What do you mean by wrap them? Would you be calling these methods from java or clojure? Are you talking about generating a new class that has the same methods or trying to monkey patch classes somehow?

Monkey patch. Specifically, I want to extend the implementation of java.sql.Connection to include some extra schema meta-data for consumption by honeyeql and graphqlize.

I'd need to do something like pass in a custom instance of java.sql.DataSource which provides an instance of my monkey-patched java.sql.Connection.

java.sql.Connection is an interface, so you don't need to monkey patch. You can just create a new instance that also implements java.sql.Connection.

I'm not super familiar with honeysql and graphqlize, but there's probably a more idiomatic way to get what you want too.

Right, but I'm going to want to use the existing MySQL implementation, and just monkey-patch the bits I need (which would include calling through to the MySQL instance.

Can you give a little more background on why the existing implementation isn't working for you?

I'm not seeing any other way to do this without forking honeyeql. The schema-related stuff is pretty much hardwired, which does make sense. The issue I'm trying to address is that MySQL views don't provide any schema info about foreign keys, so honeyeql (and hence graphqlize) don't know how to navigate from an entity defined by a MySQL view to a foreign key reference from the original table.

Given an existing instance, it would be pretty easy to create a new instance that adds the bits of functionality you're interested in using proxy or reify. It wouldn't apply to every instance, but it should be workable.

There are some limitations to proxy and reify which I don't think will get in your way, but in some cases, it may make sense to either write a small bit of java or use https://github.com/jgpc42/insn.

There's also a #C66EM8D5H channel that can probably give you more specific advice.

What did you use to generate the Swagger JSON? (Or whatever it is that ends up creating this web page)

metosin/reitit

And how is the endpoint defined?

I tried the reitit-swagger example and it uses body as the name by default.

Yes, I dont want it to use body as the name by default. Here's how my route is defined :

(def routes
  [["/"
    {:swagger {:tags ["Authentication"]}
     :middleware
     [wrap-json-params
      wrap-keyword-params
      wrap-spec
      wrap-json-response]}
    ["authentication"
     {:post
      {:spec ::authentication/authentication#post.body
       :handler handle-authentication
       :summary "Authenticate user"
       :name "Authenticate user"
       :body "Authenticate user"
       :description "TODO abstract description"
       :parameters {:body ::authentication/authentication#post.body}}}]]])

are you talking about ::authentication/authentication#post.body or that you pass this to the :body ? If it's the former you should be able to define it as you wish? If it's the latter, that's part of the http standard. You put the data you want to send into the body of a HTTP message. Sorry if it's obvious, but I'm confused.

Sorry if I'm not clear. I'm talking about the rs.metav.request.api.authentication/authentication#post.body in the screenshot. I want to change it to something else in the swagger. By default, it takes the path of the body param which I dont want.

Ah, I wondered where that # came from. :) Just in case - while it works, note that it's not a valid character for keywords. No idea whether it can lead to any problems, but I wouldn't use it.

I've tried without it, it didnt solve my problem

Yeah, that was beside the main point.

Judging by the source, you can require [spec-tools.core :as st] and then wrap you spec in (st/spec ... {:name "body data"}).

Thanks, will check it out !!!

defis a ”special function”, if that's what you mean?

more commonly called a “special form”

@U0ETXRFEW yes? I think so, haha. I am actually unfamiliar

Oh, I meant a special form. Thanks @U11BV7MTK!

@U03810HB7QS Clojure the language is bootstrapped from some first special forms, 14 I think they are. def fn, are amongst them. Then core functions and macros are build from these, and then more core functions and macros built from these and so on. You might want to try the Getting Started REPL in Calva, which tries to get this across in an interactive way, using the REPL.

Actually, looking at core.clj I see that fn is defined there.

(def
 ^{:macro true
   :added "1.0"}
 fn (fn* fn [&form &env & decl] 
         (.withMeta ^clojure.lang.IObj (cons 'fn* decl) 
                    (.meta ^clojure.lang.IMeta &form))))

yes

Or, rather, a macro using the fs* special form.

reading this closely -> https://clojure.org/reference/special_forms is filling in some blanks, for sure

If its of interest, I did a small random function generator which would show a public function (and its metadata) from clojure.core or any namespace provided https://practical.li/clojure/simple-projects/random-clojure-function/

Because type works at run time. Does ^Object not work?

It does, another case of me following a red herring 😅 thanks @U2FRKM4TW

It's CLJS, but that shouldn't matter.

that is just sort of how s/and works

I would personally use :req-un in the keys spec and if it still fails, write a generator function

req or req-un has nothing to do with it

the sad reality of spec is that most interesting schemas are going to be too hard to generate randomly and youll simply have to write generators for most of your stuff, but once you get the hang of it, it's not bad

it has to do with the internals of test.check, which is what does the generating

the way s/and is implemented is using such-that from test.check

such-that basically acts as a filter on a generator

what you're encountering is that some generator making guesses about the inputs to your arbitrary function, and it might take a very long time if you have a picky predicate

so it gives up rather than heats the universe

so (s/and int? #(< 1900 % 2150)) generates using int? and then throws away stuff that doesn't pass #(< 1900 % 2150)

so the more a/and stuff you combine together the less likely it is that you will be able to generate something successfully

sorry

s/and

hmm... maybe s/int-in will work better for year and month here

Confirmed: s/int-in is the solution here.

Thanks everybody!

And for other similar cases, s/and or not, it should be possible to write a custom generator.

I'm a lazy bastard. If the computer can do it, why should I? @U2FRKM4TW

Because sometimes the computer can't do it.

well, I expect s/int-in and related stuff to have reasonably "good" generators ootb, but again, any interesting spec, like a string representing a valid IP address, will simply need a custom generator

That's fair enough.

sometimes randomly generated data isnt specific enough to conform to your spec and give you useful samples of valid data, that's why custom generators are important

it lets you randomly generate data in a way that hopefully gives you valid samples quickly