Hi Guys. I've had some fun trying out lacinia over the last few days. There's a few "first experience" things I wanted to mention in case it is helpful.

First one was that it was confusing working out how to add interceptors. The use of a threading macro in the example code is concise but gets in the way based on how the default interceptors get included by pedestal.

My code ended up looking like

(start [this]
    (let [compiled-schema (:schema schema-provider)
          options {:graphiql true}
          service-map (-> (lp/service-map compiled-schema options)
                          (assoc ::http/allowed-origins (fn [x] true))
                          http/default-interceptors
                          http/dev-interceptors)
          server (http/create-server service-map)]
      (assoc this :server (http/start server))))

(I should have said first that the pedestal lacinia tutorial is fanastic... keep it going)

Also, it seems odd that field-resolvers don't know the key they are associated with. The default field resolver works around this. Just seems like a funny thing to not know since they are passed a map value and don't have enough context to actually know what key they are meant to be working on.

Side note: once Clojure's new "git" dependencies work their way through the toolchain (including Leiningen and Cursive), these PR releases will likely just be a tag, not built and pushed to Clojars. We'll see.

i got a really weird one for you all

Bring it!

I’m trying to use lacinia in AWS lambda

I’m able to execute queries and whatnot in the REPL just fine

I’m guessing this is an issue with antlr or clj-antlr and some classpath nonsense with the AOT compiling, but I’m a complete java newb and so I’m at a bit of a loss. hoping ya’ll might have more experience

I’ve tried unzipping the generated uberjar and poking around, and i do see the mentioned “antlr.stg”, but it appears empty

That is odd; all of our applications are Uberjars (without AOT). Are you using Leiningen? Can we see the project.clj?

I'm not sure what version of Java is supported on lambda, that could be a factor (unlikely, though).

I am using leiningen

(defproject cmb-fido-personalization-lambda "0.1.0-SNAPSHOT"
  :description "User personalization data service POC"
  :dependencies [[org.clojure/clojure "1.8.0"]
                 [org.clojure/data.json "0.2.6"]
                 [com.walmartlabs/lacinia "0.23.0"]
                 [com.amazonaws/aws-lambda-java-core "1.1.0"]]
  :source-paths ["src"]
  :resource-paths ["resources"]
  :aot :all)

Should be pretty straight forward, but AWS is not something I use (I can't even use my Walmart corporate card on Amazon's web site).

Can you try, as an experiment, to deploy the Uberjar without AOT? AOT will make your Uberjar much larger, but won't make a huge difference in startup time (maybe 10%).

I’m not sure I can do that

it’s not able to find my class. the way that AWS Lambda works, it expects a class that implements com.amazonaws.services.lambda.runtime.RequestStreamHandler. so I’m using gen-class to create it

i’m not sure how gen-class works without AOT, but when I deployed it without AOT it said it can’t find my class now

(btw, it’s Java 8 - that’s all AWS says in their docs)

Ok, I can see that. I'm afraid I'm in the middle of things right now and can't give much more help. I'd start by investigating what's in the Uberjar vs. what's in the Antlr Jars, then dig around the Leiningen docs/examples to see why some files are not being packaged. Also, check, what's in aws-lambda-java-core to see if it for some reason has Antlr repackaged inside it (this kind of crazy stuff happens for no good reason!).

hmm. OK, thanks 🙂