Skip to content

Commit

Permalink
Reduce number of tests marked spilling (#12197)
Browse files Browse the repository at this point in the history 
To save CI running time, this PR reduce the tests marked `spilling` drastically. 

An alternative to #12187

Authors:
   - Mads R. B. Kristensen (https://github.com/madsbk)

Approvers:
   - https://github.com/brandon-b-miller
   - GALI PREM SAGAR (https://github.com/galipremsagar)
  • Loading branch information
@madsbk
madsbk authored Nov 18, 2022
1 parent 3c94071 commit a2f69e4
Show file tree
Hide file tree
Showing 3 changed files with 231 additions and 163 deletions.
253 changes: 134 additions & 119 deletions python/cudf/cudf/tests/test_binops.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -14,6 +14,7 @@
import cudf
from cudf import Series
from cudf.core._compat import PANDAS_GE_150
from cudf.core.buffer.spill_manager import get_global_manager
from cudf.core.index import as_index
from cudf.testing import _utils as utils
from cudf.utils.dtypes import (
Expand All @@ -27,7 +28,6 @@

STRING_TYPES = {"str"}


_binops = [
operator.add,
operator.sub,
Expand All @@ -47,8 +47,131 @@
operator.ge,
]

_bitwise_binops = [operator.and_, operator.or_, operator.xor]

_int_types = [
"int8",
"int16",
"int32",
"int64",
"uint8",
"uint16",
"uint32",
]

_cmpops = [
operator.lt,
operator.gt,
operator.le,
operator.ge,
operator.eq,
operator.ne,
]

_reflected_ops = [
lambda x: 1 + x,
lambda x: 2 * x,
lambda x: 2 - x,
lambda x: 2 // x,
lambda x: 2 / x,
lambda x: 3 + x,
lambda x: 3 * x,
lambda x: 3 - x,
lambda x: 3 // x,
lambda x: 3 / x,
lambda x: 3 % x,
lambda x: -1 + x,
lambda x: -2 * x,
lambda x: -2 - x,
lambda x: -2 // x,
lambda x: -2 / x,
lambda x: -3 + x,
lambda x: -3 * x,
lambda x: -3 - x,
lambda x: -3 // x,
lambda x: -3 / x,
lambda x: -3 % x,
lambda x: 0 + x,
lambda x: 0 * x,
lambda x: 0 - x,
lambda x: 0 // x,
lambda x: 0 / x,
]

_operators_arithmetic = [
"add",
"radd",
"sub",
"rsub",
"mul",
"rmul",
"mod",
"rmod",
"pow",
"rpow",
"div",
"divide",
"floordiv",
"rfloordiv",
"truediv",
"rtruediv",
]

_operators_comparison = ["eq", "ne", "lt", "le", "gt", "ge"]


_cudf_scalar_reflected_ops = [
lambda x: cudf.Scalar(1) + x,
lambda x: cudf.Scalar(2) * x,
lambda x: cudf.Scalar(2) - x,
lambda x: cudf.Scalar(2) // x,
lambda x: cudf.Scalar(2) / x,
lambda x: cudf.Scalar(3) + x,
lambda x: cudf.Scalar(3) * x,
lambda x: cudf.Scalar(3) - x,
lambda x: cudf.Scalar(3) // x,
lambda x: cudf.Scalar(3) / x,
lambda x: cudf.Scalar(3) % x,
lambda x: cudf.Scalar(-1) + x,
lambda x: cudf.Scalar(-2) * x,
lambda x: cudf.Scalar(-2) - x,
lambda x: cudf.Scalar(-2) // x,
lambda x: cudf.Scalar(-2) / x,
lambda x: cudf.Scalar(-3) + x,
lambda x: cudf.Scalar(-3) * x,
lambda x: cudf.Scalar(-3) - x,
lambda x: cudf.Scalar(-3) // x,
lambda x: cudf.Scalar(-3) / x,
lambda x: cudf.Scalar(-3) % x,
lambda x: cudf.Scalar(0) + x,
lambda x: cudf.Scalar(0) * x,
lambda x: cudf.Scalar(0) - x,
lambda x: cudf.Scalar(0) // x,
lambda x: cudf.Scalar(0) / x,
]

