malli

Data-driven Schemas for Clojure/Script.

STATUS: well matured alpha

Schema definitions as data
Both Vector and Map -syntax
Validation and Value Transformation
First class Error Messages with Spell Checking
Generating values from Schemas
Inferring Schemas from sample values
Tools for Programming with Schemas
Parsing, Unparsing and Sequence Schemas
Persisting schemas, even function schemas
Immutable, Mutable, Dynamic, Lazy and Local Schema Registries
Schema Transformations to JSON Schema and Swagger2
Multi-schemas, Recursive Schemas and Default values
Function Schemas with dynamic and static schema checking
Visualizing Schemas with DOT and PlantUML
Fast

Presentations:

High-Performance Schemas in Clojure/Script with Malli 1/2
ClojureScript Podcast: Malli wtih Tommi Reiman
Structure and Interpretation of Malli Regex Schemas
LNDCLJ 9.12.2020: Designing with Malli, slides here
Malli, Data-Driven Schemas for Clojure/Script
CEST 2.6.2020: Data-driven Rapid Application Development with Malli
ClojureD 2020: Malli: Inside Data-driven Schemas, slides here

Try the online demo. Libraries using or supporting malli:

Aave, a code checking tool for Clojure.
Gungnir, a high level, data driven database library for Clojure data mapping.
Regal, Royally reified regular expressions
Reitit, a fast data-driven router for Clojure/Script.
wasm.cljc - Spec compliant WebAssembly compiler and decompiler
malli-instrument - Instrumentation for malli mimicking the clojure.spec.alpha API
Snoop - Function instrumentation using Malli schemas.
malli-key-relations - Relational schemas about map keys for malli
malli-cli - Command-line processing

Motivation

We are building dynamic multi-tenant systems where data models should be first-class: they should drive the runtime value transformations, forms and processes. We should be able to edit the models at runtime, persist them and load them back from a database and over the wire, for both Clojure and ClojureScript. Think of JSON Schema, but for Clojure/Script.

Hasn't the problem been solved (many times) already?

There is Schema, which is an awesome, proven and collaborative open-source project, and we absolutely love it. We still use it in many of our projects. The sad part: serializing & de-serializing schemas is non-trivial and there is not proper support on branching.

Spec is the de facto data specification library for Clojure. It has many great ideas, but it is opinionated with macros, global registry and it doesn't have any support for runtime transformations. Spec-tools was created to "fix" some of the things, but after five years of developing it, it's still a kind of hack and not fun to maintain.

So, we decided to spin out our own library, which would do all the things we feel is important for dynamic system development. It's based on the best parts of the existing libraries and several project-specific tools we have done over the years.

If you have expectations (of others) that aren't being met, those expectations are your own responsibility. You are responsible for your own needs. If you want things, make them.

Rich Hickey, Open Source is Not About You

The library

Syntax

Malli supports both Vector and Map syntaxes.

Vector Syntax

The default syntax uses vectors, inspired by hiccup:

type
[type & children]
[type properties & children]

Examples:

;; just a type (String)
:string

;; type with properties
[:string {:min 1, :max 10}]

;; type with properties and children
[:tuple {:title "location"} :double :double]

;; a function schema of :int -> :int
[:=> [:cat :int] :int]

Usage:

(require '[malli.core :as m])

(def non-empty-string
  (m/schema [:string {:min 1}]))

(m/schema? non-empty-string)
; => true

(m/validate non-empty-string "")
; => false

(m/validate non-empty-string "kikka")
; => true

(m/form non-empty-string)
; => [:string {:min 1}]

Map Syntax

NOTE: Map Syntax / SchemaAST is considered as alpha and subject to change.

Alternative map-syntax, similar to cljfx:

;; just a type (String)
{:type :string}

;; type with properties
{:type :string
 :properties {:min 1, :max 10}
 
;; type with properties and children
{:type :tuple
 :properties {:title "location"}
 :children [{:type :double}
            {:type :double}]}
 
;; a function schema of :int -> :int
{:type :=>
 :input {:type :cat, :children [{:type :int}]}
 :output :int}

Usage:

(def non-empty-string
  (m/from-ast {:type :string
               :properties {:min 1}}))

(m/schema? non-empty-string)
; => true

(m/validate non-empty-string "")
; => false

(m/validate non-empty-string "kikka")
; => true

(m/ast non-empty-string)
; => {:type :string,
;     :properties {:min 1}}

Map-syntax is also called the Schema AST.

Why Two Syntaxes?

We have found out that the overhead of parsing large amount of vector-syntaxes can be a deal-breaker when running on slow single-threaded environments like Javascript on mobile phones. Instantiating schemas using the Schema AST can be much faster.

Validation

Validating values against a schema:

;; with schema instances
(m/validate (m/schema :int) 1)
; => true

;; with vector syntax
(m/validate :int 1)
; => true

(m/validate :int "1")
; => false

(m/validate [:and :int [:> 6]] 7)
; => true

(m/validate [:qualified-keyword {:namespace :aaa}] :aaa/bbb)
; => true

;; optimized (pure) validation function for best performance
(def valid?
  (m/validator
    [:map
     [:x :boolean]
     [:y {:optional true} :int]
     [:z :string]]))

(valid? {:x true, :z "kikka"})
; => true

Schemas can have properties:

(def Age
  [:and
   {:title "Age"
    :description "It's an age"
    :json-schema/example 20}
   :int [:> 18]])
   
(m/properties Age)
; => {:title "Age"
;     :description "It's an age"
;     :json-schema/example 20}

Maps are open by default:

(m/validate
  [:map [:x :int]]
  {:x 1, :extra "key"})
; => true

Maps can be closed with :closed property:

(m/validate
  [:map {:closed true} [:x :int]]
  {:x 1, :extra "key"})
; => false

Maps keys are not limited to keywords:

(m/validate
  [:map
   ["status" [:enum "ok"]]
   [1 :any]
   [nil :any]
   [::a :string]]
  {"status" "ok"
   1 'number
   nil :yay
   ::a "properly awesome"})
; => true

Most core-predicates are mapped to Schemas:

(m/validate string? "kikka")
; => true

See the full list of default schemas.

Qualified keys in a map

You can also use decomplected maps keys and values using registry references. References must be either qualified keywords or strings.

(m/validate
  [:map {:registry {::id int?
                    ::country string?}}
   ::id
   [:name string?]
   [::country {:optional true}]]
  {::id 1
   :name "kikka"})
; => true

Homogeneous Maps

Other times, we use a map as a homogeneous index. In this case, all our key-value pairs have the same type. For this use case, we can use the :map-of schema.

(m/validate 
  [:map-of :string [:map [:lat number?] [:long number?]]]
  {"oslo" {:lat 60 :long 11}
   "helsinki" {:lat 60 :long 24}})
;; => true

Sequence Schemas

You can use :sequential for any homogeneous Clojure sequence, :vector for vectors and :set for sets.

(m/validate [:sequential any?] (list "this" 'is :number 42))
;; => true

(m/validate [:vector int?] [1 2 3])
;; => true

(m/validate [:vector int?] (list 1 2 3))
;; => false

A :tuple describes a fixed length Clojure vector of heterogeneous elements:

(m/validate [:tuple keyword? string? number?] [:bing "bang" 42])
;; => true

Malli also supports sequence regexes like Seqexp and Spec. The supported operators are :cat & :catn for concatenation / sequencing

(m/validate [:cat string? int?] ["foo" 0]) ; => true

(m/validate [:catn [:s string?] [:n int?]] ["foo" 0]) ; => true

:alt & :altn for alternatives

(m/validate [:alt keyword? string?] ["foo"]) ; => true

(m/validate [:altn [:kw keyword?] [:s string?]] ["foo"]) ; => true

and :?, :*, :+ & :repeat for repetition:

(m/validate [:? int?] []) ; => true
(m/validate [:? int?] [1]) ; => true
(m/validate [:? int?] [1 2]) ; => false

(m/validate [:* int?] []) ; => true
(m/validate [:* int?] [1 2 3]) ; => true

(m/validate [:+ int?] []) ; => false
(m/validate [:+ int?] [1]) ; => true
(m/validate [:+ int?] [1 2 3]) ; => true

(m/validate [:repeat {:min 2, :max 4} int?] [1]) ; => false
(m/validate [:repeat {:min 2, :max 4} int?] [1 2]) ; => true
(m/validate [:repeat {:min 2, :max 4} int?] [1 2 3 4]) ; => true (:max is inclusive, as elsewhere in Malli)
(m/validate [:repeat {:min 2, :max 4} int?] [1 2 3 4 5]) ; => false

