You should mention that you posted this in #ring https://clojurians.slack.com/archives/C0A5GSC6T/p1593506540098400 Follow ups should go there

@U06BE1L6T ok sorry. i did modify the original, hoping this is ok

(defn calc-expiry-date [expiry-days]
  (->> (.plusDays (LocalDate/now) expiry-days)
       (.format (DateTimeFormatter/ofPattern "yyyy-MM-dd"))))

I found it the simplest solution for me

the example is not exactly what you need but it shows how it works

I don’t use shema so I can’t help more in this context

I use spec

https://stackoverflow.com/questions/22463062/how-to-parse-format-dates-with-localdatetime-java-8#:~:text=To%20create%20a%20LocalDateTime%20object,specify%20the%20date%2Ftime%20pattern.

Yeah, that's not the case though. My issue is: • Request coming to Compojure endpoint • Exception is thrown, because provided value is not an instance of LocalDateTime This looks like that somewhere during middleware stuff it doesn't parse the string to LDT correctly. And I don't know where to add it.

Just do it with your code

again I don’t use Compojure so I don’t know if there is some trick for this

I don’t think ring middleware parse dates in reqquest

but I can be wrong

Well, there must be a point where JSON data is being decoded.

And I can't find the place where I would inject the support for LocalDateTime object. Because it only has problem with types like this (they don't support it, I have to do it manually for some reason).

Also, it doesn't matter if it is Ring or Compojure, because Compojure accepts Ring wrappers too. And even custom wrappers. But I don't think providing custom decoder for whole body is performant. It's like a matter of missing protocol, I don't know.

I just found out that all I have to do is use Jsonista inside my Ring middleware.

But I can't figure out how to insert it there, ah.

The docs are so bad...

They are promoting the cooperation between those 2 libraries, but there is nothing about their common configuration.

I use this (wrap-json-body {:keywords? true :bigdecimals? true}) for JSON

I thin if you will describe your issue with past of exception, show exact moment of this exception, some etc. it will be easier to help

On the end I am not sure what is your problem

words are too general 😉

I created a Pastebin of the exception

https://pastebin.com/P7umc8TG

I don’t know schema but is just looks like you have to coerce date to java.time.LocalDateTime before validation. I think you have to id yourself by your own code.

unless schema has some coerce features, but I don’t know this

so you have to parse this string into LocalDateTime

I don’t think any middleware or anything will guess this string has to be convert into LocalDateTime

Hmmmm. That sounds reasonable. Thank you a lot. I was getting confused. I already spent a lot of time on that. I am going to try that and keep you posted.

Looks like it wasn't a mistake of schema 🙂

Cause schema can't do anything about it. So, like I said, it was screwed somewhere in the middleware chain.

(defmethod ring-swagger/custom-matcher LocalDateTime [_] #(LocalDateTime/parse %))

This fixed it 🙂

Thanks! Can I somehow use resolve and not have to use ns-resolve? I'm using clojurescript and ns-resolve doesn't seem to be available here

In CLJS this doesn't work. You can maybe use analyzer/resolve-var in a macro

Ask in #clojurescript

Thank you

(reduce-kv (fn [m k v] (assoc m k (:uri v))) {} table)

Thanks

Sean’s version will work just as well. Destructuring the uri key being the only difference

These are always fun.

(zipmap (keys table)
        (map :uri (vals table)))

They tell you to create beans like https://github.com/rohanKanojia/podsetoperatorinjava/blob/master/src/main/java/io/fabric8/podset/operator/crd/PodSetSpec.java and then pass those classes into function calls

the implements keyword means that it's only using an interface and doing no concrete inheritance

that means you can use defrecord, deftype, or reify instead of gen-class (and all of those are much simpler to use)

if you only ever need one instance, and don't need the class to have a statically known name, reify is simplest, then use deftype if you don't need map behavior, or defrecord if you do

