After seeing the #Performance section of malli's readme I'm quite interested in trying it out. One thing I was wondering is how it fares at compile-time? I'm just assuming that it achieves high runtime speed by offloading as much as possible to macros. Are there any numbers how this affects eg lein startup time?

@vale good question. There are no macros in malli (ok, one mandatory). the “schema compile time” is at runtime, you can have a separate step to create an immutable & optimized validator/transformer/parser and reuse that.

in many cases, even without reuse, it’s still faster than with alternatives, e.g. with spec / spec-tools - one-shot “gimme a validator, use it once and forget it”.

malli.core (without sci) should load under 1sec from source and 0.1sec from classes.

(= int? (m/validator [:and int? int? [:and int? [:and int?]]]))
; => true

(= identity
   (m/decoder
     [:map
      [:id :int]
      [:name :string]
      [:tags [:vector :string]]]
     (mt/json-transformer)))
; => true

was hoping that humanize would use language like 'at least one string? is required' here, but instead, i get "unknown error". am i doing it wrong, or is this a TODO for humanize?

(m/explain [:sequential {:min 1} string?] ())

  ;; {:schema [:sequential {:min 1} string?]
  ;;  :value  ()
  ;;  :errors (#Error{:path   []
  ;;                  :in     []
  ;;                  :schema [:sequential {:min 1} string?]
  ;;                  :value  ()
  ;;                  :type   :malli.core/limits})}

  (-> (m/explain [:sequential {:min 1} string?] ())
      me/humanize)

  ;; ["unknown error"]

@robert-stuttaford just missing an error mapping for :malli.core/limits. PR welcome to malli.error.

you can also play with the mappings by passing them into me/humanize:

(-> (m/explain [:sequential {:min 1} string?] ())
    (me/humanize {:errors {::m/limits {:error/message {:en "there's no limits"}}}}))
; => ["there's no limits"]

uncomplete, but close:

(defn min-max-sequence-error-message [error options]
  (str ((me/-pred-min-max-error-fn {:pred identity}) error options) " elements"))

(-> (m/explain [:sequential {:min 1} string?] ())
    (me/humanize {:errors {::m/limits {:error/fn {:en min-max-sequence-error-message}}}}))
; => ["should be at least 1 elements"]

The following scheme systematically produces a stack overflow. It is a recursive definition where one item (`:E`) contains another item (which might be :E as well, or not). I am surprised this stack overflow is that systematic. There must be something wrong somewhere unless I am missing the obvious. Tried with :or instead of :multi, same thing.

(def registry
     {:A    [:tuple [:= :A]]
      :B    [:tuple [:= :B]]
      :C    [:tuple [:= :C]]
      :D    [:tuple [:= :D]]
      :E    [:tuple
             [:= :E]
             :item]
      :item [:multi
             {:dispatch first}
             [:A :A]
             [:B :B]
             [:C :C]
             [:D :D]
             [:E :E]]})


(malli.gen/generate [:schema
                     {:registry registry}
                     :item])

Execution error (StackOverflowError) at malli.core/-schema (core.cljc:242).

Validation results in the same, I probably should open an issue actually

@adam678 would [:ref :item] work? It's a lazy reference type

there is no recursion detection for eager references.

Greetings everyone!! Thank so much for this super neat lib, really cool and useful. I have a bit of an issue, whereby requiring malli.core triggers a ClassNotFoundException:

nREPL server started on port 55188 on host 127.0.0.1 - 
REPL-y 0.4.4, nREPL 0.8.3
Clojure 1.10.3
Java HotSpot(TM) 64-Bit Server VM 11.0.9+7-LTS
    Docs: (doc function-name-here)
          (find-doc "part-of-name-here")
  Source: (source function-name-here)
 Javadoc: (javadoc java-object-or-class-here)
    Exit: Control+D or (exit) or (quit)
 Results: Stored in vars *1, *2, *3, an exception in *e

user=> (require '[malli.core :as m])
Execution error (ClassNotFoundException) at java.net.URLClassLoader/findClass (URLClassLoader.java:471).
malli.impl.util.SchemaError

user=>

wonderful thank you @ikitommi!

A little while ago, there was a conversation about stripping nulls from the return back to the client

hjmmm...let me think a bit more...

So, okay, I'm doing this in the setup of the ring/router

:muuntaja (m/create
                      (assoc-in m/default-options
                                [:formats "application/json" :opts]
                                {:mapper (-> (j/object-mapper)
                                             (.setSerializationInclusion JsonInclude$Include/NON_EMPTY))}))

Following throught the code, I believe I'm handed back an instance of the ObjectMapper from jsonista.core when I then set that option

the thing is, if I do this (now!), then all of my coercion fails (using malli)

i.e.,

"coercion": "malli",
    "errors": [
        {
            "in": [
                "repId"
            ],
            "message": "missing required key",
            "path": [
                "repId"
            ],
....
....

if I put it back to :muuntaja m/instance, then all is well with the world

I'm a bit lost as to why it fails

it feels like it's not able to parse the incoming JSON data from the request

I realise this is in the wrong channel (maybe I should put into #reitit)

pushed out [metosin/malli "0.4.0"]. It’s a new MINOR as the :multi parse is changed (fixed) and leaning on the old (broken) code can break someone. https://cljdoc.org/d/metosin/malli/0.4.0/doc/changelog