malli

Data-driven Schemas for Clojure/Script and babashka.

Metosin Open Source Status: Active. Stability: well matured alpha.

• Schema definitions as data
• Vector, Map and Lite syntaxes
• Validation and Value Transformation
• First class Error Messages with Spell Checking
• Generating values from Schemas
• Inferring Schemas from sample values and Destructuring.
• Tools for Programming with Schemas
• Parsing and Unparsing values
• Enumeration, Sequence, Vector, and Set Schemas
• Persisting schemas, even function schemas
• Immutable, Mutable, Dynamic, Lazy and Local Schema Registries
• Schema Transformations to JSON Schema, Swagger2, and descriptions in english
• Multi-schemas, Recursive Schemas and Default values
• Function Schemas with dynamic and static schema checking
  • Integrates with both clj-kondo and Typed Clojure
• Visualizing Schemas with DOT and PlantUML
• Pretty development time errors
• Fast

Presentations:

• Transforming Data With Malli and Meander
• High-Performance Schemas in Clojure/Script with Malli 1/2
• ClojureStream 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, see also some 3rd Party Libraries.

Want to contribute? See the Development guide.

Hi! We are Metosin, a consulting company. These libraries have evolved out of the work we do for our clients. We maintain & develop this project, for you, for free. Issues and pull requests welcome! However, if you want more help using the libraries, or want us to build something as cool for you, consider our commercial support.

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 no 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

Malli requires Clojure 1.11.

Malli is tested with the LTS releases Java 8, 11, 17 and 21.

Quickstart

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

(def UserId :string)

(def Address
  [:map
   [:street :string]
   [:country [:enum "FI" "UA"]]])

(def User
  [:map
   [:id #'UserId]
   [:address #'Address]
   [:friends [:set {:gen/max 2} [:ref #'User]]]])

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

(mg/generate User)
;{:id "AC",
; :address {:street "mf", :country "UA"},
; :friends #{{:id "1dm",
;             :address {:street "8", :country "UA"},
;             :friends #{}}}}

(m/validate User *1)
; => true

Syntax

Malli supports Vector, Map and Lite 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]
[:-> :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

Alternative map-syntax, similar to cljfx:

NOTE: For now, Map syntax in considered as internal, so don't use it as a database persistency model.

;; 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}
{:type :->
 :children [{:type :int} {:type :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 multiple syntaxes?

Malli started with just the Vector syntax. It's really powerful and relatively easy to read, but not optimal for all use cases.

We introduced Map Syntax as we 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. Map-syntax allows lazy and parseless Schema Creation.

We added Lite Syntax for simplified schema creation for special cases, like to be used with reitit coercion and for easy migration from data-specs.

Example Address schema

Following example schema is assumed in many of the following examples.

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

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 [:= 1] 1)
; => true

(m/validate [:enum 1 2] 1)
; => true

(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.

Enumeration schemas

:enum schemas [:enum V1 V2 ...] represent an enumerated set of values V1 V2 ....

This mostly works as you'd expect, with values passing the schema if it is contained in the set and generators returning one of the values, shrinking to the left-most value.

There are some special cases to keep in mind around syntax. Since schema properties can be specified with a map or nil, enumerations starting with a map or nil must use slightly different syntax.

If your :enum does not have properties, you must provide nil as the properties.

[:enum nil {}]  ;; singleton schema of {}
[:enum nil nil] ;; singleton schema of nil

If your :enum has properties, the leading map with be interpreted as properties, not an enumerated value.

[:enum {:foo :bar} {}]  ;; singleton schema of {}, with properties {:foo :bar}
[:enum {:foo :bar} nil] ;; singleton schema of nil, with properties {:foo :bar}

In fact, these syntax rules apply to all schemas, but :enum is the most common schema where this is relevant so it deserves a special mention.

