I've been playing with @borkdude's carve, and I've started using it as a basis for building code rewriting tools. Mostly one-off things. But the code (and the clever loop that borkdude came up with) is great for rewriting old functions to new ones in an automatic way.

I'm looking for a graph database that's easy to set up and has a mature Clojure client library, any suggestions?

I don't need web scale, but it has to be reasonably performant.

I have a code in namespace X, that code has caching via memoize. So in I have X: (def get-val (memoize load-val)) (def some-map (get-val)) (def some-key (:key some-map). Now when I import namespace X into namespace Y and try and call (prn X/some-key) - > I am getting: #object[clojure.core$constantly$fn__5394 0x50ecef36 "clojure.core$constantly$fn__5394@50ecef36"] While if I call in the same namespace, its just a plain value.

try calling it (X/some-key) - that will return the same thing no matter how many args it is given

sounds like your namespaces are out of sync

yeah I know

but its inconsistent

what you provide there isn't enough to know what's wrong, as it's not how clojure behaves, something else is up on your local repl state

from the same namespace i get value, from someother namespace I have function.

ok

any tips for the witch-hunt 🙂?

I'd start with a full shutdown of the repl and clean up of any cache

(eg. class files you might have generated under target/)

it happens in unittest, so I am not particularly placed in repl env

you(or some tool you are using) is deleting clojure namespaces, and defining them without redefining the namespaces they depend on

scratch that

some tool is just defing the var

somewhere something is doing (alter-var-root #'get-val (constantly 'whatever))

the reason you see different values is because you doing things at different times

ok, thanks for hints

and in the time in between other side effecting (var value changing) code is running

given it is in tests it is likely some heavy handed test mocking

and depending, your test mocking may be stepping on itself, a lot of mocking code that mutates vars is not thread safe, and may not be safe to nest

(my wild irresponsible guess would be you are using midje)

Question about writing libraries wrapping java libraries: It makes sense (to me) to map enums to keywords and use utility functions to transform POJOs to maps. I don't want my users to .getThis and .getThat every time they want to work with Just Data. Would you rather the keywords be namespace qualified? i.e. :state vs. :domain.object/state. The first option looks like it might be easier to work with, at the cost of information and context loss. What do you think?

I think it's best not to wrap things unless the API is designed to waste your time or be error prone

I guess there's also the use case of portability but that usually doesn't work as well as planned

You mean not to wrap Java libs/clients at all?

I mean not to wrap them unless there's an urgency, in the long run it's usually causes more problems than it solves (wrappers are incomplete, the author didn't really understand the API, ...)

but that's very generic advice, and I don't mean to be combative

there might be a valid reason to make your wrapper, I just want people to consider the question carefully

You aren't coming off as combative, it's okay. Giving me a lot to consider, is all

I'd usually use non-namespaced keys if there is a reason to wrap, but in the big picture there's less reason to wrap a POJO, and more reason to wrap things that require inheritance (clumsy in clojure) or factories (very easy to simplify in clojure)

Alright, thank you 🙂

@ben.sless Also, when dealing with POJOs, look at clojure.core/bean for simple cases and org.clojure/java.data library for more complex stuff.

The latter handles the builder syntax as well as the bean syntax, and knows how to handle constructors and various other things.

Will do, thanks 🙃

Should I refer to next.jdbc.datafy for examples?

@ben.sless https://github.com/clojure/java.data

LMK if you have specific questions about it.

What am I doing wrong with buddy here?

(require '[buddy.hashers]
         '[buddy.auth.backends :as backends]
         '[buddy.auth.middleware :refer [wrap-authentication]])

(defn authfn [request {:keys [username password]}]
  (when-let [password-hash (get {"admin" (buddy.hashers/derive "admin")} username)]
    (buddy.hashers/check password password-hash)))

(authfn {} {:username "admin" :password "admin"}) ; => true
(authfn {} {:username "admin" :password "wrong"}) ; => false

(defn handler [request]
  {:status 200 :body "ok" :req request})

(let [backend (buddy.auth.backends/http-basic {:authfn authfn} )
      app (-> handler
              (buddy.auth.middleware/wrap-authorization backend)
              (buddy.auth.middleware/wrap-authentication backend))
      request {:headers {"Authorization" "incorrect password"}}]
  (app request)) ; should return a 403, but returns the 200 response

Got it, the handler should throw an exception if not authenticated:

(defn handler [request]
  (if (buddy.auth/authenticated? request)
    {:status 200 :body "ok" :req request}
    (buddy.auth/throw-unauthorized)))

I expected the middleware to do it all for me

Maybe i'll write an auth library with an honest name some day

Not 'buddy' or 'friend', but 'nemesis'