I'm the one who wrote that, but I'd point to @ekmett. The Applicative involved was stolen from lens, where it's used to implement the aptly-named confusing. By applying it in a finer-grained manner here (generic representations rather than whole containers), we maximize the benefit. It's really all about avoiding unnecessarily stacked fmap applications. I can try to explain a bit later today, but it's fine if @sjoerdvisscher gets to it first.

Right, I don't think I understand this enough to document it. I played around enough with the types at some point to get what was going on, but that has long since been garbage collected from my brain.

@treeowl

I'm the one who wrote that

Indeed, I now realise that #316 was coauthored 🙂 (I'd like to bring it up to master, now, but it's a non-trivial amount of work to merge master in it. I don't really have time to do so, would you?).

Regarding this PR: I don't necessarily ask for a detailed explanation (though it would be welcome), but enough to get an intuition of why these types specifically?

I will work on bringing that branch (or its main ideas) up to master as soon as I can. I'm still getting wiped out by this cold that won't let me sleep. I will try to write an explanation of that, however brief, today.

I wrote a little documentation for Curried in #385. It's not great; hopefully someone will be able to flesh it out later.