:catn and :altn allow naming the subsequences / alternatives

(m/explain
  [:* [:catn [:prop string?] [:val [:altn [:s string?] [:b boolean?]]]]]
  ["-server" "foo" "-verbose" 11 "-user" "joe"])
;; => {:schema [:* [:map [:prop string?] [:val [:map [:s string?] [:b boolean?]]]]],
;;     :value ["-server" "foo" "-verbose" 11 "-user" "joe"],
;;     :errors (#Error{:path [0 :val :s], :in [3], :schema string?, :value 11}
;;              #Error{:path [0 :val :b], :in [3], :schema boolean?, :value 11})}

while :cat and :alt just use numeric indices for paths:

(m/explain 
  [:* [:cat string? [:alt string? boolean?]]]
  ["-server" "foo" "-verbose" 11 "-user" "joe"])
;; => {:schema [:* [:cat string? [:alt string? boolean?]]],
;;     :value ["-server" "foo" "-verbose" 11 "-user" "joe"],
;;     :errors (#Error{:path [0 1 0], :in [3], :schema string?, :value 11}
;;              #Error{:path [0 1 1], :in [3], :schema boolean?, :value 11})}

As all these examples show, the "seqex" operators take any non-seqex child schema to mean a sequence of one element that matches that schema. To force that behaviour for a seqex child :schema can be used:

(m/validate
  [:cat [:= :names] [:schema [:* string?]] [:= :nums] [:schema [:* number?]]]
  [:names ["a" "b"] :nums [1 2 3]])
; => true

;; whereas
(m/validate
  [:cat [:= :names] [:* string?] [:= :nums] [:* number?]]
  [:names "a" "b" :nums 1 2 3]) 
; => true

Although a lot of effort has gone into making the seqex implementation fast

(require '[clojure.spec.alpha :as s])
(require '[criterium.core :as cc])

(let [valid? (partial s/valid? (s/* int?))]
  (cc/quick-bench (valid? (range 10)))) ; Execution time mean : 27µs

(let [valid? (m/validator [:* int?])]
  (cc/quick-bench (valid? (range 10)))) ; Execution time mean : 2.7µs

it is always better to use less general tools whenever possible:

(let [valid? (partial s/valid? (s/coll-of int?))]
  (cc/quick-bench (valid? (range 10)))) ; Execution time mean : 1.8µs

(let [valid? (m/validator [:sequential int?])]
  (cc/quick-bench (valid? (range 10)))) ; Execution time mean : 0.12µs

String schemas

Using a predicate:

(m/validate string? "kikka")

Using :string Schema:

(m/validate :string "kikka")
;; => true

(m/validate [:string {:min 1, :max 4}] "")
;; => false

Using regular expressions:

(m/validate #"a+b+c+" "abbccc")
;; => true

;; :re with string
(m/validate [:re ".{3,5}"] "abc")
;; => true

;; :re with regex
(m/validate [:re #".{3,5}"] "abc")
;; => true

;; NB: re-find semantics
(m/validate [:re #"\d{4}"] "1234567")
;; => true

;; anchor with ^...$ if you want to strictly match the whole string
(m/validate [:re #"^\d{4}$"] "1234567")
;; => false

Maybe schemas

Use :maybe to express that an element should match some schema OR be nil:

(m/validate [:maybe string?] "bingo")
;; => true

(m/validate [:maybe string?] nil)
;; => true

(m/validate [:maybe string?] :bingo)
;; => false

Fn schemas

:fn allows any predicate function to be used:

(def my-schema
  [:and
   [:map
    [:x int?]
    [:y int?]]
   [:fn (fn [{:keys [x y]}] (> x y))]])
   
(m/validate my-schema {:x 1, :y 0})
; => true

(m/validate my-schema {:x 1, :y 2})
; => false

Error Messages

Detailed errors with m/explain:

(def Address
  [:map
   [:id string?]
   [:tags [:set keyword?]]
   [:address
    [:map
     [:street string?]
     [:city string?]
     [:zip int?]
     [:lonlat [:tuple double? double?]]]]])

(m/explain
  Address
  {:id "Lillan"
   :tags #{:artesan :coffee :hotel}
   :address {:street "Ahlmanintie 29"
             :city "Tampere"
             :zip 33100
             :lonlat [61.4858322, 23.7854658]}})
; => nil

(m/explain
  Address
  {:id "Lillan"
   :tags #{:artesan "coffee" :garden}
   :address {:street "Ahlmanintie 29"
             :zip 33100
             :lonlat [61.4858322, nil]}})
;{:schema [:map
;          [:id string?]
;          [:tags [:set keyword?]]
;          [:address [:map
;                     [:street string?]
;                     [:city string?]
;                     [:zip int?]
;                     [:lonlat [:tuple double? double?]]]]],
; :value {:id "Lillan",
;         :tags #{:artesan :garden "coffee"},
;         :address {:street "Ahlmanintie 29"
;                   :zip 33100
;                   :lonlat [61.4858322 nil]}},
; :errors (#Error{:path [:tags 0]
;                 :in [:tags 0]
;                 :schema keyword?
;                 :value "coffee"}
;          #Error{:path [:address :city],
;                 :in [:address :city],
;                 :schema [:map
;                          [:street string?]
;                          [:city string?]
;                          [:zip int?]
;                          [:lonlat [:tuple double? double?]]],
;                 :type :malli.core/missing-key}
;          #Error{:path [:address :lonlat 1]
;                 :in [:address :lonlat 1]
;                 :schema double?
;                 :value nil})}

Custom Error Messages

Explain results can be humanized with malli.error/humanize:

(require '[malli.error :as me])

(-> Address
    (m/explain
      {:id "Lillan"
       :tags #{:artesan "coffee" :garden}
       :address {:street "Ahlmanintie 29"
                 :zip 33100
                 :lonlat [61.4858322, nil]}})
    (me/humanize))
;{:tags #{["should be a keyword"]}
; :address {:city ["missing required key"]
;           :lonlat [nil ["should be a double"]]}}

Error messages can be customized with :error/message and :error/fn properties:

(-> [:map
     [:id int?]
     [:size [:enum {:error/message "should be: S|M|L"} 
             "S" "M" "L"]]
     [:age [:fn {:error/fn '(fn [{:keys [value]} _] (str value ", should be > 18"))}
            '(fn [x] (and (int? x) (> x 18)))]]]
    (m/explain {:size "XL", :age 10})
    (me/humanize
      {:errors (-> me/default-errors
                   (assoc ::m/missing-key {:error/fn (fn [{:keys [in]} _] (str "missing key " (last in)))}))}))
;{:id ["missing key :id"]
; :size ["should be: S|M|L"]
; :age ["10, should be > 18"]}

Messages can be localized:

(-> [:map
     [:id int?]
     [:size [:enum {:error/message {:en "should be: S|M|L"
                                    :fi "pitäisi olla: S|M|L"}}
             "S" "M" "L"]]
     [:age [:fn {:error/fn {:en '(fn [{:keys [value]} _] (str value ", should be > 18"))
                            :fi '(fn [{:keys [value]} _] (str value ", pitäisi olla > 18"))}}
            '(fn [x] (and (int? x) (> x 18)))]]]
    (m/explain {:size "XL", :age 10})
    (me/humanize
      {:locale :fi
       :errors (-> me/default-errors
                   (assoc-in ['int? :error-message :fi] "pitäisi olla numero")
                   (assoc ::m/missing-key {:error/fn {:en '(fn [{:keys [in]} _] (str "missing key " (last in)))
                                                      :fi '(fn [{:keys [in]} _] (str "puuttuu avain " (last in)))}}))}))
;{:id ["puuttuu avain :id"]
; :size ["pitäisi olla: S|M|L"]
; :age ["10, pitäisi olla > 18"]}

Top-level humanized map-errors are under :malli/error:

(-> [:and [:map
           [:password string?]
           [:password2 string?]]
     [:fn {:error/message "passwords don't match"}
      '(fn [{:keys [password password2]}]
         (= password password2))]]
    (m/explain {:password "secret"
                :password2 "faarao"})
    (me/humanize))
; {:malli/error ["passwords don't match"]}

Errors can be targeted using :error/path property:

(-> [:and [:map
           [:password string?]
           [:password2 string?]]
     [:fn {:error/message "passwords don't match"
           :error/path [:password2]}
      '(fn [{:keys [password password2]}]
         (= password password2))]]
    (m/explain {:password "secret"
                :password2 "faarao"})
    (me/humanize))
; {:password2 ["passwords don't match"]}

By default, only direct erroneous schema properties are used:

(-> [:map
     [:foo {:error/message "entry-failure"} :int]] ;; here, :int fails, no error props
    (m/explain {:foo "1"})
    (me/humanize))
; => {:foo ["should be an integer"]}

Looking up humanized errors from parent schemas with custom :resolve (BETA, subject to change):

(-> [:map
     [:foo {:error/message "entry-failure"} :int]]
    (m/explain {:foo "1"})
    (me/humanize {:resolve me/-resolve-root-error}))
; => {:foo ["entry-failure"]}

Spell Checking

For closed schemas, key spelling can be checked with:

(-> [:map [:address [:map [:street string?]]]]
    (mu/closed-schema)
    (m/explain
      {:name "Lie-mi"
       :address {:streetz "Hämeenkatu 14"}})
    (me/with-spell-checking)
    (me/humanize))
;{:address {:street ["missing required key"]
;           :streetz ["should be spelled :street"]}
; :name ["disallowed key"]}

Value Transformation

(require '[malli.transform :as mt])

Two-way schema-driven value transformations with m/decode and m/encode using a m/Transformer.

