Open devin-petersohn opened 2 years ago
just forked: not obvious how to invoke the tests. pytest modin
produces a bunch of errors
just forked: not obvious how to invoke the tests. pytest modin produces a bunch of errors
@jbrockmendel thank you for your feedback. pytest modin
also gives me many errors:
We do have instructions saying to use pytest modin/pandas/test
here in the contributing guide. Do you think that instruction could go somewhere more obvious?
By the way, the entire test suite takes way too long on my laptop (and doesn't even cover all the operating systems and engines that the CI). When I make a change, I run the test files or even just the cases that I think are relevant, then open a PR and let Modin's Github CI Actions run all the test. Should the notes in this paragraph also be documented?
Do you think that instruction could go somewhere more obvious?
That's a good location, I just forgot to check it. Thanks.
I get 6 failures locally (on Mac). Will take a look and see if they already have issues.
FAILED modin/pandas/test/test_io.py::TestParquet::test_read_parquet_s3[object] - ModuleNotFoundError: No module named 's3fs'
FAILED modin/pandas/test/dataframe/test_udf.py::test_agg_apply[agg-str-1] - AssertionError: Series are different
FAILED modin/pandas/test/dataframe/test_udf.py::test_agg_apply[apply-str-1] - AssertionError: Series are different
FAILED modin/pandas/test/dataframe/test_udf.py::test_agg_apply_axis_names[agg-str-columns] - AssertionError: Series are different
FAILED modin/pandas/test/dataframe/test_udf.py::test_agg_apply_axis_names[apply-str-columns] - AssertionError: Series are diff...
FAILED modin/pandas/test/internals/test_benchmark_mode.py::test_syncronous_mode - AssertionError: assert False
@jbrockmendel,
modin/pandas/test/internals/test_benchmark_mode.py::test_syncronous_mode
you should set MODIN_BENCHMARK_MODE=True
as we do here. https://github.com/modin-project/modin/blob/05933a5f27fb96f5a7ff6025ae2573d033a31b11/.github/workflows/ci.yml#L417We should probably set the config value in the code to avoid such a confusion. What do you think?
I saw the failures of modin/pandas/test/dataframe/test_udf.py::test_agg_apply
as well some time ago when I ran pytest modin/pandas/test
. However, when I ran pytest modin/pandas/test/dataframe/test_udf.py
, the tests passed. We should investigate this behavior.
Not sure why modin/pandas/test/test_io.py::TestParquet::test_read_parquet_s3
is failing on your side. It would be great if you take a closer look at it.
Not sure why modin/pandas/test/test_io.py::TestParquet::test_read_parquet_s3 is failing on your side.
This looks like it's because @jbrockmendel doesn't have s3fs
, which is in the dev requirements. @jbrockmendel , did you run pip install -r requirements-dev.txt
?
We should probably set the config value in the code to avoid such a confusion. What do you think?
This should be useful IMHO.
Issue to generate release notes: https://github.com/modin-project/modin/issues/4747
From #4571:
We should use this thread to discuss improvements to the developer experience.
cc @NickCrews