[python][tests] Migrates test_basic.py to use pytest

[thomasjpfan]

Related to #3732

This PR migrates the unitest-based test to use pytest. Additionally, this PR uses the tmp_path fixture when the test needs to create a temporary file.

[ghost]

All CLA requirements met.

[StrikerRUS]

@thomasjpfan Thanks a lot for this PR! I'll check it a little bit later today.

@jameslamb A lot of diff tools I tried sucked to produce a nice diff. But this one looks quite readable. At least it is much better than one which GitHub offers to review. Feel free to use it to save (a lot of, I believe) time 🙂

https://www.textcompare.org/index.html?id=6000f45b143e0d0017c81ac3

[jameslamb]

@thomasjpfan Thanks a lot for this PR! I'll check it a little bit later today.

@jameslamb A lot of diff tools I tried sucked to produce a nice diff. But this one looks quite readable. At least it is much better than one which GitHub offers to review. Feel free to use it to save (a lot of, I believe) time

https://www.textcompare.org/index.html?id=6000f45b143e0d0017c81ac3

thanks so much @StrikerRUS ! I'm glad you said that, I also found GitHub's diff for this type of PR really hard to read.

[jameslamb]

Thanks so much for the help @thomasjpfan ! Awesome PR.

I reviewed this with the diff tool @StrikerRUS mentioned, and it looks great to me. I also compared the results to those on the recent jobs from master, and confirm all the tests ran.

this PR

https://github.com/microsoft/LightGBM/pull/3764/checks?check_run_id=1702945130

====== 207 passed, 2 skipped, 2 xfailed, 72 warnings in 384.25s (0:06:24) ======

most recent tests on master

https://github.com/microsoft/LightGBM/runs/1697195023

====== 207 passed, 2 skipped, 2 xfailed, 72 warnings in 120.19s (0:02:00) ======

Do you have any idea what the tests now take 3 times as long to run? Is there some significant overhead introduced by the tmp_path fixture? I'd be surprised to learn that, but I'm not sure what else the cause could be.

comparing total runtime for this PR to most recent master, for the Python-package tests on GitHub Actions:

	master	this PR
regular	7m22s	31m24s
sdist	6m35s	9m20s
bdist	6m26s	7m16s
if-else	3m28s	6m40s
mpi source	10m45s	12m7s
mpi pip	11m34s	9m40s
mpi wheel	8m8s	9m44s

I understand that these differences could be random and not related to the changes in this PR, but I'd like to know if you have any theories.

[StrikerRUS]

@jameslamb
Speaking about the time difference you've mentioned between master and this PR,

if-else | 3m28s | 6m40s

we don't run test_basic.py for if-else job...

LightGBM/.ci/test.sh

Lines 62 to 69 in a15a370

	if [[ $TASK == "if-else" ]]; then
	conda install -q -y -n $CONDA_ENV numpy
	mkdir $BUILD_DIRECTORY/build && cd $BUILD_DIRECTORY/build && cmake .. && make lightgbm -j4 || exit -1
	cd $BUILD_DIRECTORY/tests/cpp_test && ../../lightgbm config=train.conf convert_model_language=cpp convert_model=../../src/boosting/gbdt_prediction.cpp && ../../lightgbm config=predict.conf output_result=origin.pred || exit -1
	cd $BUILD_DIRECTORY/build && make lightgbm -j4 || exit -1
	cd $BUILD_DIRECTORY/tests/cpp_test && ../../lightgbm config=predict.conf output_result=ifelse.pred && python test.py || exit -1
	exit 0
	fi

[thomasjpfan]

I reran the tests and the timings are more reasonable. My first guess would be that the environment took longer for conda to solve.

[StrikerRUS]

@thomasjpfan Just great! Many thanks for contribution!

I have only one non-blocking the merge question below.

[StrikerRUS]

pytest simply doesn't need this line, right?

[jameslamb]

I reran the tests and the timings are more reasonable. My first guess would be that the environment took longer for conda to solve.

Ok thanks, sounds good to me!

[jameslamb]

thanks again for the help @thomasjpfan ! If you have time and interest, we'd welcome PRs for the other files that still use unittest classes and assertions (#3732 (comment))

[github-actions]

This pull request has been automatically locked since there has not been any recent activity since it was closed. To start a new related discussion, open a new issue at https://github.com/microsoft/LightGBM/issues including a reference to this.