Default Transformers include: string-transformer, json-transformer, strip-extra-keys-transformer, default-value-transformer and key-transformer.

NOTE: the included transformers are best-effort, i.e. they won't throw on bad input, they will just pass the input value through unchanged. You should make sure your schema validation catches these non-transformed values. Custom transformers should follow the same idiom.

(m/decode int? "42" mt/string-transformer)
; 42

(m/encode int? 42 mt/string-transformer)
; "42"

Transformations are recursive:

(m/decode
  Address
  {:id "Lillan",
   :tags ["coffee" "artesan" "garden"],
   :address {:street "Ahlmanintie 29"
             :city "Tampere"
             :zip 33100
             :lonlat [61.4858322 23.7854658]}}
  mt/json-transformer)
;{:id "Lillan",
; :tags #{:coffee :artesan :garden},
; :address {:street "Ahlmanintie 29"
;           :city "Tampere"
;           :zip 33100
;           :lonlat [61.4858322 23.7854658]}}

Transform map keys:

(m/encode
  Address
  {:id "Lillan",
   :tags ["coffee" "artesan" "garden"],
   :address {:street "Ahlmanintie 29"
             :city "Tampere"
             :zip 33100
             :lonlat [61.4858322 23.7854658]}}
  (mt/key-transformer {:encode name}))
;{"id" "Lillan",
; "tags" ["coffee" "artesan" "garden"],
; "address" {"street" "Ahlmanintie 29"
;            "city" "Tampere"
;            "zip" 33100
;            "lonlat" [61.4858322 23.7854658]}}

Transformers can be composed with mt/transformer:

(def strict-json-transformer
  (mt/transformer
    mt/strip-extra-keys-transformer
    mt/json-transformer))

(m/decode
  Address
  {:id "Lillan",
   :EVIL "LYN"
   :tags ["coffee" "artesan" "garden"],
   :address {:street "Ahlmanintie 29"
             :DARK "ORKO"
             :city "Tampere"
             :zip 33100
             :lonlat [61.4858322 23.7854658]}}
  strict-json-transformer)
;{:id "Lillan",
; :tags #{:coffee :artesan :garden},
; :address {:street "Ahlmanintie 29"
;           :city "Tampere"
;           :zip 33100
;           :lonlat [61.4858322 23.7854658]}}

Schema properties can be used to override default transformations:

(m/decode
  [string? {:decode/string 'str/upper-case}]
  "kerran" mt/string-transformer)
; => "KERRAN"

Decoders and encoders as interceptors (with :enter and :leave stages):

(m/decode
  [string? {:decode/string {:enter 'str/upper-case}}]
  "kerran" mt/string-transformer)
; => "KERRAN"

(m/decode
  [string? {:decode/string {:enter '#(str "olipa_" %)
                            :leave '#(str % "_avaruus")}}]
  "kerran" mt/string-transformer)
; => "olipa_kerran_avaruus"

To access Schema (and options) use :compile:

(m/decode
  [int? {:math/multiplier 10
         :decode/math {:compile '(fn [schema _]
                                  (let [multiplier (:math/multiplier (m/properties schema))]
                                    (fn [x] (* x multiplier))))}}]
  12
  (mt/transformer {:name :math}))
; => 120

Going crazy:

(m/decode
  [:map
   {:decode/math {:enter '#(update % :x inc)
                  :leave '#(update % :x (partial * 2))}}
   [:x [int? {:decode/math {:enter '(partial + 2)
                            :leave '(partial * 3)}}]]]
  {:x 1} 
  (mt/transformer {:name :math}))
; => {:x 24}

Default values

Applying default values:

(m/decode [:and {:default 42} int?] nil mt/default-value-transformer)
; => 42

With custom key and type defaults:

(m/decode
  [:map
   [:user [:map
           [:name :string]
           [:description {:ui/default "-"} :string]]]]
  nil
  (mt/default-value-transformer
    {:key :ui/default
     :defaults {:map (constantly {})
                :string (constantly "")}}))
; => {:user {:name "", :description "-"}}

Single sweep of defaults & string encoding:

(m/encode
  [:map {:default {}}
   [:a [int? {:default 1}]]
   [:b [:vector {:default [1 2 3]} int?]]
   [:c [:map {:default {}}
        [:x [int? {:default 42}]]
        [:y int?]]]
   [:d [:map
        [:x [int? {:default 42}]]
        [:y int?]]]
   [:e int?]]
  nil
  (mt/transformer
    mt/default-value-transformer
    mt/string-transformer))
;{:a "1"
; :b ["1" "2" "3"]
; :c {:x "42"}}

Programming with Schemas

(require '[malli.util :as mu])

Updating Schema properties:

(mu/update-properties [:vector int?] assoc :min 1)
; => [:vector {:min 1} int?]

Lifted clojure.core function to work with schemas: select-keys, dissoc, get, assoc, update, get-in, assoc-in, update-in

(mu/get-in Address [:address :lonlat])
; => [:tuple double? double?]

(mu/update-in Address [:address] mu/assoc :country [:enum "fi" "po"])
;[:map
; [:id string?]
; [:tags [:set keyword?]]
; [:address
;  [:map [:street string?]
;   [:city string?]
;   [:zip int?]
;   [:lonlat [:tuple double? double?]]
;   [:country [:enum "fi" "po"]]]]]

(-> Address
    (mu/dissoc :address)
    (mu/update-properties assoc :title "Address"))
;[:map {:title "Address"} 
; [:id string?] 
; [:tags [:set keyword?]]]

Making keys optional or required:

(mu/optional-keys [:map [:x int?] [:y int?]])
;[:map 
; [:x {:optional true} int?] 
; [:y {:optional true} int?]]

(mu/required-keys [:map [:x {:optional true} int?] [:y int?]])
;[:map 
; [:x int?] 
; [:y int?]]

Closing and opening all :map schemas recursively:

(def abcd
  [:map {:title "abcd"}
   [:a int?]
   [:b {:optional true} int?]
   [:c [:map
        [:d int?]]]])

(mu/closed-schema abcd)
;[:map {:title "abcd", :closed true}
; [:a int?]
; [:b {:optional true} int?]
; [:c [:map {:closed true}
;      [:d int?]]]]

(-> abcd 
    mu/closed-schema
    mu/open-schema)
;[:map {:title "abcd"}
; [:a int?]
; [:b {:optional true} int?]
; [:c [:map
;      [:d int?]]]]

Merging Schemas (last value wins):

(mu/merge
  [:map
   [:name string?]
   [:description string?]
   [:address
    [:map
     [:street string?]
     [:country [:enum "finland" "poland"]]]]]
  [:map
   [:description {:optional true} string?]
   [:address
    [:map
     [:country string?]]]])
;[:map
; [:name string?]
; [:description {:optional true} string?]
; [:address [:map
;            [:street string?]
;            [:country string?]]]]

With :and, first child is used in merge:

(mu/merge
  [:and {:type "entity"}
   [:map {:title "user"} 
    [:name :string]]
   map?]
  [:map {:description "aged"} [:age :int]])
;[:and {:type "entity"} 
; [:map {:title "user", :description "aged"} 
;  [:name :string] 
;  [:age :int]] 
; map?]

Schema unions (merged values of both schemas are valid for union schema):

(mu/union
  [:map
   [:name string?]
   [:description string?]
   [:address
    [:map
     [:street string?]
     [:country [:enum "finland" "poland"]]]]]
  [:map
   [:description {:optional true} string?]
   [:address
    [:map
     [:country string?]]]])
;[:map
; [:name string?]
; [:description {:optional true} string?]
; [:address [:map
;            [:street string?]
;            [:country [:or [:enum "finland" "poland"] string?]]]]]

