Except that spec is not part of the description, we are already using simplifications here and there, and the JS tests are not unordered either ( they could have been, with assertDeepEquals, but they are not, with assertSimilar ). Because JS does have something of an ordering in objects, at least sometimes ( and sometimes not ).

It's a bit of a wasps' nest. But I tend towards wanting to preserve existing ordering, and the source has an ordering.

( I may not have thought of all this when translating or when having visions of nightmares. )

Note that I don't get notifications for this thread. If the translation must be updated, please comment on the translation suggestion.

That's a nightmare to test, at least in Haskell.

Only because it uses the wrong type.
https://tools.ietf.org/html/rfc8259

An object is an unordered collection...

It should be a Map in the first place.

Please not. That's a nightmare to test, at least in Haskell.

Should tests ignore the order of pairs parsed in Object?

Using only TypeFamilies. No flexible instances. I'm pretty happy with that.

What do you mean?

isoUnMaybe has to be nonconstructive, so I'm really not a fan of it.