properties are not preserved by the `:merge` schema

[vise890]

I would expect (:merge malli.util/schemas) to behave like malli.util/merge:

(ns schemas
  (:require [malli.core :as ma]
            [malli.util :as mu]))

;; this is the behavior of malli.util/merge: properties are preserved (expected)
(ma/properties
 (ma/schema (mu/merge [:map {:p1 1}
                       [:x :int]]
                      [:map
                       [:y :int]])
           ))
;; => {:p1 1}

;; this is the behavior of the malli.util :merge schema: properties are lost in the process
(ma/properties
 (ma/schema [:merge
             [:map {:p1 1}
              [:x :int]]
             [:map
              [:y :int]]]
            {:registry
             (merge (ma/default-schemas)
                    (mu/schemas))}))
;; => nil

This seems to stem from the implementation of into-schema of malli.util/-util-schema. Perhaps it could fall back to the merged schema's properties like so:

(defn -util-schema [{:keys [type min max childs type-properties fn]}]
  ^{:type ::m/into-schema}
  (reify m/IntoSchema
    (-type [_] type)
    (-type-properties [_] type-properties)
    (-properties-schema [_ _])
    (-children-schema [_ _])
    (-into-schema [parent properties children options]
      (m/-check-children! type properties children min max)
      (let [[children forms schema] (fn properties (vec children) options)
            properties (or properties (m/properties schema)) ; 👈 HERE
            form (delay (m/-create-form type properties forms options))

(works, but I'm unsure of the implications)

Tested on malli 0.14.0

Thank you!

[ikitommi]

Makes sense, PR welcome.

[vise890]

Thank you for the quick reply @ikitommi !

So i put the proposed fix in and ran all the tests, and one unintended consequence is this:

(m/form
 (m/schema
  [:merge
   [:map {:closed true}
    [:x :int]]
   [:map
    [:y :int]]]
  {:registry (merge (m/default-schemas)
                    (mu/schemas))}))
;; => [:merge {:closed true} [:map {:closed true} [:x :int]] [:map [:y :int]]]
;;            ^^^^^^^^^^^^^^

Perhaps it's best to leave the form alone, an just rely on m/deref to get at the properties of the result after all.. I'm happy to close this unless you have another suggestion.

Thank you again, and keep up the amazing work!

[ikitommi]

don't have other suggestions, let's close this. thanks for your work on this!