Adding generated example values to Schemas:

(m/walk
  [:map
   [:name string?]
   [:description string?]
   [:address
    [:map
     [:street string?]
     [:country [:enum "finland" "poland"]]]]]
  (m/schema-walker
    (fn [schema]
      (mu/update-properties schema assoc :examples (mg/sample schema {:size 2, :seed 20})))))
;[:map
; {:examples ({:name "", :description "", :address {:street "", :country "poland"}}
;             {:name "W", :description "x", :address {:street "8", :country "finland"}})}
; [:name [string? {:examples ("" "")}]]
; [:description [string? {:examples ("" "")}]]
; [:address
;  [:map
;   {:examples ({:street "", :country "finland"} {:street "W", :country "poland"})}
;   [:street [string? {:examples ("" "")}]]
;   [:country [:enum {:examples ("finland" "poland")} "finland" "poland"]]]]]

Finding first value (prewalk):

(mu/find-first
  [:map
   [:x int?]
   [:y [:vector [:tuple
                 [:or [:and {:salaisuus "turvassa"} boolean?] int?]
                 [:schema {:salaisuus "vaarassa"} false?]]]]
   [:z [:string {:salaisuus "piilossa"}]]]
  (fn [schema _ _]
    (-> schema m/properties :salaisuus)))
; => "turvassa"

Finding all subschemas with paths, retaining order:

(def Schema
  (m/schema
    [:maybe
     [:map
      [:id string?]
      [:tags [:set keyword?]]
      [:address
       [:and
        [:map
         [:street {:optional true} string?]
         [:lonlat {:optional true} [:tuple double? double?]]]
        [:fn '(fn [{:keys [street lonlat]}] (or street lonlat))]]]]]))

(mu/subschemas Schema)
;[{:path [], :in [], :schema [:maybe
;                             [:map
;                              [:id string?]
;                              [:tags [:set keyword?]]
;                              [:address
;                               [:and
;                                [:map
;                                 [:street {:optional true} string?]
;                                 [:lonlat {:optional true} [:tuple double? double?]]]
;                                [:fn (fn [{:keys [street lonlat]}] (or street lonlat))]]]]]}
; {:path [0], :in [], :schema [:map
;                              [:id string?]
;                              [:tags [:set keyword?]]
;                              [:address
;                               [:and
;                                [:map
;                                 [:street {:optional true} string?]
;                                 [:lonlat {:optional true} [:tuple double? double?]]]
;                                [:fn (fn [{:keys [street lonlat]}] (or street lonlat))]]]]}
; {:path [0 :id], :in [:id], :schema string?}
; {:path [0 :tags], :in [:tags], :schema [:set keyword?]}
; {:path [0 :tags :malli.core/in], :in [:tags :malli.core/in], :schema keyword?}
; {:path [0 :address], :in [:address], :schema [:and
;                                               [:map
;                                                [:street {:optional true} string?]
;                                                [:lonlat {:optional true} [:tuple double? double?]]]
;                                               [:fn (fn [{:keys [street lonlat]}] (or street lonlat))]]}
; {:path [0 :address 0], :in [:address], :schema [:map
;                                                 [:street {:optional true} string?]
;                                                 [:lonlat {:optional true} [:tuple double? double?]]]}
; {:path [0 :address 0 :street], :in [:address :street], :schema string?}
; {:path [0 :address 0 :lonlat], :in [:address :lonlat], :schema [:tuple double? double?]}
; {:path [0 :address 0 :lonlat 0], :in [:address :lonlat 0], :schema double?}
; {:path [0 :address 0 :lonlat 1], :in [:address :lonlat 1], :schema double?}
; {:path [0 :address 1], :in [:address], :schema [:fn (fn [{:keys [street lonlat]}] (or street lonlat))]}]

Collecting unique value paths and their schema paths:

(->> Schema
     (mu/subschemas)
     (mu/distinct-by :id)
     (mapv (juxt :in :path)))
;[[[] []]
; [[] [0]]
; [[:id] [0 :id]]
; [[:tags] [0 :tags]]
; [[:tags :malli.core/in] [0 :tags :malli.core/in]]
; [[:address] [0 :address]]
; [[:address] [0 :address 0]]
; [[:address :street] [0 :address 0 :street]]
; [[:address :lonlat] [0 :address 0 :lonlat]]
; [[:address :lonlat 0] [0 :address 0 :lonlat 0]]
; [[:address :lonlat 1] [0 :address 0 :lonlat 1]]
; [[:address] [0 :address 1]]]

Schema paths can be converted into value paths:

(mu/get-in Schema [0 :address 0 :lonlat])
; => [:tuple double? double?]

(mu/path->in Schema [0 :address 0 :lonlat])
; => [:address :lonlat]

and back, returning all paths:

(mu/in->paths Schema [:address :lonlat])
; => [[0 :address 0 :lonlat]]

Declarative Schema Transformation

There are also declarative versions of schema transforming utilities in malli.util/schemas. These include :merge, :union and :select-keys:

(def registry (merge (m/default-schemas) (mu/schemas)))

(def Merged
  (m/schema
    [:merge
     [:map [:x :string]]
     [:map [:y :int]]]
    {:registry registry}))

Merged
;[:merge
; [:map [:x :string]]
; [:map [:y :int]]]

(m/deref Merged)
;[:map 
; [:x :string] 
; [:y :int]]

(m/validate Merged {:x "kikka", :y 6})
; => true

Persisting Schemas

Writing and Reading schemas as EDN, no eval needed.

(require '[malli.edn :as edn])

(-> [:and
     [:map
      [:x int?]
      [:y int?]]
     [:fn '(fn [{:keys [x y]}] (> x y))]]
    (edn/write-string)
    (doto prn) ; => "[:and [:map [:x int?] [:y int?]] [:fn (fn [{:keys [x y]}] (> x y))]]"
    (edn/read-string)
    (doto (-> (m/validate {:x 0, :y 1}) prn)) ; => false
    (doto (-> (m/validate {:x 2, :y 1}) prn))) ; => true
;[:and 
; [:map 
;  [:x int?] 
;  [:y int?]] 
; [:fn (fn [{:keys [x y]}] (> x y))]]

Multi Schemas

Closed dispatch with :multi schema and :dispatch property:

(m/validate
  [:multi {:dispatch :type}
   [:sized [:map [:type keyword?] [:size int?]]]
   [:human [:map [:type keyword?] [:name string?] [:address [:map [:country keyword?]]]]]]
  {:type :sized, :size 10})
; true

Default branch with ::m/default:

(def valid?
  (m/validator
    [:multi {:dispatch :type}
     ["object" [:map-of :keyword :string]]
     [::m/default :string]]))

(valid? {:type "object", :key "1", :value "100"})
; => true

(valid? "SUCCESS!")
; => true

(valid? :failure)
; => false

Any (serializable) function can be used for :dispatch:

(m/validate
  [:multi {:dispatch 'first}
   [:sized [:tuple keyword? [:map [:size int?]]]]
   [:human [:tuple keyword? [:map [:name string?] [:address [:map [:country keyword?]]]]]]]
  [:human {:name "seppo", :address {:country :sweden}}])
; true

:dispatch values should be decoded before actual values:

(m/decode
  [:multi {:dispatch :type
           :decode/string '#(update % :type keyword)}
   [:sized [:map [:type [:= :sized]] [:size int?]]]
   [:human [:map [:type [:= :human]] [:name string?] [:address [:map [:country keyword?]]]]]]
  {:type "human"
   :name "Tiina"
   :age "98"
   :address {:country "finland"
             :street "this is an extra key"}}
  (mt/transformer mt/strip-extra-keys-transformer mt/string-transformer))
;{:type :human
; :name "Tiina"
; :address {:country :finland}}

Recursive Schemas

To create a recursive schema, introduce a local registry and wrap all recursive positions in the registry with :ref. Now you may reference the recursive schemas in the body of the schema.

For example, here is a recursive schema using :schema for singly-linked lists of positive integers:

(m/validate
  [:schema {:registry {::cons [:maybe [:tuple pos-int? [:ref ::cons]]]}}
   ::cons]
  [16 [64 [26 [1 [13 nil]]]]])
; => true

Without the :ref keyword, malli eagerly expands the schema until a stack overflow error is thrown:

(m/validate
  [:schema {:registry {::cons [:maybe [:tuple pos-int? ::cons]]}}
   ::cons]
  [16 [64 [26 [1 [13 nil]]]]])
; StackOverflowError

Mutual recursion works too. Thanks to the :schema construct, many schemas could be defined in the local registry, the top-level one being promoted by the :schema second parameter:

(m/validate
  [:schema {:registry {::ping [:maybe [:tuple [:= "ping"] [:ref ::pong]]]
                       ::pong [:maybe [:tuple [:= "pong"] [:ref ::ping]]]}}
   ::ping]
  ["ping" ["pong" ["ping" ["pong" ["ping" nil]]]]])
; => true

Nested registries, the last definition wins:

(m/validate
  [:schema {:registry {::ping [:maybe [:tuple [:= "ping"] [:ref ::pong]]]
                       ::pong any?}} ;; effectively unreachable
   [:schema {:registry {::pong [:maybe [:tuple [:= "pong"] [:ref ::ping]]]}}
    ::ping]]
  ["ping" ["pong" ["ping" ["pong" ["ping" nil]]]]])
; => true

Value Generation

Schemas can be used to generate values:

(require '[malli.generator :as mg])

;; random
(mg/generate keyword?)
; => :?

