[TST] functions/(add_columns|conditional_join).py scripts are not covered by test cases

[Zeroto521]

The codecov/project of latest commit 3687219 on dev branch is 77.60% (-20.04%).

The following methods or scripts aren't covered by test cases.
Please go codecov to see details.

• functions/add_columns.py::add_columns
• functions/conditional_join.py

We could learn that the coverage doesn't like our thought after adding --cov-append option.
pytest tests != pytest tests -m "turtle" + pytest tests -m "not turtle"

[Zeroto521]

But that is weird. Testing logic is the same for whether it is a PR or dev branch now.
I see the PR's codecov/project is 97.66%, not 77.60%.

[Zeroto521]

Or another solution is that directly use pytest tests.
Don't split test cases into "turtle" and "not turtle" two parts.
Don't figure out the reason

[samukweku]

if i remember correctly , and correct me @ericmjl , we used the turtle because some tests were taking quite some time to run, and we needed a way to let them run without timing out?

[Zeroto521]

The aim of #1185 and #1190 these two PRs are to keep the same testing logic no matter whether it's the dev branch or a PR.

The dev branch codecov/project history.

• Before [INF] Merge codecov.yml into tests.yml #1185, the dev branch used pytest tests, and the codecov/project was 97%.
• After [INF] Merge codecov.yml into tests.yml #1185 and before [INF/CI] Add --cov-append for pytest #1190, the dev branch used pytest tests -m "turtle" + pytest tests -m "not turtle" but without --cov-append. So the codecov/project was 47.30%.
• After [INF/CI] Add --cov-append for pytest #1190, the dev branch used pytest tests -m "turtle" + pytest tests -m "not turtle" with --cov-append. the codecov/project was 77.60%.

[ericmjl]

if i remember correctly , and correct me @ericmjl , we used the turtle because some tests were taking quite some time to run, and we needed a way to let them run without timing out?

That's almost right, @samukweku! We wanted the test suite to complete faster to minimize feedback loop timing. Without splitting, we would take ~40 minutes for all tests to pass. Now it's a maximum of 26 minutes, with a possibility of further reducing if we split out testing of the longest-running tests. Of course, the alternative is to reduce the testing time of those long-running tests.

[ericmjl]

@Zeroto521 let's look at the latest test runs. --cov-append generally does the right thing here.

[Zeroto521]

For a single environment is working. Speed up via splitting different parts (parallelizing test cases).
But the aim of #1143 is to set multi-environments and compat with them. So here is to parallelize multi-environments, not test cases. Splitting test cases are not working to speed up. Splitting test cases get more clear for CI.

[Zeroto521]

I suddenly found there has an option addopts in pyproject.toml.
It may affect the code coverage.

pyjanitor/pyproject.toml

Lines 33 to 34 in c5ba7ef

	[tool.pytest.ini_options]
	addopts = "--cov=janitor/ --cov-report=html --cov-report=xml --cov-report=term-missing --durations=0"

[Zeroto521]

I suddenly found there has an option addopts in pyproject.toml. It may affect the code coverage.

pyjanitor/pyproject.toml

Lines 33 to 34 in c5ba7ef

	[tool.pytest.ini_options]
	addopts = "--cov=janitor/ --cov-report=html --cov-report=xml --cov-report=term-missing --durations=0"

This is the reason. The code coverage is back to normal. See #1204

[Zeroto521]

this issue could be closed. It was already fixed in #1204