pytest_xfail = pytest.mark.xfail
pytestmark = pytest.mark.spilling

# If spilling is enabled globally, we skip many test permutations
# to reduce running time.
if get_global_manager() is not None:
_binops = _binops[:1]
_binops_compare = _binops_compare[:1]
_int_types = _int_types[-1:]
_cmpops = _cmpops[:1]
_reflected_ops = _reflected_ops[:1]
_operators_arithmetic = _operators_arithmetic[:1]
_operators_comparison = _operators_comparison[:1]
_cudf_scalar_reflected_ops = _cudf_scalar_reflected_ops[:1]
DATETIME_TYPES = {"datetime64[ms]"} # noqa: F811
NUMERIC_TYPES = {"float32"} # noqa: F811
FLOAT_TYPES = {"float64"} # noqa: F811
INTEGER_TYPES = {"int16"} # noqa: F811
TIMEDELTA_TYPES = {"timedelta64[s]"} # noqa: F811
# To save time, we skip tests marked "pytest.mark.xfail"
pytest_xfail = pytest.mark.skipif


@pytest.mark.parametrize("obj_class", ["Series", "Index"])
@pytest.mark.parametrize("binop", _binops)
Expand Down Expand Up @@ -114,20 +237,6 @@ def test_series_binop_scalar(nelem, binop, obj_class, use_cudf_scalar):
np.testing.assert_almost_equal(result.to_numpy(), binop(arr, rhs))


_bitwise_binops = [operator.and_, operator.or_, operator.xor]


_int_types = [
"int8",
"int16",
"int32",
"int64",
"uint8",
"uint16",
"uint32",
]


@pytest.mark.parametrize("obj_class", ["Series", "Index"])
@pytest.mark.parametrize("binop", _bitwise_binops)
@pytest.mark.parametrize(
Expand All @@ -152,16 +261,6 @@ def test_series_bitwise_binop(binop, obj_class, lhs_dtype, rhs_dtype):
np.testing.assert_almost_equal(result.to_numpy(), binop(arr1, arr2))


_cmpops = [
operator.lt,
operator.gt,
operator.le,
operator.ge,
operator.eq,
operator.ne,
]


@pytest.mark.parametrize("obj_class", ["Series", "Index"])
@pytest.mark.parametrize("cmpop", _cmpops)
@pytest.mark.parametrize(
Expand Down Expand Up @@ -385,37 +484,6 @@ def test_series_cmpop_mixed_dtype(cmpop, lhs_dtype, rhs_dtype, obj_class):
np.testing.assert_array_equal(result.to_numpy(), cmpop(lhs, rhs))


_reflected_ops = [
lambda x: 1 + x,
lambda x: 2 * x,
lambda x: 2 - x,
lambda x: 2 // x,
lambda x: 2 / x,
lambda x: 3 + x,
lambda x: 3 * x,
lambda x: 3 - x,
lambda x: 3 // x,
lambda x: 3 / x,
lambda x: 3 % x,
lambda x: -1 + x,
lambda x: -2 * x,
lambda x: -2 - x,
lambda x: -2 // x,
lambda x: -2 / x,
lambda x: -3 + x,
lambda x: -3 * x,
lambda x: -3 - x,
lambda x: -3 // x,
lambda x: -3 / x,
lambda x: -3 % x,
lambda x: 0 + x,
lambda x: 0 * x,
lambda x: 0 - x,
lambda x: 0 // x,
lambda x: 0 / x,
]


@pytest.mark.parametrize("obj_class", ["Series", "Index"])
@pytest.mark.parametrize(
"func, dtype", list(product(_reflected_ops, utils.NUMERIC_TYPES))
Expand Down Expand Up @@ -458,37 +526,6 @@ def test_cudf_scalar_reflected_ops_scalar(func, dtype):
assert np.isclose(expected, actual)


_cudf_scalar_reflected_ops = [
lambda x: cudf.Scalar(1) + x,
lambda x: cudf.Scalar(2) * x,
lambda x: cudf.Scalar(2) - x,
lambda x: cudf.Scalar(2) // x,
lambda x: cudf.Scalar(2) / x,
lambda x: cudf.Scalar(3) + x,
lambda x: cudf.Scalar(3) * x,
lambda x: cudf.Scalar(3) - x,
lambda x: cudf.Scalar(3) // x,
lambda x: cudf.Scalar(3) / x,
lambda x: cudf.Scalar(3) % x,
lambda x: cudf.Scalar(-1) + x,
lambda x: cudf.Scalar(-2) * x,
lambda x: cudf.Scalar(-2) - x,
lambda x: cudf.Scalar(-2) // x,
lambda x: cudf.Scalar(-2) / x,
lambda x: cudf.Scalar(-3) + x,
lambda x: cudf.Scalar(-3) * x,
lambda x: cudf.Scalar(-3) - x,
lambda x: cudf.Scalar(-3) // x,
lambda x: cudf.Scalar(-3) / x,
lambda x: cudf.Scalar(-3) % x,
lambda x: cudf.Scalar(0) + x,
lambda x: cudf.Scalar(0) * x,
lambda x: cudf.Scalar(0) - x,
lambda x: cudf.Scalar(0) // x,
lambda x: cudf.Scalar(0) / x,
]


