This is a very high-hit (performance sensitive) part of the code. I'd prefer this be written as:

 (if (and (= 1 (count acc))
          (not (#{:list :set} (-> context (get p-iri) :container)))
   (first acc) 
   acc)

Doing two map "get" operations then a set comparison every time is not cheap in the context of possibly millions of times for a single query... at least with the above we are only doing that check when necessary.

We're passing both the context value and the compacting function compact here. Was there an issue with adding the new 2-arity function to compact data maps by accepting both a context and the data value that needs compacting?

The json-ld/compact-fn function take a context and optionally a "used-atom" to track which context entries are used. It then transforms the supplied context and closes over the result in order to do efficient compaction and returns a single-arity function that takes an iri.

I considered adding another arity to the returned compact-fn that would accept a context along with the iri, but that forces us to re-transform the context for every compaction, an operation that can be quite costly if the context is large. So I elected to do the more clumsy approach of passing both.

Now that I think about it, though, if we did go with the 2-arity compact-fn it would be called in multiple places with the same context, so if we added an LRU cache of size 1 in front of it that could make the code more elegant and not incur the penalty of re-processing the context for each compaction. Or just use a delay?

The atom isn't used for any of the compacting functions in this pull request; it's only used for commits.

As I mentioned on our last call, I think we should define a new function that takes a context and data needed to be compacted using that context and returns the compacted data for these usages, and we should only pass the context through these call stacks, not any compacting functions.

My understanding is that compact is called multiple times in the same call stack, and doing the work to compact is expensive, which is why we rely on the compact-fn to do the expensive work once.

In the the fluree.db.query.exec.select namespace it looks like each of the different Selectors can call compact to format the results, and my understanding is that several Selectors may be called during one query. If that's the case, then re-creating the compact-fn for each of those calls could get expensive.

But! My understanding of how Selectors work may be wrong - if only one of them is used than we would only have one compact-fn per query.

No, your understanding is correct. I think we can leave the extra contexts in for now possibly to re-think how we do the compaction and expansion later.