I may have picked a bad example class to link as the others all seem to extend CustomResource or some variant, though I am not certain that they have to so I will test that out, thanks! I think I have to pass the class around for things to reference, and unfortunately I think I'll have to set up the getters and setters for jackson databind, I'll see where that leaves me 🙂

also, if you need concrete inheritance but a generated class-name is OK, proxy is easier to use compared to gen-class

that's a good point, hadn't looked at proxy yet. Also just seen I can annotate fields in deftype (at least) with :volatile-mutable, not sure I like the idea but I might see if jackson will fall back to setting the field directly if a setter doesn't exist

you can crate a setter via a protocol (defprotocol ImyFoo (setBar [this value])) (deftype Foo [^:volatile-mutable bar] ImyFoo (setBar [this v] (set! bar v))) something like that

NB

Note well that mutable fields are extremely difficult to use
  correctly, and are present only to facilitate the building of higher
  level constructs, such as Clojure's reference types, in Clojure
  itself. They are for experts only - if the semantics and
  implications of :volatile-mutable or :unsynchronized-mutable are not
  immediately apparent to you, you should not be using them.

you can put an atom in a field, and swap inside the setter for thread safety

you'd wan't a getter that derefs of course, it would be part of the same protocol defining the setter

Yeah that's reasonable, I'm hopeful that these are very much at the edge and really only jackson will use them and then I'll immediately convert to a map and throw them away so I'm not too worried about the thread safety, and I started in java so know mostly what I'm dealing with, but since performance doesn't matter here anyway the atom idea is a good one

I'm kind of hoping with the mutable fields I can avoid some of the extra protocols that I really don't want to be defining in the first place, but that's probably too much laziness!

it's easy to write correct code that uses an atom and convert to a volatile once you know you need it

the reverse is harder

well, you can't have getter/setters as a bean expects with fields, mutable or no

I mean yeah you could name the field "getFoo" and then you have half as many things to implement, but that seems a little hacky

Oh I agree, I'm just wondering if the bean is a hard requirement for jackson to work. I remember using JAXB or something an age ago and it would use a setter if it was there, otherwise it would try to access the field directly, otherwise it would use the constructor. There was a whole host of fallbacks before it failed

But thanks for the help, I think I have a few things to go and tinker with from this and if I mess up mutable state it will only be my own fault now! 😉

quick repl experiment

(ins)user=> (defrecord whatever [a b])
user.whatever
(ins)user=> (bean (->whatever 1 2))
{:class user.whatever, :empty false}
(ins)user=> (defrecord whatever [getA getB])
user.whatever
(cmd)user=> (bean (->whatever 1 2))
{:class user.whatever, :empty false}
(ins)user=> (defprotocol Iwhatever (getA [this]) (setA [this val]) (getB [this]) (setB [this val]))
Iwhatever
(ins)user=> (defrecord whatever [^:volatile-mutable a ^:volatile-mutable b] Iwhatever (getA [this] a) (setA [this v] (set! a v)) (getB [this] b) (setB [this v] (set! b v)))
Syntax error macroexpanding defrecord at (REPL:1:1).
:volatile-mutable or :unsynchronized-mutable not supported for record fields
(ins)user=> (deftype whatever [^:volatile-mutable a ^:volatile-mutable b] Iwhatever (getA [this] a) (setA [this v] (set! a v)) (getB [this] b) (setB [this v] (set! b v)))
user.whatever
(cmd)user=> (bean (->whatever 1 2))
{:a 1, :b 2, :class user.whatever}

"bean" is just a function that tries to give you a hash map version of a bean

yeah good point, if I can't leverage some nice conversion out from the type I create it definitely weighs in favour of not being lazy and setting up my protocols!

this https://github.com/wjoel/clj-bean and https://github.com/clojure/java.data also looked kind of useful from my travels, first one requires aot unfortuantely (I think)

if you were feeling cheeky you could make a pojo macro that defines a randomly named interface and a deftype with mutable fields all in one go

since the naming is all trivial and machine generatable

user=> (gensym "Ipojo")                                                                                                                                                                               
Ipojo1278  

you could even through the class name in there for clarity

👍 I had started looking at that actually, I don't have much experience with macros but probably enough to get started, it would be really nice, I have to pass in a 'kind' which defined the kubernetes kind to listen for so I can use that as the class name

this is definitely the easiest kind of macro - just expanding from a template in a loop for each field, rather than weird bindings and conditionals

weird bindings and conditions is the bit I know though 😆

Well it will definitely need some work to make pretty 😆 but it's functioning! Thanks for the help today 😊

ahh just realised I need an empty constructor 😢 alas, might have to fall back on gen-class or proxy, one for tomorrow anyway!

