The b-side for this test currently shows non-alpha-normalized names from the import -- but should not all imported values be fully alpha normalized, since that's the value that would come out of the cache? Seems like the result should not be different when an import is cached or not.
Does it change the correctness of the test?
My preference is to use the form that one would get for an uncached import.
I think one should get the same form if the import is cached or not, no?
Oh, I see what you are saying. Yeah, we could require that an import protected by a semantic integrity check is 伪-normalized so that you get the same result if it is cached or uncached
Would we want to do this for any expression protected by an integrity check, or just imports?
@philandstuff: I'm guessing probably any expression protected by an integrity check, not just imports. My reasoning is that out now that we're decoupling imports from integrity checks, it seems like the 伪-normalization behavior "belongs" to the integrity check rather than the import.
Most helpful comment
@philandstuff: I'm guessing probably any expression protected by an integrity check, not just imports. My reasoning is that out now that we're decoupling imports from integrity checks, it seems like the 伪-normalization behavior "belongs" to the integrity check rather than the import.