Closed ivirshup closed 1 month ago
Attention: Patch coverage is 96.22642%
with 2 lines
in your changes are missing coverage. Please review.
Project coverage is 91.12%. Comparing base (
10db44f
) to head (e06dca6
).
Files | Patch % | Lines |
---|---|---|
api/python/cellxgene_census/tests/conftest.py | 75.00% | 2 Missing :warning: |
:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.
For this kind of change, would you prefer that it's done incrementally or in one big PR?
If all you intend to do is change the scope
of the census
pytest.fixture
from the default of function
to session
, then I think it is reasonable to make that change in a single PR in all the tests. But I would also like @ebezzi opinion here too
Cool. So the two other sets of tests where I think this can be used is test_lts_compat
and test_get_helpers.py
. I would think we should also share the fixture between the tests, so can move it to the conftest.
I propose we re-use SOMA objects in testing to cut down on total test time.
From a single measurement, I see a little more than a 10% (~40 seconds) decrease in test time with this change. I've run the
get_anndata
tests on this commit and the one before it. I haven't run this many times, so I don't know what the variance is. You can see that with this commit there is far lower time taken on setup for each task (since the resource is simply removed)Timings on this commit
Timings on main