Spoke with Bobbi Fox and David at Code4Lib 2017; we are in agreement that for the particular vocabularies for which Harvard is implementing converter code, that the tests for their converter code should be associated with the converter code (i.e. junit tests if java ...).
It makes sense for tests to live as close to the converter code as possible.
The validator should have tests only if
it will be used for direct a comparison across converter implementations (marc -> BF2, specialized vocabs ... anything with multiple converters to be compared)
there is a single converter for which additional validation, beyond the converter's own test code, is necessary.
The validator can only have tests if
there is a spec that can be translated into tests.
Spoke with Bobbi Fox and David at Code4Lib 2017; we are in agreement that for the particular vocabularies for which Harvard is implementing converter code, that the tests for their converter code should be associated with the converter code (i.e. junit tests if java ...).
It makes sense for tests to live as close to the converter code as possible.
The validator should have tests only if
The validator can only have tests if