Such dependence on the magnitude of the target value may lead to false positives in our case, so it's probably a good idea to write an expect_ variant or any other solution that always uses absolute tolerances, such as setting scale = 1 to end up in all.equal.
Testthat's
expect_equalcan work with a tolerance, but whether this tolerance is absolute or relative actually depends on the version (waldo:::num_equalin 3e,testthat:::compare.numericin 2e), possibly its size compared to the target value, or the specific path taken bytestthat:::compare.numericandall.equal.Such dependence on the magnitude of the target value may lead to false positives in our case, so it's probably a good idea to write an
expect_variant or any other solution that always uses absolute tolerances, such as settingscale = 1to end up inall.equal.