This PR should make some major changes to how we run tests.
Isolate tests
Each pipeline is run individually within the corresponding test, rather than using run_pipelines(). This has some downsides but a major upside is that we can now isolate and test pipelines individually.
#This will run ONLY the test for the PEW pipeline
devtools::test(filter = "PEW")
Change test file names
To allow this filtering process, test files should be named test-XXX.R (notice the hyphen rather than underscore). Anything after the hyphen can then be passed to the filter argument in devtools::test()
Skip test with wrong OS
Instead of writing if/else statements to check OS, we now skip tests that can't run on Mac using testthat::skip_on_os("mac").
Pipeline warnings are now a problem
Previously when we ran everything together with run_pipelines() in the file setup-xyz.R pipelines could throw warnings (e.g. NA coercion warnings) and these were hidden/not a problem. Now each pipeline is run in the test and it will flag all these warnings. We need to make sure that pipelines do not return superfluous warnings or that warnings can be turned off by default (e.g. in cases where warnings might help flag weird records).
This PR should make some major changes to how we run tests.
Isolate tests
Each pipeline is run individually within the corresponding test, rather than using
run_pipelines(). This has some downsides but a major upside is that we can now isolate and test pipelines individually.Change test file names
To allow this filtering process, test files should be named
test-XXX.R(notice the hyphen rather than underscore). Anything after the hyphen can then be passed to thefilterargument indevtools::test()Skip test with wrong OS
Instead of writing if/else statements to check OS, we now skip tests that can't run on Mac using
testthat::skip_on_os("mac").Pipeline warnings are now a problem
Previously when we ran everything together with
run_pipelines()in the filesetup-xyz.Rpipelines could throw warnings (e.g. NA coercion warnings) and these were hidden/not a problem. Now each pipeline is run in the test and it will flag all these warnings. We need to make sure that pipelines do not return superfluous warnings or that warnings can be turned off by default (e.g. in cases where warnings might help flag weird records).