;; using seed
(mg/generate [:enum "a" "b" "c"] {:seed 42})
;; => "a"

;; using seed and size
(mg/generate pos-int? {:seed 10, :size 100})
;; => 55740

;; regexs work too (only clj and if [com.gfredericks/test.chuck "0.2.10"+] available)
(mg/generate 
  [:re #"^[a-zA-Z0-9._%+-]+@[a-zA-Z0-9.-]+\.[a-zA-Z]{2,63}$"] 
  {:seed 42, :size 10})
; => "[email protected]"

;; :gen/elements (note, are not validated)
(mg/generate
  [:and {:gen/elements ["kikka" "kukka" "kakka"]} string?]
  {:seed 10})
; => "kikka"

;; portable :gen/fmap
(mg/generate
  [:and {:gen/fmap '(partial str "kikka_")} string?]
  {:seed 10, :size 10})
;; => "kikka_WT3K0yax2"

;; :gen/schema
(mg/generate
  [:any {:gen/schema [:int {:min 10, :max 20}]}]
  {:seed 10})
; => 19

;; :gen/min & :gen/max for numbers and collections
(mg/generate 
  [:vector {:gen/min 4, :gen/max 4} :int] '
  {:seed 1})
; => [-8522515 -1433 -1 1]

;; :gen/infinite? & :gen/NaN? for :double
(mg/generate 
  [:double {:gen/infinite? true, :gen/NaN? true}] 
  {:seed 1})
; => ##Inf

(require '[clojure.test.check.generators :as gen])

;; gen/gen (note, not serializable)
(mg/generate
  [:sequential {:gen/gen (gen/list gen/neg-int)} int?]
  {:size 42, :seed 42})
; => (-37 -13 -13 -24 -20 -11 -34 -40 -22 0 -10)

Generated values are valid:

(mg/generate Address {:seed 123, :size 4})
;{:id "H7",
; :tags #{:v?.w.t6!.QJYk-/-?s*4
;         :_7U
;         :QdG/Xi8J
;         :*Q-.p*8*/n-J9u}
; :address {:street "V9s"
;           :city ""
;           :zip 3
;           :lonlat [-2.75 -0.625]}}

(m/validate Address (mg/generate Address))
; => true

Sampling values:

;; sampling
(mg/sample [:and int? [:> 10] [:< 100]] {:seed 123})
; => (25 39 51 13 53 43 57 15 26 27)

Integration with test.check:

(require '[clojure.test.check.generators :as gen])
(gen/sample (mg/generator pos-int?))
; => (2 1 2 2 2 2 8 1 55 83)

Inferring Schemas

Inspired by F# Type providers:

(require '[malli.provider :as mp])

(def samples
  [{:id "Lillan"
    :tags #{:artesan :coffee :hotel}
    :address {:street "Ahlmanintie 29"
              :city "Tampere"
              :zip 33100
              :lonlat [61.4858322, 23.7854658]}}
   {:id "Huber",
    :description "Beefy place"
    :tags #{:beef :wine :beer}
    :address {:street "Aleksis Kiven katu 13"
              :city "Tampere"
              :zip 33200
              :lonlat [61.4963599 23.7604916]}}])

(mp/provide samples)
;[:map
; [:id string?]
; [:tags [:set keyword?]]
; [:address
;  [:map
;   [:street string?]
;   [:city string?]
;   [:zip number?]
;   [:lonlat [:vector double?]]]]
; [:description {:optional true} string?]]

All samples are valid against the inferred schema:

(every? (partial m/validate (mp/provide samples)) samples)
; => true

For order of magnitude better performance, use mp/provider instead:

;; 3.6ms -> 2.1ms (1.7x)
(p/bench (mp/provide [1 2 3]))

;; 2.5ms -> 82µs (30x)
(let [provider (mp/provider)]
  (p/bench (provider [1 2 3])))

Parsing values

Schemas can be used to parse values using m/parse and m/parser:

m/parse for one-time things:

(m/parse
  [:* [:catn
       [:prop string?]
       [:val [:altn
              [:s string?]
              [:b boolean?]]]]]
  ["-server" "foo" "-verbose" true "-user" "joe"])
;[{:prop "-server", :val [:s "foo"]}
; {:prop "-verbose", :val [:b true]}
; {:prop "-user", :val [:s "joe"]}]

m/parser to create an optimized parser:

(def Hiccup
  [:schema {:registry {"hiccup" [:orn
                                 [:node [:catn
                                         [:name keyword?]
                                         [:props [:? [:map-of keyword? any?]]]
                                         [:children [:* [:schema [:ref "hiccup"]]]]]]
                                 [:primitive [:orn
                                              [:nil nil?]
                                              [:boolean boolean?]
                                              [:number number?]
                                              [:text string?]]]]}}
   "hiccup"])

(def parse-hiccup (m/parser Hiccup))

(parse-hiccup
  [:div {:class [:foo :bar]}
   [:p "Hello, world of data"]])
;[:node
; {:name :div
;  :props {:class [:foo :bar]}
;  :children [[:node
;              {:name :p
;               :props nil
;               :children [[:primitive [:text "Hello, world of data"]]]}]]}]

Parsing returns tagged values for :orn, :catn, :altn and :multi.

(def Multi
  [:multi {:dispatch :type}
   [:user [:map [:size :int]]]
   [::m/default :any]])

(m/parse Multi {:type :user, :size 1})
; => [:user {:type :user, :size 1}]

(m/parse Multi {:type "sized", :size 1})
; => [:malli.core/default {:type "sized", :size 1}]

Unparsing values

The inverse of parsing, using m/unparse and m/unparser:

(->> [:div {:class [:foo :bar]}
      [:p "Hello, world of data"]]
     (m/parse Hiccup)
     (m/unparse Hiccup))
;[:div {:class [:foo :bar]}
; [:p "Hello, world of data"]]

Serializable Functions

Enabling serializable function schemas requires sci as external dependency. If it is not present, the malli function evaluator throws :sci-not-available exception.

For ClojureScript, you also need to require sci.core manually, either directly or via :preloads.

For GraalVM, you need to require sci.core manually, before requiring any malli namespaces.

(def my-schema
  [:and
   [:map
    [:x int?]
    [:y int?]]
   [:fn '(fn [{:keys [x y]}] (> x y))]])

(m/validate my-schema {:x 1, :y 0})
; => true

(m/validate my-schema {:x 1, :y 2})
; => false

NOTE: sci is not termination safe so be wary of sci functions from untrusted sources. You can explicitly disable sci with option ::m/disable-sci and set the default options with ::m/sci-options.

(m/validate [:fn 'int?] 1 {::m/disable-sci true})
; Execution error
; :malli.core/sci-not-available {:code int?}

Schema AST

NOTE: Map Syntax / SchemaAST is considered as alpha and subject to change.

Implemented with protocol malli.core/AST. Allows lossless round-robin with faster schema creation.

NOTE: For now, the AST syntax in concidered as internal, e.g. don't use it as a database persistency model.

(def ?schema
  [:map
   [:x boolean?]
   [:y {:optional true} int?]
   [:z [:map
        [:x boolean?]
        [:y {:optional true} int?]]]])

(m/form ?schema)
;[:map
; [:x boolean?]
; [:y {:optional true} int?]
; [:z [:map
;      [:x boolean?]
;      [:y {:optional true} int?]]]]

(m/ast ?schema)
;{:type :map,
; :keys {:x {:order 0
;            :value {:type boolean?}},
;        :y {:order 1, :value {:type int?}
;            :properties {:optional true}},
;        :z {:order 2,
;            :value {:type :map,
;                    :keys {:x {:order 0
;                               :value {:type boolean?}},
;                           :y {:order 1
;                               :value {:type int?}
;                               :properties {:optional true}}}}}}}

(-> ?schema
    (m/schema) ;; 3.4µs
    (m/ast)
    (m/from-ast) ;; 180ns (18x, lazy)
    (m/form)
    (= (m/form ?schema)))
; => true

Schema Transformation

Schemas can be transformed using post-walking, e.g. the Visitor Pattern.

The identity walker:

(m/walk 
  Address 
  (m/schema-walker identity))
;[:map
; [:id string?]
; [:tags [:set keyword?]]
; [:address
;  [:map
;   [:street string?]
;   [:city string?]
;   [:zip int?]
;   [:lonlat [:tuple double? double?]]]]]