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

Map with default schemas

Map schemas can define a special :malli.core/default key to handle extra keys:

(m/validate
 [:map
  [:x :int]
  [:y :int]
  [::m/default [:map-of :int :int]]]
 {:x 1, :y 2, 1 1, 2 2})
; => true

default branching can be arbitrarily nested:

(m/validate
 [:map
  [:x :int]
  [::m/default [:map
                [:y :int]
                [::m/default [:map-of :int :int]]]]]
 {:x 1, :y 2, 1 1, 2 2})
; => true

Seqable schemas

The :seqable and :every schemas describe seqable? collections. They differ in their handling of collections that are neither counted? nor indexed?, and their parsers:

1. :seqable parses its elements but :every does not and returns the identical input, and
2. valid unparsed :seqable values lose the original collection type while :every returns the identical input.

:seqable validates the entire collection, while :every checks only the largest of :min, (inc :max), and (::m/coll-check-limit options 101), or the entire collection if the input is counted? or indexed?.

;; :seqable and :every validate identically with small, counted, or indexed collections.
(m/validate [:seqable :int] #{1 2 3})
;=> true
(m/validate [:seqable :int] [1 2 3])
;=> true
(m/validate [:seqable :int] (sorted-set 1 2 3))
;=> true
(m/validate [:seqable :int] (range 1000))
;=> true
(m/validate [:seqable :int] (conj (vec (range 1000)) nil))
;=> false

(m/validate [:every :int] #{1 2 3})
;=> true
(m/validate [:every :int] [1 2 3])
;=> true
(m/validate [:every :int] (sorted-set 1 2 3))
;=> true
(m/validate [:every :int] (vec (range 1000)))
;=> true
(m/validate [:every :int] (conj (vec (range 1000)) nil))
;=> false

;; for large uncounted and unindexed collections, :every only checks a certain length
(m/validate [:seqable :int] (concat (range 1000) [nil]))
;=> false
(m/validate [:every :int] (concat (range 1000) [nil]))
;=> true

Sequence schemas

You can use :sequential to describe homogeneous sequential Clojure collections.

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

(m/validate [:sequential int?] [42 105])
;; => true

(m/validate [:sequential int?] #{42 105})
;; => false

Malli also supports sequence regexes (also called sequence expressions) 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 ({:path [0 :val :s], :in [3], :schema string?, :value 11}
;;              {: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 ({:path [0 1 0], :in [3], :schema string?, :value 11}
;;              {:path [0 1 1], :in [3], :schema boolean?, :value 11})}

As all these examples show, the sequence expression (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

Vector schemas

You can use :vector to describe homogeneous Clojure vectors.

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

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

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

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

To create a vector schema based on a seqex, use :and.

;; non-empty vector starting with a keyword
(m/validate [:and [:cat :keyword [:* :any]]
                  vector?]
            [:a 1])
; => true

(m/validate [:and [:cat :keyword [:* :any]]
                  vector?]
            (:a 1))
; => false

Note: To generate values from a vector seqex, see :and generation.

Set schemas

You can use :set to describe homogeneous Clojure sets.

(m/validate [:set int?] #{42 105})
;; => true

(m/validate [:set int?] #{:a :b})
;; => false

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:

(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 ({:path [:tags 0]
;           :in [:tags 0]
;           :schema keyword?
;           :value "coffee"}
;          {:path [:address :city],
;           :in [:address :city],
;           :schema [:map
;                    [:street string?]
;                    [:city string?]
;                    [:zip int?]
;                    [:lonlat [:tuple double? double?]]],
;           :type :malli.core/missing-key}
;          {:path [:address :lonlat 1]
;           :in [:address :lonlat 1]
;           :schema double?
;           :value nil})}

Under :errors, you get a list of errors with the following keys:

• :path, error location in Schema
• :in, error location in value
• :schema, schema in error
• :value, value in error

(def Schema [:map [:x [:maybe [:tuple :string]]]])

(def value {:x [1]})

(def error (-> Schema
               (m/explain value)
               :errors
               first))

error
;{:path [:x 0 0]
; :in [:x 0]
; :schema :string
; :value 1}

(get-in value (:in error))
; => 1

(mu/get-in Schema (:path error))
; => :string

Note! If you need error messages that serialize neatly to EDN/JSON, use malli.util/explain-data instead.

