I found the bug. I was querying entities on 3 levels (simplified example) [:a/id {:a/bs [:b/id {:b/cs [:c/id]}]}] for :a resolver I declared :a/bs as output, pulled the data from datomic, and renamed the :db/id keys but in addition for whatever reason I pulled the rest of the graph but didn’t rename the keys. Pulling and returning only what I was supposed to return in the :a resolver fixed the issue. It’s still unclear to me why it sometimes worked, but doesn’t really matter.

Hi. Is there support in pathom for federation of resolvers between providers please? I’d like to federate or hop between resolvers running on different services - is such a thing supported or does one simply create a facade and forward on each resolution request?

My second question is one of style. I’d like to build a set of resolvers that supports different ontologies, thus abstracting implementation details and providing graph edges and graph exploration using a standardised vocabulary - this would mean clients can walk - for example, organisations across different countries/regions without concern about underlying schema differences. For example, I’m building a graph of health and care organisations in the UK and would like to represent relationships via an ontology namespace such as https://www.w3.org/ns/prov - so that a relationship ends up being https://www.w3.org/ns/prov#wasDerivedFrom - I’m planning to use reverse URL notation in the clojure namespace to represent a particular attribute but is there a better way? For instance, this means clients can fetch the name of a given organization using :org.w3.http://www.2004.02.skos.core/prefLabel I quite like the first-class nature of naming here, but can’t use traditional URL/value tuples http://www.w3.org/2004/02/skos/core#prefLabel because that wouldn’t be a valid clojure namespace.