Adding :title property to schemas:

(m/walk
  Address
  (m/schema-walker #(mu/update-properties % assoc :title (name (m/type %)))))
;[:map {:title "map"}
; [:id [string? {:title "string?"}]]
; [:tags [:set {:title "set"} [keyword? {:title "keyword?"}]]]
; [:address
;  [:map {:title "map"}
;   [:street [string? {:title "string?"}]]
;   [:city [string? {:title "string?"}]]
;   [:zip [int? {:title "int?"}]]
;   [:lonlat [:tuple {:title "tuple"} [double? {:title "double?"}] [double? {:title "double?"}]]]]]]

Transforming schemas into maps:

(m/walk
  Address
  (fn [schema _ children _]
    (-> (m/properties schema)
        (assoc :malli/type (m/type schema))
        (cond-> (seq children) (assoc :malli/children children)))))
;{:malli/type :map,
; :malli/children [[:id nil {:malli/type string?}]
;                  [:tags nil {:malli/type :set
;                              :malli/children [{:malli/type keyword?}]}]
;                  [:address nil {:malli/type :map,
;                                 :malli/children [[:street nil {:malli/type string?}]
;                                                  [:city nil {:malli/type string?}]
;                                                  [:zip nil {:malli/type int?}]
;                                                  [:lonlat nil {:malli/type :tuple
;                                                                :malli/children [{:malli/type double?}
;                                                                                 {:malli/type double?}]}]]}]]}

JSON Schema

Transforming Schemas into JSON Schema:

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

(json-schema/transform Address)
;{:type "object",
; :properties {:id {:type "string"},
;              :tags {:type "array"
;                     :items {:type "string"}
;                     :uniqueItems true},
;              :address {:type "object",
;                        :properties {:street {:type "string"},
;                                     :city {:type "string"},
;                                     :zip {:type "integer", :format "int64"},
;                                     :lonlat {:type "array",
;                                              :items [{:type "number"} {:type "number"}],
;                                              :additionalItems false}},
;                        :required [:street :city :zip :lonlat]}},
; :required [:id :tags :address]}

Custom transformation via :json-schema namespaced properties:

(json-schema/transform
  [:enum
   {:title "Fish"
    :description "It's a fish"
    :json-schema/type "string"
    :json-schema/default "perch"}
   "perch" "pike"])
;{:title "Fish"
; :description "It's a fish"
; :type "string"
; :default "perch"
; :enum ["perch" "pike"]}

Full override with :json-schema property:

(json-schema/transform 
  [:map {:json-schema {:type "file"}} 
   [:file any?]])
; {:type "file"}

Swagger2

Transforming Schemas into Swagger2 Schema:

(require '[malli.swagger :as swagger])

(swagger/transform Address)
;{:type "object",
; :properties {:id {:type "string"},
;              :tags {:type "array"
;                     :items {:type "string"}
;                     :uniqueItems true},
;              :address {:type "object",
;                        :properties {:street {:type "string"},
;                                     :city {:type "string"},
;                                     :zip {:type "integer", :format "int64"},
;                                     :lonlat {:type "array",
;                                              :items {},
;                                              :x-items [{:type "number", :format "double"}
;                                                        {:type "number", :format "double"}]}},
;                        :required [:street :city :zip :lonlat]}},
; :required [:id :tags :address]}

Custom transformation via :swagger and :json-schema namespaced properties:

(swagger/transform
  [:enum
   {:title "Fish"
    :description "It's a fish"
    :swagger/type "string"
    :json-schema/default "perch"}
   "perch" "pike"])
;{:title "Fish"
; :description "It's a fish"
; :type "string"
; :default "perch"
; :enum ["perch" "pike"]}

Full override with :swagger property:

(swagger/transform 
  [:map {:swagger {:type "file"}} 
   [:file any?]])
; {:type "file"}

Custom Schema Types

Schema Types are described using m/IntoSchema protocol, which has a factory method (-into-schema [this properties children options]) to create the actual Schema instances. See malli.core for example implementations.

Simple Schema

For simple cases, there is m/-simple-schema:

(require '[clojure.test.check.generators :as gen])

(def Over6
  (m/-simple-schema
    {:type :user/over6
     :pred #(and (int? %) (> % 6))
     :type-properties {:error/message "should be over 6"
                       :decode/string mt/-string->long
                       :json-schema/type "integer"
                       :json-schema/format "int64"
                       :json-schema/minimum 6
                       :gen/gen (gen/large-integer* {:min 7})}}))

(m/into-schema? Over6)
; => true

m/IntoSchema can be both used as Schema (creating a Schema instance with nil properties and children) and as Schema type to create new Schema instances without needing to register the types:

(m/schema? (m/schema Over6))
; => true

(m/schema? (m/schema [Over6 {:title "over 6"}]))
; => true

:pred is used for validation:

(m/validate Over6 2)
; => false

(m/validate Over6 7)
; => true

:type-properties are shared for all schema instances and are used just like Schema (instance) properties by many Schema applications, including error messages, value generation and json-schema transformations.

(json-schema/transform Over6)
; => {:type "integer", :format "int64", :minimum 6}

(json-schema/transform [Over6 {:json-schema/example 42}])
; => {:type "integer", :format "int64", :minimum 6, :example 42}

Content Dependent Simple Schema

You can also build content-dependent schemas by using a callback function of properties children -> opts instead of static opts:

(def Over
  (m/-simple-schema
    (fn [{:keys [value]} _]
      (assert (int? value))
      {:type :user/over
       :pred #(and (int? %) (> % value))
       :type-properties {:error/fn (fn [error _] (str "should be over " value ", was " (:value error)))
                         :decode/string mt/-string->long
                         :json-schema/type "integer"
                         :json-schema/format "int64"
                         :json-schema/minimum value
                         :gen/gen (gen/large-integer* {:min (inc value)})}})))

(-> [Over {:value 12}]
    (m/explain 10)
    (me/humanize))
; => ["should be over 12, was 10"]

Schema Registry

Schemas are looked up using a malli.registry/Registry protocol, which is effectively a map from schema type to a schema recipe (Schema, IntoSchema or vector-syntax schema). Maps can also be used as a registry.

Custom Registry can be passed into all/most malli public APIs via the optional options map using :registry key. If omitted, malli.core/default-registry is used.

;; the default registry
(m/validate [:maybe string?] "kikka")
; => true

;; registry as explicit options
(m/validate [:maybe string?] "kikka" {:registry m/default-registry})
; => true

The default immutable registry is merged from multiple parts, enabling easy re-composition of custom schema sets. See built-in schemas for list of all Schemas.

Custom registry

Here's an example to create a custom registry without the default core predicates and with :neg-int and :pos-int Schemas:

(def registry
  (merge
    (m/class-schemas)
    (m/comparator-schemas)
    (m/base-schemas)
    {:neg-int (m/-simple-schema {:type :neg-int, :pred neg-int?})
     :pos-int (m/-simple-schema {:type :pos-int, :pred pos-int?})}))

(m/validate [:or :pos-int :neg-int] 'kikka {:registry registry})
; => false

(m/validate [:or :pos-int :neg-int] 123 {:registry registry})
; => true

We did not register normal predicate schemas:

(m/validate pos-int? 123 {:registry registry})
; Syntax error (ExceptionInfo) compiling
; :malli.core/invalid-schema {:schema pos-int?}

Local registry

Any schema can define a local registry using :registry schema property:

(def Adult
  [:map {:registry {::age [:and int? [:> 18]]}}
   [:age ::age]])

(mg/generate Adult {:size 10, :seed 1})
; => {:age 92}

Local registries can be persisted:

(-> Adult
    (malli.edn/write-string)
    (malli.edn/read-string)
    (m/validate {:age 46}))
; => true

Changing the default registry

Passing in custom options to all public methods is a lot of boilerplate. For the lazy, there is an easier way - we can swap the (global) default registry:

(require '[malli.registy :as mr])

;; the default registry
(-> m/default-registry (mr/schemas) (count)) 
;=> 140

;; global side-effects! free since 0.7.0!
(mr/set-default-registry!
  {:string (m/-string-schema)
   :maybe (m/-maybe-schema)
   :map (m/-map-schema)})
   
(-> m/default-registry (mr/schemas) (count))
; => 3
   
(m/validate
  [:map [:maybe [:maybe :string]]]
  {:maybe "sheep"})
; => true

(m/validate :int 42)
; =throws=> :malli.core/invalid-schema {:schema :int}

NOTE: mr/set-default-registry! is an imperative api with global side-effects. Easy, but not simple. If you want to disable the api, you can define the following compiler/jvm bootstrap:

cljs: :closure-defines {malli.registry/mode "strict"}
clj: :jvm-opts ["-Dmalli.registry/mode=strict"]

DCE and Schemas

The default schema registry is defined as a Var, so all Schema implementation (100+) are dragged in. For ClojureScript, this means the schemas implementations are not removed via Dead Code Elimination (DCE), resulting a large (37KB, zipped) js-bundle.