This sounds like something we've gone through in an ongoing project.

Is your session middleware set up?

I am not sure what you mean exactly by setup. It's part of the site-defaults: https://github.com/ring-clojure/ring-defaults/blob/master/src/ring/middleware/defaults.clj#L101 So it's definitely called, is there something else to setup?

I'm trying to rattle my own brain to try and remember what exactly our issue was and how we solved it 😝. IIRC it had something to do with the session.

Yea, it must have, somehow, well, thanks for trying 🙂

What does your config to wrap-defaults look like?

Looks like we ended up disabling [:security :anti-forgery] in wrap-defaults config and manually calling wrap-anti-forgery when necessary 😟

Is the ring-session cookie set by the server successfully?

Do you send the post request to the server with credentials enabled?

https://developer.mozilla.org/en-US/docs/Web/API/XMLHttpRequest/withCredentials

Look in the dev console at the actual headers sent in the post request, ring-session should be there.

Is that a cross-origin post request?

So, I figured what the problem actually is. I have the following code:

(routes
      (wrap-api (partial api-routes db) db dev?)
      (wrap-base home-routes dev?)
      (wrap-base (partial user-routes config db) dev?)))

(wrap-base
      (routes (home-routes) (user-routes config db))
      dev?)

what seems to work is to wrap it all in routes again like this:

(routes
      (wrap-api (api-routes db) db dev?)
      (wrap-base
        (routes (home-routes) (user-routes config db))
        dev?))

Hmmm :thinking_face:. I'll have a look, thanks.

You could always separate out the generation of the coords from the actual updating of the board. Maybe something like this:

(def board (atom nil))

(defn random-coord []
  (inc (rand-int 9)))

(defn generate-mole-coords [b]
  (->> (repeatedly #(vector (random-coord) (random-coord)))
       distinct
       (remove #(get-in b %))
       (take 10)))

(defn place-moles [b moles]
  (reduce #(assoc-in %1 %2 :mole) b moles))

(defn place-moles-on-board! []
  (swap! board place-moles (generate-mole-coords @board)))

(comment

  (place-moles-on-board!)
  @board
  )

That way you can always keep the domain code that generates the new state separate from the stateful code that updates the atom

Bear in mind though that if you want the update to the atom to guarantee to add 10 unique moles, you'll need to move more code inside the function that swap! calls, as both these solutions have a read-then-modify bug which will break what I think you want to do if you run it from multiple threads (for example)

so the read bug basically everything would have to happen in swap to make sure so thats why you made functions for everything

It's only really a problem if you have 2 threads updating the board at once. You could fix it by moving the check to see if each coord is already on the board into the function called by swap!

but if you're just using the board as a global variable and not sharing it between threads, then it's not really a problem

cool. and whats the relevance of the ! at the end of the place-moles-on-board! function. I notice theres like some unspoken conventions on method naming based on what it does.

Its obviously a pretty simple game im making not looking to spread this across hundreds of threads or something ha

the ! tends to be used to signify that running this function performs some side effects

if you're playing with a game to learn idiomatic clojure, it's often best to try to learn to write small pure functions and reduce the amount of side effecting code as much as you can

yea im fairly new to clojure which is why I posted. I felt like I was just falling into writing declaritve stuff and it didn't "look" clojure. Ive been trying to just get a simple event loop just to play with structure https://github.com/ThrowsException/whack-a-mole

I think you mean "imperative" when you say "declarative" ... Declarative code tends to be stateless, for example, using a map as a look up table

yes.