@pytest.mark.parametrize("obj_class", ["Series", "Index"])
@pytest.mark.parametrize(
"funcs, dtype",
Expand Down Expand Up @@ -652,28 +689,6 @@ def test_boolean_scalar_binop(op):
utils.assert_eq(op(psr, False), op(gsr, cudf.Scalar(False)))


_operators_arithmetic = [
"add",
"radd",
"sub",
"rsub",
"mul",
"rmul",
"mod",
"rmod",
"pow",
"rpow",
"div",
"divide",
"floordiv",
"rfloordiv",
"truediv",
"rtruediv",
]

_operators_comparison = ["eq", "ne", "lt", "le", "gt", "ge"]


@pytest.mark.parametrize("func", _operators_arithmetic)
@pytest.mark.parametrize("has_nulls", [True, False])
@pytest.mark.parametrize("fill_value", [None, 27])
Expand Down Expand Up @@ -887,7 +902,7 @@ def test_binop_bool_uint(func, rhs):
(
pytest.param(
np.bool_,
marks=pytest.mark.xfail(
marks=pytest_xfail(
reason=(
"Pandas handling of division by zero-bool is too strange"
)
Expand Down Expand Up @@ -918,7 +933,7 @@ def test_floordiv_zero_float64(series_dtype, divisor_dtype, scalar_divisor):
(
pytest.param(
np.bool_,
marks=pytest.mark.xfail(
marks=pytest_xfail(
reason=(
"Pandas handling of division by zero-bool is too strange"
)
Expand Down Expand Up @@ -1624,7 +1639,7 @@ def test_scalar_null_binops(op, dtype_l, dtype_r):
"microseconds",
pytest.param(
"nanoseconds",
marks=pytest.mark.xfail(
marks=pytest_xfail(
condition=not PANDAS_GE_150,
reason="https://github.com/pandas-dev/pandas/issues/36589",
),
Expand Down Expand Up @@ -1676,19 +1691,19 @@ def test_datetime_dateoffset_binaryop(
{"months": 2, "years": 5, "seconds": 923, "microseconds": 481},
pytest.param(
{"milliseconds": 4},
marks=pytest.mark.xfail(
marks=pytest_xfail(
reason="Pandas gets the wrong answer for milliseconds"
),
),
pytest.param(
{"milliseconds": 4, "years": 2},
marks=pytest.mark.xfail(
marks=pytest_xfail(
reason="Pandas construction fails with these keywords"
),
),
pytest.param(
{"nanoseconds": 12},
marks=pytest.mark.xfail(
marks=pytest_xfail(
reason="Pandas gets the wrong answer for nanoseconds"
),
),
Expand Down Expand Up @@ -1732,7 +1747,7 @@ def test_datetime_dateoffset_binaryop_multiple(date_col, kwargs, op):
"microseconds",
pytest.param(
"nanoseconds",
marks=pytest.mark.xfail(
marks=pytest_xfail(
condition=not PANDAS_GE_150,
reason="https://github.com/pandas-dev/pandas/issues/36589",
),
Expand Down Expand Up @@ -2754,7 +2769,7 @@ def test_binops_decimal_comp_mixed_integer(args, integer_dtype, reflected):
),
],
)
@pytest.mark.xfail(
@pytest_xfail(
reason="binop operations not supported for different "
"bit-width decimal types"
)
Expand Down Expand Up @@ -2928,7 +2943,7 @@ def decimal_series(input, dtype):
],
)
@pytest.mark.parametrize("reflected", [True, False])
@pytest.mark.xfail(
@pytest_xfail(
reason="binop operations not supported for different bit-width "
"decimal types"
)
Expand Down Expand Up @@ -3117,7 +3132,7 @@ def test_empty_column(binop, data, scalar):
cudf.DataFrame([[1, 2, 3, 4], [5, 6, 7, 8]]),
pytest.param(
cudf.DataFrame([[1, None, None, 4], [5, 6, 7, None]]),
marks=pytest.mark.xfail(
marks=pytest_xfail(
reason="Cannot access Frame.values if frame contains nulls"
),
),
Expand Down
Loading

0 comments on commit a2f69e4

Please sign in to comment.