> Perhaps this is an acceptable solution for your use cases.
>

It isn't, I'm afraid.

By the way, I'm well aware of the problem with effects and value
restriction and such (hence my qualification above, with purity). However,
I don't care if it fails from time to time (and I think I remember some
check about purity being done somewhere in the module system). For my
use-case, I don't care if I need to have access to the original source-code
-- I mean to see the concrete definition of the functor rather than its
signature -- either (though it seems a limitation in general).

Am I weird to run into this issue all the time or is it common?


PS: now that there are both applicative functors and generative functors,
could applicative functors be restricted to being pure, so that this sort
of manipulation become possible in a generic way?