Define Child/Parent relationship to Schemas

[ikitommi]

If schemas knew their parent, we would be able to do:

• use properties of mother in children, e.g. error definitions (Feature request or bug: adding properties directly to map keys. #86)

;; use error message of a parent (here: map-entry schema)
[:map [:x {:error/message "not an int"} int?]]

• use mother for complex transformations feat: call map-key w/ parent-schema information #119

Is this good or bad?

[rschmukler]

Just echoing some conversation seen in other issues.

For complex transformations, having interceptors (#122, #114) takes care of most things. Any transformer that requires parent access can simply become an :after transformer on the parent type. I will admit that this, theoretically, doesn't compose as well, nor remain as focused - however in my use cases this ended up not being a big difference.

[flow-danny]

I think I have a use-case for this.

In reitit nesting of routes allows one to define what sub-routes share access roles. I would like to define in a similar way the access level for a group of attributes. Then I could filter/transform data (anonymize, or omit keys) based on the user access level.

[ikitommi]

#212 is merged in master. Is allows all but error messages to be read from entry schemas.

(require '[malli.json-schema :as json-schema])

(json-schema/transform
  [:map
   [:x {:json-schema/example 42} int?]
   [:y {:json-schema {:type "string"}} [:fn (fn [x] (string? x))]]])
;{:type "object"
; :properties {:x {:type "integer"
;                  :format "int64"
;                  :example 42}
;              :y {:type "string"}}
; :required [:x :y]}

because of that, we don't need childs to know about their parents.