Humanized 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"]]}}

Or if you already have a malli validation exception (e.g. in a catch form):

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

(try
  (m/validate Address {:not "an address"})
  (catch Exception e
    (-> e ex-data :data :explain me/humanize)))

Custom error messages

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"]}

Values in error

Just to get parts of the value that are in error:

(-> Address
    (m/explain
     {:id "Lillan"
      :tags #{:artesan "coffee" :garden "ground"}
      :address {:street "Ahlmanintie 29"
                :zip 33100
                :lonlat [61.4858322, "23.7832851,17"]}})
    (me/error-value))
;{:tags #{"coffee" "ground"}
; :address {:lonlat [nil "23.7832851,17"]}}

Masking irrelevant parts:

(-> Address
    (m/explain
     {:id "Lillan"
      :tags #{:artesan "coffee" :garden "ground"}
      :address {:street "Ahlmanintie 29"
                :zip 33100
                :lonlat [61.4858322, "23.7832851,17"]}})
    (me/error-value {::me/mask-valid-values '...}))
;{:id ...
; :tags #{"coffee" "ground" ...}
; :address {:street ...
;           :zip ...
;           :lonlat [... "23.7832851,17"]}}

Pretty errors

There are two ways to get pretty errors:

Development mode

Start development mode:

((requiring-resolve 'malli.dev/start!))

Now, any exception thrown via malli.core/-fail! is being captured and pretty printed before being thrown. Pretty printing is extendable using virhe.

Pretty Coercion:

Custom exception (with default layout):

Pretty printing in being backed by malli.dev.virhe/-format multimethod using (-> exception (ex-data) :data) as the default dispatch key. As fallback, exception class - or exception subclass can be used, e.g. the following will handle all java.sql.SQLException and it's parent exceptions:

(require '[malli.dev.virhe :as v])

(defmethod v/-format java.sql.SQLException [e _ printer]
  {:title "Exception thrown"
   :body [:group
          (v/-block "SQL Exception" (v/-color :string (ex-message e) printer) printer) :break :break
          (v/-block "More information:" (v/-link "https://cljdoc.org/d/metosin/malli/CURRENT" printer) printer)]})

pretty/explain

For pretty development-time error printing, try malli.dev.pretty/explain

Value transformation

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

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

Default Transformers include:

name	description
mt/string-transformer	transform between strings and EDN
mt/json-transformer	transform between JSON and EDN
mt/strip-extra-keys-transformer	drop extra keys from maps
mt/default-value-transformer	applies default values from schema properties
mt/key-transformer	transforms map keys
mt/collection-transformer	conversion between collections (e.g. set -> vector)

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.

Simple usage:

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

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

For performance, precompute the transformations with m/decoder and m/encoder:

(def decode (m/decoder int? mt/string-transformer))

(decode "42")
; 42

(def encode (m/encoder int? mt/string-transformer))

(encode 42)
; "42"

Coercion

For both decoding + validating the results (throwing exception on error), there is m/coerce and m/coercer:

(m/coerce :int "42" mt/string-transformer)
; 42

((m/coercer :int mt/string-transformer) "42")
; 42

(m/coerce :int "invalid" mt/string-transformer)
; =throws=> :malli.core/invalid-input {:value "invalid", :schema :int, :explain {:schema :int, :value "invalid", :errors ({:path [], :in [], :schema :int, :value "invalid"})}}

Coercion can be applied without transformer, doing just validation:

(m/coerce :int 42)
; 42

(m/coerce :int "42")
; =throws=> :malli.core/invalid-input {:value "42", :schema :int, :explain {:schema :int, :value "42", :errors ({:path [], :in [], :schema :int, :value "42"})}}

Exception-free coercion with continuation-passing style:

(m/coerce :int "fail" nil (partial prn "success:") (partial prn "error:"))
; =prints=> "error:" {:value "fail", :schema :int, :explain ...}

Advanced Transformations

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]}}

Transforming homogenous :enum or :=s (supports automatic type detection of :keyword, :symbol, :int and :double):

(m/decode [:enum :kikka :kukka] "kukka" mt/string-transformer)
; => :kukka

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 clojure.string/upper-case}]
  "kerran" mt/string-transformer)
; => "KERRAN"

This works too:

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

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

(m/decode
  [string? {:decode/string {:enter clojure.string/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}

:and accumulates the transformed value left-to-right.

