",
         f"2000-01-31 00:00:{components}",

python/cudf/cudf/tests/test_categorical.py

@@ -11,6 +11,7 @@
 import pytest
 
 import cudf
+from cudf.core._compat import PANDAS_CURRENT_SUPPORTED_VERSION, PANDAS_VERSION
 from cudf.testing import assert_eq
 from cudf.testing._utils import NUMERIC_TYPES, assert_exceptions_equal
 
@@ -858,6 +859,10 @@ def test_cat_from_scalar(scalar):
     assert_eq(ps, gs)
 
 
+@pytest.mark.skipif(
+    PANDAS_VERSION < PANDAS_CURRENT_SUPPORTED_VERSION,
+    reason="Does not warn on older versions of pandas",
+)
 def test_cat_groupby_fillna():
     ps = pd.Series(["a", "b", "c"], dtype="category")
     gs = cudf.from_pandas(ps)

python/cudf/cudf/tests/test_concat.py

@@ -9,6 +9,7 @@
 import pytest
 
 import cudf
+from cudf.core._compat import PANDAS_GE_220
 from cudf.core.dtypes import Decimal32Dtype, Decimal64Dtype, Decimal128Dtype
 from cudf.testing import assert_eq
 from cudf.testing._utils import assert_exceptions_equal, expect_warning_if
@@ -451,45 +452,75 @@ def test_concat_mixed_input():
         [pd.Series([1, 2, 3]), pd.DataFrame({"a": []})],
         [pd.Series([], dtype="float64"), pd.DataFrame({"a": []})],
         [pd.Series([], dtype="float64"), pd.DataFrame({"a": [1, 2]})],
-        [
-            pd.Series([1, 2, 3.0, 1.2], name="abc"),
-            pd.DataFrame({"a": [1, 2]}),
-        ],
-        [
-            pd.Series(
-                [1, 2, 3.0, 1.2], name="abc", index=[100, 110, 120, 130]
-            ),
-            pd.DataFrame({"a": [1, 2]}),
-        ],
-        [
-            pd.Series(
-                [1, 2, 3.0, 1.2], name="abc", index=["a", "b", "c", "d"]
+        pytest.param(
+            [
+                pd.Series([1, 2, 3.0, 1.2], name="abc"),
+                pd.DataFrame({"a": [1, 2]}),
+            ],
+            marks=pytest.mark.skipif(
+                not PANDAS_GE_220,
+                reason="https://github.com/pandas-dev/pandas/pull/56365",
             ),
-            pd.DataFrame({"a": [1, 2]}, index=["a", "b"]),
-        ],
-        [
-            pd.Series(
-                [1, 2, 3.0, 1.2, 8, 100],
-                name="New name",
-                index=["a", "b", "c", "d", "e", "f"],
+        ),
+        pytest.param(
+            [
+                pd.Series(
+                    [1, 2, 3.0, 1.2], name="abc", index=[100, 110, 120, 130]
+                ),
+                pd.DataFrame({"a": [1, 2]}),
+            ],
+            marks=pytest.mark.skipif(
+                not PANDAS_GE_220,
+                reason="https://github.com/pandas-dev/pandas/pull/56365",
             ),
-            pd.DataFrame(
-                {"a": [1, 2, 4, 10, 11, 12]},
-                index=["a", "b", "c", "d", "e", "f"],
+        ),
+        pytest.param(
+            [
+                pd.Series(
+                    [1, 2, 3.0, 1.2], name="abc", index=["a", "b", "c", "d"]
+                ),
+                pd.DataFrame({"a": [1, 2]}, index=["a", "b"]),
+            ],
+            marks=pytest.mark.skipif(
+                not PANDAS_GE_220,
+                reason="https://github.com/pandas-dev/pandas/pull/56365",
             ),
-        ],
-        [
-            pd.Series(
-                [1, 2, 3.0, 1.2, 8, 100],
-                name="New name",
-                index=["a", "b", "c", "d", "e", "f"],
+        ),
+        pytest.param(
+            [
+                pd.Series(
+                    [1, 2, 3.0, 1.2, 8, 100],
+                    name="New name",
+                    index=["a", "b", "c", "d", "e", "f"],
+                ),
+                pd.DataFrame(
+                    {"a": [1, 2, 4, 10, 11, 12]},
+                    index=["a", "b", "c", "d", "e", "f"],
+                ),
+            ],
+            marks=pytest.mark.skipif(
+                not PANDAS_GE_220,
+                reason="https://github.com/pandas-dev/pandas/pull/56365",
             ),
-            pd.DataFrame(
-                {"a": [1, 2, 4, 10, 11, 12]},
-                index=["a", "b", "c", "d", "e", "f"],
+        ),
+        pytest.param(
+            [
+                pd.Series(
+                    [1, 2, 3.0, 1.2, 8, 100],
+                    name="New name",
+                    index=["a", "b", "c", "d", "e", "f"],
+                ),
+                pd.DataFrame(
+                    {"a": [1, 2, 4, 10, 11, 12]},
+                    index=["a", "b", "c", "d", "e", "f"],
+                ),
+            ]
+            * 7,
+            marks=pytest.mark.skipif(
+                not PANDAS_GE_220,
+                reason="https://github.com/pandas-dev/pandas/pull/56365",
             ),
-        ]
-        * 7,
+        ),
     ],
 )
 def test_concat_series_dataframe_input(objs):

python/cudf/cudf/tests/test_csv.py

@@ -16,9 +16,13 @@
 
 import cudf
 from cudf import read_csv
-from cudf.core._compat import PANDAS_CURRENT_SUPPORTED_VERSION, PANDAS_VERSION
+from cudf.core._compat import (
+    PANDAS_CURRENT_SUPPORTED_VERSION,
+    PANDAS_GE_220,
+    PANDAS_VERSION,
+)
 from cudf.testing import assert_eq
-from cudf.testing._utils import assert_exceptions_equal
+from cudf.testing._utils import assert_exceptions_equal, expect_warning_if
 
 
 def make_numeric_dataframe(nrows, dtype):
@@ -1270,14 +1274,14 @@ def test_csv_reader_delim_whitespace():
     # with header row
     with pytest.warns(FutureWarning):
         cu_df = read_csv(StringIO(buffer), delim_whitespace=True)
-    with pytest.warns(FutureWarning):
+    with expect_warning_if(PANDAS_GE_220):
         pd_df = pd.read_csv(StringIO(buffer), delim_whitespace=True)
     assert_eq(pd_df, cu_df)
 
     # without header row
     with pytest.warns(FutureWarning):
         cu_df = read_csv(StringIO(buffer), delim_whitespace=True, header=None)
-    with pytest.warns(FutureWarning):
+    with expect_warning_if(PANDAS_GE_220):
         pd_df = pd.read_csv(
             StringIO(buffer), delim_whitespace=True, header=None
         )