Open kevinushey opened 3 weeks ago
Could/should we just make an empty test name an error unconditionally? (or maybe make it a warning and then upgrade it to an error in a couple of years)
We could, but there's a lot of tests in testthat itself which also use empty names, so I wasn't sure if that was intentionally allowed.
Hmmm yeah, mostly used for testing though. I think we could fix those.
Currently, testthat accepts empty strings as test names, but this causes issues with snapshot tests. For example:
But every time this test file is run, testthat doesn't recognize the existing snapshot file, and so always prints:
One user encountered this during the package development workshop at posit::conf, so learners might unintentionally fall into this trap.