(m/decode
  [:and
   [:string {:decode/string '{:enter #(str "1_" %), :leave #(str % "_2")}}]
   [:string {:decode/string '{:enter #(str "3_" %), :leave #(str % "_4")}}]]
  "kerran" mt/string-transformer)
;; => "3_1_kerran_2_4"

:or transforms using the first successful schema, left-to-right.

(m/decode
  [:or
   [:string {:decode/string '{:enter #(str "1_" %), :leave #(str % "_2")}}]
   [:string {:decode/string '{:enter #(str "3_" %), :leave #(str % "_4")}}]]
  "kerran" mt/string-transformer)
;; => "1_kerran_2"

(m/decode
  [:or
   :map
   [:string {:decode/string '{:enter #(str "3_" %), :leave #(str % "_4")}}]]
  "kerran" mt/string-transformer)
;; => "3_kerran_4"

Proxy schemas like :merge and :union transform as if m/derefed.

(m/decode
  [:merge
   [:map [:name [:string {:default "kikka"}]] ]
   [:map [:description {:optional true} [:string {:default "kikka"}]]]]
  {}
  {:registry (merge (mu/schemas) (m/default-schemas))}
  (mt/default-value-transformer {::mt/add-optional-keys true}))
;; => {:name "kikka"
;;     :description "kikka"}

To and from JSON

The m/encode and m/decode functions work on clojure data. To go from clojure data to JSON, you need a JSON library like jsonista. Additionally, since m/decode doesn't check the schema, you need to run m/validate (or m/explain) if you want to make sure your data conforms to your schema.

To JSON:

(def Tags
  (m/schema [:map
             {:closed true}
             [:tags [:set :keyword]]]))
(jsonista.core/write-value-as-string
 (m/encode Tags
           {:tags #{:bar :quux}}
           mt/json-transformer))
; => "{\"tags\":[\"bar\",\"quux\"]}"

From JSON without validation:

(m/decode Tags
          (jsonista.core/read-value "{\"tags\":[\"bar\",[\"quux\"]]}"
                                    jsonista.core/keyword-keys-object-mapper)
          mt/json-transformer)
; => {:tags #{:bar ["quux"]}}

From JSON with validation:

(m/explain Tags
           (m/decode Tags
                     (jsonista.core/read-value "{\"tags\":[\"bar\",[\"quux\"]]}"
                                               jsonista.core/keyword-keys-object-mapper)
                     mt/json-transformer))
; => {:schema [:map {:closed true} [:tags [:set :keyword]]],
;     :value {:tags #{:bar ["quux"]}},
;     :errors ({:path [:tags 0], :in [:tags ["quux"]], :schema :keyword, :value ["quux"]})}

(m/validate Tags
            (m/decode Tags
                      (jsonista.core/read-value "{\"tags\":[\"bar\",\"quux\"]}" ; <- note! no error
                                                jsonista.core/keyword-keys-object-mapper)
                      mt/json-transformer))
; => true

For performance, it's best to prebuild the validator, decoder and explainer:

(def validate-Tags (m/validator Tags))
(def decode-Tags (m/decoder Tags mt/json-transformer))
(-> (jsonista.core/read-value "{\"tags\":[\"bar\",\"quux\"]}"
                              jsonista.core/keyword-keys-object-mapper)
    decode-Tags
    validate-Tags)
; => true

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 "-"}}

With custom function:

(m/decode
 [:map
  [:os [:string {:property "os.name"}]]
  [:timezone [:string {:property "user.timezone"}]]]
 {}
 (mt/default-value-transformer
  {:key :property
   :default-fn (fn [_ x] (System/getProperty x))}))
; => {:os "Mac OS X", :timezone "Europe/Helsinki"}

Optional Keys are not added by default:

(m/decode
 [:map
  [:name [:string {:default "kikka"}]]
  [:description {:optional true} [:string {:default "kikka"}]]]
 {}
 (mt/default-value-transformer))
; => {:name "kikka"}

Adding optional keys too via ::mt/add-optional-keys option:

(m/decode
 [:map
  [:name [:string {:default "kikka"}]]
  [:description {:optional true} [:string {:default "kikka"}]]]
 {}
 (mt/default-value-transformer {::mt/add-optional-keys true}))
; => {:name "kikka", :description "kikka"}

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

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

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

:union is similar to :or, except :union combines map schemas in different disjuncts with :or. For example, UnionMaps is equivalent to [:map [:x [:or :int :string]] [:y [:or :int :string]]].

