test: xfail cuDF tests that use total_seconds

[LiamConnors]

What type of PR is this? (check all applicable)

• 💾 Refactor
• ✨ Feature
• 🐛 Bug Fix
• 🔧 Optimization
• 📝 Documentation
• ✅ Test
• 🐳 Other

Related issues

• Related issue # test: Get all cuDF tests passing #862
• Closes #

Checklist

• Code follows style guide (ruff)
• Tests added
• Documented the changes

If you have comments or can explain your changes, please do so below.

There are ten tests here currently failing because it looks like TimedeltaProperties in cuDF doesn't have a total_seconds and all these tests depend on it.

https://github.com/rapidsai/cudf/blob/branch-24.10/docs/cudf/source/user_guide/api_docs/series.rst#timedelta-properties
docs start here: https://docs.rapids.ai/api/cudf/stable/user_guide/api_docs/api/cudf.core.series.timedeltaproperties.days/

FAILED tests/expr_and_series/dt/datetime_duration_test.py::test_duration_attributes[cudf_constructor-total_minutes-expected_a0-expected_b0-expected_c0] - AttributeError: 'TimedeltaProperties' object has no attribute 'total_seconds'
FAILED tests/expr_and_series/dt/datetime_duration_test.py::test_duration_attributes[cudf_constructor-total_seconds-expected_a1-expected_b1-expected_c1] - AttributeError: 'TimedeltaProperties' object has no attribute 'total_seconds'
FAILED tests/expr_and_series/dt/datetime_duration_test.py::test_duration_attributes[cudf_constructor-total_milliseconds-expected_a2-expected_b2-expected_c2] - AttributeError: 'TimedeltaProperties' object has no attribute 'total_seconds'
FAILED tests/expr_and_series/dt/datetime_duration_test.py::test_duration_attributes[cudf_constructor-total_microseconds-expected_a3-expected_b3-expected_c3] - AttributeError: 'TimedeltaProperties' object has no attribute 'total_seconds'
FAILED tests/expr_and_series/dt/datetime_duration_test.py::test_duration_attributes[cudf_constructor-total_nanoseconds-expected_a4-expected_b4-expected_c4] - AttributeError: 'TimedeltaProperties' object has no attribute 'total_seconds'
FAILED tests/expr_and_series/dt/datetime_duration_test.py::test_duration_attributes_series[cudf_constructor-total_minutes-expected_a0-expected_b0-expected_c0] - AttributeError: 'TimedeltaProperties' object has no attribute 'total_seconds'
FAILED tests/expr_and_series/dt/datetime_duration_test.py::test_duration_attributes_series[cudf_constructor-total_seconds-expected_a1-expected_b1-expected_c1] - AttributeError: 'TimedeltaProperties' object has no attribute 'total_seconds'
FAILED tests/expr_and_series/dt/datetime_duration_test.py::test_duration_attributes_series[cudf_constructor-total_milliseconds-expected_a2-expected_b2-expected_c2] - AttributeError: 'TimedeltaProperties' object has no attribute 'total_seconds'
FAILED tests/expr_and_series/dt/datetime_duration_test.py::test_duration_attributes_series[cudf_constructor-total_microseconds-expected_a3-expected_b3-expected_c3] - AttributeError: 'TimedeltaProperties' object has no attribute 'total_seconds'
FAILED tests/expr_and_series/dt/datetime_duration_test.py::test_duration_attributes_series[cudf_constructor-total_nanoseconds-expected_a4-expected_b4-expected_c4] - AttributeError: 'TimedeltaProperties' object has no attribute 'total_seconds'

@MarcoGorelli do you think it would be a good idea to document these and any others that aren't currently supported. I was thinking that it might be helpful, but also that it may be hard to keep this info up to date. Though, I suppose when running the tests, it would be easy enough to see if any of these pass in the future.

[MarcoGorelli]

thanks @LiamConnors ! 🙏

for drop_first, I reported this to them and they reacted very quickly rapidsai/cudf#16795

I think it wouldn't hurt to add a docs note to some methods saying "not available in cudf", but .dt.total_seconds seems quite common...perhaps let's see if they respond to rapidsai/cudf#16802 ?

there's some unrelated CI failures, i'll try to get ci green, then we can merge

[MarcoGorelli]

unrelated ci failures are already being tracked, let's merge now then to reduce the queue - thanks @LiamConnors !

[MarcoGorelli]

@LiamConnors if it interests you, it might be possible to workaround cudf's limitation here by reimplementing total_seconds in terms of the other components?

[LiamConnors]

thanks @MarcoGorelli. that sounds good to me! I'll take a look.