Malli allows the default registry to initialized with empty schemas, using the following compiler/jvm bootstrap:

cljs: :closure-defines {malli.registry/type "custom"}
clj: :jvm-opts ["-Dmalli.registry/type=custom"]

;; with the flag set on
(-> m/default-registry (mr/schemas) (count))
; => 0

With this, you can register just what you need and rest are DCE'd. The previous example results in just a 3KB gzip bundle.

Registry implemenations

Malli supports multiple type of registries.

Immutable registry

Just just a Map.

(require '[malli.registry :as mr])

(mr/set-default-registry!
  {:string (m/-string-schema)
   :maybe (m/-maybe-schema)
   :map (m/-map-schema)})

(m/validate
  [:map [:maybe [:maybe :string]]]
  {:maybe "sheep"})
; => true

Mutable registry

clojure.spec introduces a mutable global registry for specs. The mutable registry in malli forced you to bring in your own state atom and functions how to work with it:

Using a custom registry atom:

(def registry*
  (atom {:string (m/-string-schema)
         :maybe (m/-maybe-schema)
         :map (m/-map-schema)}))

(defn register! [type ?schema]
  (swap! registry* assoc type ?schema))

(mr/set-default-registry!
  (mr/mutable-registry registry*))

(register! :non-empty-string [:string {:min 1}])

(m/validate :non-empty-string "malli")
; => true

The mutable registry can also be passed in as an explicit option:

(def registry (mr/mutable-registry registry*))

(m/validate :non-empty-string "malli" {:registry registry})
; => true

Dynamic Registry

If you know what you are doing, you can also use dynamic scope to pass in default schema registry:

(mr/set-default-registry!
  (mr/dynamic-registry))

(binding [mr/*registry* {:string (m/-string-schema)
                         :maybe (m/-maybe-schema)
                         :map (m/-map-schema)
                         :non-empty-string [:string {:min 1}]}]
  (m/validate :non-empty-string "malli"))
; => true

Lazy Registries

You can provide schemas at runtime using mr/lazy-registry - it takes a local registry and a provider function of type registry -> schema as arguments:

(def registry
  (mr/lazy-registry
    (m/default-schemas)
    (fn [type registry]
      ;; simulates pulling CloudFormation Schemas when needed
      (let [lookup {"AWS::ApiGateway::UsagePlan" [:map {:closed true}
                                                  [:Type [:= "AWS::ApiGateway::UsagePlan"]]
                                                  [:Description {:optional true} string?]
                                                  [:UsagePlanName {:optional true} string?]]
                    "AWS::AppSync::ApiKey" [:map {:closed true}
                                            [:Type [:= "AWS::AppSync::ApiKey"]]
                                            [:ApiId string?]
                                            [:Description {:optional true} string?]]}]
        (println "... loaded" type)
        (some-> type lookup (m/schema {:registry registry}))))))

;; lazy multi, doesn't realize the schemas
(def CloudFormation
  (m/schema
    [:multi {:dispatch :Type, :lazy-refs true}
     "AWS::ApiGateway::UsagePlan"
     "AWS::AppSync::ApiKey"]
    {:registry registry}))

(m/validate
  CloudFormation
  {:Type "AWS::ApiGateway::UsagePlan"
   :Description "laiskanlinna"})
; ... loaded AWS::ApiGateway::UsagePlan
; => true

(m/validate
  CloudFormation
  {:Type "AWS::ApiGateway::UsagePlan"
   :Description "laiskanlinna"})
; => true

Composite Registry

Registries can be composed, a full example:

(require '[malli.core :as m])
(require '[malli.registry :as mr])

(def registry (atom {}))

(defn register! [type schema]
  (swap! registry assoc type schema))

(mr/set-default-registry!
  ;; linear search
  (mr/composite-registry
    ;; immutable registry
    {:map (m/-map-schema)}
    ;; mutable (spec-like) registry
    (mr/mutable-registry registry)
    ;; on the perils of dynamic scope
    (mr/dynamic-registry)))

;; mutate like a boss
(register! :maybe (m/-maybe-schema))

;; ☆.。.:*・°☆.。.:*・°☆.。.:*・°☆.。.:*・°☆
(binding [mr/*registry* {:string (m/-string-schema)}]
  (m/validate
    [:map [:maybe [:maybe :string]]]
    {:maybe "sheep"}))
; => true

Function Schemas

See Working with functions.

Instrumentation

See Instrumentation.

Clj-kondo

Clj-kondo is a linter for Clojure code that sparks joy.

Given functions and function Schemas:

(defn square [x] (* x x))
(m/=> square [:=> [:cat int?] nat-int?])

(defn plus
  ([x] x)
  ([x y] (+ x y)))

(m/=> plus [:function
            [:=> [:cat int?] int?]
            [:=> [:cat int? int?] int?]])

Generating clj-kondo configuration from current namespace:

(require '[malli.clj-kondo :as mc])

(-> (mc/collect *ns*) (mc/linter-config))
;{:lint-as #:malli.schema{defn schema.core/defn},
; :linters
; {:type-mismatch
;  {:namespaces
;   {user {square {:arities {1 {:args [:int]
;                               :ret :pos-int}}}
;          plus {:arities {1 {:args [:int]
;                             :ret :int},
;                          2 {:args [:int :int]
;                             :ret :int}}}}}}}}

Emitting confing into ./.clj-kondo/configs/malli/config.edn:

(mc/emit!)

To enable the integration, you need to have the following config:

✗ cat .clj-kondo/config.edn
{:config-paths ["configs/malli"]}

In action:

Visualizing Schemas

DOT

Transforming Schemas into [DOT Language](https://en.wikipedia.org/wiki/DOT_(graph_description_language):

(require '[malli.dot :as md])

(def Address
  [:schema
   {:registry {"Country" [:map
                          [:name [:enum :FI :PO]]
                          [:neighbors [:vector [:ref "Country"]]]]
               "Burger" [:map
                         [:name string?]
                         [:description {:optional true} string?]
                         [:origin [:maybe "Country"]]
                         [:price pos-int?]]
               "OrderLine" [:map
                            [:burger "Burger"]
                            [:amount int?]]
               "Order" [:map
                        [:lines [:vector "OrderLine"]]
                        [:delivery [:map
                                    [:delivered boolean?]
                                    [:address [:map
                                               [:street string?]
                                               [:zip int?]
                                               [:country "Country"]]]]]]}}
   "Order"])

(md/transform Address)
; "digraph { ... }"

Visualized with Graphviz:

PlantUML

Transforming Schemas into PlantUML:

(require '[malli.plantuml :as plantuml])

(plantuml/transform Address)
; "@startuml ... @enduml"

Visualized with PlantText:

Performance

Malli tries to be really, really fast.

Validation Perfomance

Usually as fast (or faster) as idiomatic Clojure.

(require '[criterium.core :as cc])

(def valid {:x true, :y 1, :z "zorro"})

;; idomatic clojure (54ns)
(let [valid? (fn [{:keys [x y z]}]
               (and (boolean? x)
                    (if y (int? y) true)
                    (string? z)))]
  (assert (valid? valid))
  (cc/quick-bench (valid? valid)))

(require '[malli.core :as m])

;; malli (39ns)
(let [valid? (m/validator
               [:map
                [:x :boolean]
                [:y {:optional true} :int]
                [:z :string]])]
  (assert (valid? valid))
  (cc/quick-bench (valid? valid)))

Same with Clojure Spec and Plumatic Schema:

(require '[clojure.spec.alpha :as spec])
(require '[schema.core :as schema])

(spec/def ::x boolean?)
(spec/def ::y int?)
(spec/def ::z string?)

;; clojure.spec (450ns)
(let [spec (spec/keys :req-un [::x ::z] :opt-un [::y])]
  (assert (spec/valid? spec valid))
  (cc/quick-bench (spec/valid? spec valid)))

;; plumatic schema (660ns)
(let [valid? (schema/checker
               {:x schema/Bool
                (schema/optional-key :y) schema/Int
                :z schema/Str})]
  (assert (not (valid? valid)))
  (cc/quick-bench (valid? valid)))

Transformation Performance

Usually faster than idiomatic Clojure.

(def data {:x "true", :y "1", :z "kikka"})
(def expexted {:x true, :y 1, :z "kikka"})

;; idiomatic clojure (290ns)
(let [transform (fn [{:keys [x y] :as m}]
                  (cond-> m
                    (string? x) (update :x #(Boolean/parseBoolean %))
                    (string? y) (update :y #(Long/parseLong %))))]
  (assert (= expexted (transform data)))
  (cc/quick-bench (transform data)))

;; malli (72ns)
(let [schema [:map
              [:x :boolean]
              [:y {:optional true} int?]
              [:z string?]]
      transform (m/decoder schema (mt/string-transformer))]
  (assert (= expexted (transform data)))
  (cc/quick-bench (transform data)))

Same with Clojure Spec and Plumatic Schema:

(require '[spec-tools.core :as st])
(require '[schema.coerce :as sc])

(spec/def ::x boolean?)
(spec/def ::y int?)
(spec/def ::z string?)

