Feature/query iri map

[zonotope]

This patch allows users to reference subjects with an iri map in the object position containing one entry whose key is "@id" and value is the referenced iri.

After parsing the iri map, the query executor first looks up the subject id corresponding to the supplied iri and then processes the rest of the query normally.

[dpetran]

We have a constant for the @id string (const/iri-id) that we may want to use here.

[cap10morgan]

In other places where this construct is used (transaction data mainly) it is often with "id" "@id" in the context and so these can be written as either {"id" ...} or {"@id" ...}. I bet one of the first things that would come up from users of this is "Why the inconsistency?" It's a bit tricky, though. Since transaction data is JSON-LD those get expanded for you. But queries aren't JSON-LD, so you don't get that for free. But I do wonder how hard it would be to support context-aliases for "@id" in here? Perhaps not a this-PR thing either way, just wanted to flag it.

[bplatz]

I think this is good to go, but not fully accurate. Based on the context being used, we don't know what key will translate to @id. If this is run after json-ld expansion, it will always be @id. If run before expansion, it could literally be anything (we often alias id to @id in our example contexts).

One optimization we did in the JSON-LD library is we include :type-key so we don't have to, in some scenarios, fully expand to perform some logic. I think we should also consider including :id-key in the parser. Then this logic could be as an example:

(defn iri-key?
  [x parsed-context]
  (or (= (:id-key parsed-context) x)
        (= "@id" x))

The way it is I think we should merge though, so we can make this work tomorrow, but make another ticket to enhance the json-ld library, and then modify this once that is done.

[zonotope]

When writing this code, I had assumed that this "@id" key would be expanded by the time it got to this part of the code, but there's nothing here right now that does the expanding. I think this code is fine as is (or with minor changes anyone would like to suggest) because this is the only thing that can show up as a map like this in the object position.

I do however like the idea of using this map format for more than just specifying an iri eventually, so we'd have to figure out proper id key expansion before we can do that.

[zonotope]

I've made changes to expand the key of the iri map as well so the equality check against "@id" should always work.

[cap10morgan]

LGTM! 🎉