;; clojure.spec (19µs)
(let [spec (spec/keys :req-un [::x ::z] :opt-un [::y])
      transform #(st/coerce spec % st/string-transformer)]
  (assert (= expexted (transform data)))
  (cc/quick-bench (transform data)))

;; plumatic schema (2.2µs)
(let [schema {:x schema/Bool
              (schema/optional-key :y) schema/Int
              :z schema/Str}
      transform (sc/coercer schema sc/string-coercion-matcher)]
  (assert (= expexted (transform data)))
  (cc/quick-bench (transform data)))

The transformation engine is smart enough to just transform parts of the schema that need to be transformed. If there is nothing to transform, identity function is returned.

(def json->user
  (m/decoder
    [:map
     [:id :int]
     [:name :string]
     [:address [:map
                [:street :string]
                [:rural :boolean]
                [:country [:enum "finland" "poland"]]]]]
    (mt/json-transformer)))

(= identity json->user)
; => true

;; 5ns
(cc/quick-bench 
  (json->user 
    {:id 1
     :name "tiina"
     :address {:street "kotikatu"
               :rural true
               :country "poland"}}))

Parsing performance

;; 37µs
(let [spec (s/* (s/cat :prop string?,
                       :val (s/alt :s string?
                                   :b boolean?)))
      parse (partial s/conform spec)]
  (cc/quick-bench
    (parse ["-server" "foo" "-verbose" "-verbose" "-user" "joe"])))

;; 2.4µs
(let [schema [:* [:catn
                  [:prop string?]
                  [:val [:altn
                         [:s string?]
                         [:b boolean?]]]]]
      parse (m/parser schema)]
  (cc/quick-bench
    (parse ["-server" "foo" "-verbose" "-verbose" "-user" "joe"])))

Built-in Schemas

`malli.core/predicate-schemas`

Contains both function values and unqualified symbol representations for all relevant core predicates. Having both representations enables reading forms from both code (function values) and EDN-files (symbols): any?, some?, number?, integer?, int?, pos-int?, neg-int?, nat-int?, float?, double?, boolean?, string?, ident?, simple-ident?, qualified-ident?, keyword?, simple-keyword?, qualified-keyword?, symbol?, simple-symbol?, qualified-symbol?, uuid?, uri?, decimal?, inst?, seqable?, indexed?, map?, vector?, list?, seq?, char?, set?, nil?, false?, true?, zero?, rational?, coll?, empty?, associative?, sequential?, ratio?, bytes?, ifn? and fn?.

`malli.core/class-schemas`

Class-based schemas, contains java.util.regex.Pattern & js/RegExp.

`malli.core/comparator-schemas`

Comparator functions as keywords: :>, :>=, :<, :<=, := and :not=.

`malli.core/type-schemas`

Type-like schemas: :any, :nil, :string, :int, :double, :boolean, :keyword, :symbol, :qualified-symbol, :qualified-keyword and :uuid.

`malli.core/sequence-schemas`

Sequence/regex-schemas: :+, :*, :?, :repeat, :cat, :alt, :catn, :altn.

`malli.core/base-schemas`

Contains :and, :or, :orn, :not, :map, :map-of, :vector, :sequential, :set, :enum, :maybe, :tuple, :multi, :re, :fn, :ref, :=>, :function and :schema.

`malli.util/schemas`

:merge, :union and :select-keys.

Links (and thanks)

Schema https://github.com/plumatic/schema
Clojure.spec https://clojure.org/guides/spec
Spell-spec https://github.com/bhauman/spell-spec
JSON Schema https://json-schema.org/understanding-json-schema/index.html
Spec-provider: https://github.com/stathissideris/spec-provider
F# Type Providers: https://docs.microsoft.com/en-us/dotnet/fsharp/tutorials/type-providers/
Minimallist https://github.com/green-coder/minimallist
malli-instrument https://github.com/setzer22/malli-instrument
Core.typed https://github.com/clojure/core.typed
TypeScript https://www.typescriptlang.org/
Struct https://funcool.github.io/struct/latest/
Seqexp https://github.com/cgrand/seqexp
yup https://github.com/jquense/yup
JOI https://github.com/hapijs/joi

Alpha

The public API of Malli has been quite stable already in pre-alpha and in alpha, we try not to break things. Still, the library is evolving and things like value destructuring could affect public APIs and most likely affect the library extenders, e.g. need to implement a new protocol method for custom schemas.

All changes (breaking or not) will be documented in the CHANGELOG and there will be a migration guide and path if needed.

The API layers and stability:

public API: public vars, name doesn't start with -, e.g. malli.core/validate. The most stable part of the library, should not change (much) in alpha
extender API: public vars, name starts with -, e.g. malli.core/-collection-schema. Not needed with basic use cases, might evolve during the alpha, follow CHANGELOG for details
private API: private vars and malli.impl namespaces, all bets are off.

Supported Java versions

Malli aims to support the LTS releases Java 8 and 11 and the latest Java release, Java 15.

Running tests

We use Kaocha and cljs-test-runner as a test runners. Before running the tests, you need to install NPM dependencies.

npm install
./bin/kaocha
./bin/node

Installing locally

clj -Mjar
clj -Minstall

Bundle size for cljs

With default registry (37KB+ Gzipped)

# no sci
npx shadow-cljs run shadow.cljs.build-report app /tmp/report.html

# with sci
npx shadow-cljs run shadow.cljs.build-report app-sci /tmp/report.html

With minimal registry (2.4KB+ Gzipped)

# no sci
npx shadow-cljs run shadow.cljs.build-report app2 /tmp/report.html

# with sci
npx shadow-cljs run shadow.cljs.build-report app2-sci /tmp/report.html

Checking the generated code

npx shadow-cljs release app --pseudo-names

Testing on GraalVM

Without sci (11Mb)

./bin/native-image demo
./demo '[:set :keyword]' '["kikka" "kukka"]'

With sci (18Mb):

./bin/native-image demosci
./demosci '[:fn (fn [x] (and (int? x) (> x 10)))]]' '12'

License

Available under the terms of the Eclipse Public License 2.0, see LICENSE.

Name		Name	Last commit message	Last commit date
Latest commit History 1,547 Commits
.github/workflows		.github/workflows
app/malli		app/malli
bin		bin
docs		docs
graal-test/src/malli/graalvm		graal-test/src/malli/graalvm
perf/malli/perf		perf/malli/perf
src/malli		src/malli
test/malli		test/malli
.gitignore		.gitignore
CHANGELOG.md		CHANGELOG.md
CONTRIBUTING.md		CONTRIBUTING.md
LICENSE		LICENSE
README.md		README.md
build.clj		build.clj
deps.edn		deps.edn
jmh.edn		jmh.edn
package-lock.json		package-lock.json
package.json		package.json
pom.xml		pom.xml
shadow-cljs.edn		shadow-cljs.edn
tests.edn		tests.edn

License

evg-tso/malli

Folders and files

Latest commit

History

Repository files navigation

malli

Motivation

The library

Syntax

Vector Syntax

Map Syntax

Why Two Syntaxes?

Validation

Qualified keys in a map

Homogeneous Maps

Sequence Schemas

String schemas

Maybe schemas

Fn schemas

Error Messages

Custom Error Messages

Spell Checking

Value Transformation

Default values

Programming with Schemas

Declarative Schema Transformation

Persisting Schemas

Multi Schemas

Recursive Schemas

Value Generation

Inferring Schemas

Parsing values

Unparsing values

Serializable Functions

Schema AST

Schema Transformation

JSON Schema

Swagger2

Custom Schema Types

Simple Schema

Content Dependent Simple Schema

Schema Registry

Custom registry

Local registry

Changing the default registry

DCE and Schemas

Registry implemenations

Immutable registry

Mutable registry

Dynamic Registry

Lazy Registries

Composite Registry

Function Schemas

Instrumentation

Clj-kondo

Visualizing Schemas

DOT

PlantUML

Performance

Validation Perfomance

Transformation Performance

Parsing performance

Built-in Schemas

malli.core/predicate-schemas

malli.core/class-schemas

malli.core/comparator-schemas

malli.core/type-schemas

malli.core/sequence-schemas

malli.core/base-schemas

malli.util/schemas

Links (and thanks)

Alpha

Supported Java versions

Running tests

Installing locally

Bundle size for cljs

Checking the generated code

Testing on GraalVM

License

`malli.core/predicate-schemas`

`malli.core/class-schemas`

`malli.core/comparator-schemas`

`malli.core/type-schemas`

`malli.core/sequence-schemas`

`malli.core/base-schemas`

`malli.util/schemas`

Packages