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

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

(m/validate OrMaps {:x "kikka" :y "kikka"})
; => false

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

:merge and :union differ on schemas with common keys. :merge chooses the right-most schema of common keys, and :union combines them with :or. For example, MergedCommon is equivalent to [:map [:x :int]], and UnionCommon is equivalent to [:map [:x [:or :string :int]]].

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

(def UnionCommon
  (m/schema
    [:union
     [:map [:x :string]]
     [:map [:x :int]]]
    {:registry registry}))

(m/validate MergedCommon {:x "kikka"})
; => true
(m/validate MergedCommon {:x 1})
; => false
(m/validate UnionCommon {:x "kikka"})
; => true
(m/validate UnionCommon {:x 1})
; => true

Distributive schemas

:merge also distributes over :multi in a similar way to how multiplication distributes over addition in arithmetic. There are two transformation rules, applied in the following order:

;; right-distributive
[:merge [:multi M1 M2 ...] M3]
=>
[:multi [:merge M1 M3] [:merge M2 M3] ...]

;; left-distributive
[:merge M1 [:multi M2 M3 ...]]
=>
[:multi [:merge M1 M2] [:merge M1 M3] ...]

For :merge with more than two arguments, the rules are applied iteratively left-to-right as if the following transformation was applied:

[:merge M1 M2 M3 M4 ...]
=>
[:merge
 [:merge
  [:merge M1 M2]
  M3]
 M4]
...

The distributive property of :multi is useful combined with :merge if you want all clauses of a :multi to share extra entries.

Here are concrete examples of applying the rules:

;; left-distributive
(m/deref
 [:merge
  [:map [:x :int]]
  [:multi {:dispatch :y}
   [1 [:map [:y [:= 1]]]]
   [2 [:map [:y [:= 2]]]]]]
 {:registry registry})
; => [:multi {:dispatch :y}
;     [1 [:map [:x :int] [:y [:= 1]]]]
;     [2 [:map [:x :int] [:y [:= 2]]]]]

;; right-distributive
(m/deref
 [:merge
  [:multi {:dispatch :y}
   [1 [:map [:y [:= 1]]]]
   [2 [:map [:y [:= 2]]]]]
  [:map [:x :int]]]
 {:registry registry})
; => [:multi {:dispatch :y}
;     [1 [:map [:y [:= 1]] [:x :int]]]
;     [2 [:map [:y [:= 2]] [:x :int]]]]

It is not recommended to use local registries in schemas that are transformed. Also be aware that merging non-maps via the distributive property inherits the same semantics as :merge, which is based on meta-merge.

Persisting schemas

Writing and Reading schemas as EDN, no eval needed.

Following example requires SCI or cherry as external dependency because it includes a (quoted) function definition. See Serializable functions.

(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 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]]]}}
   [:ref ::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

Technically, you only need the :ref in recursive positions. However, it is best practice to :ref all references to recursive variables for better-behaving generators:

;; Note:
[:schema {:registry {::cons [:maybe [:tuple pos-int? [:ref ::cons]]]}}
 ::cons]
;; produces the same generator as the "unfolded"
[:maybe [:tuple pos-int? [:schema {:registry {::cons [:maybe [:tuple pos-int? [:ref ::cons]]]}} ::cons]]]
;; while 
[:schema {:registry {::cons [:maybe [:tuple pos-int? [:ref ::cons]]]}}
 [:ref ::cons]]
;; has a direct correspondance to the following generator:
(gen/recursive-gen
  (fn [rec] (gen/one-of [(gen/return nil) (gen/tuple rec)]))
  (gen/return nil))

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})
; => "CaR@MavCk70OHiX.yZ"

;; :gen/return (note, not validated)
(mg/generate
 [:and {:gen/return 42} :int])
; => 42

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

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

;; portable :gen/fmap (requires `org.babashka/sci` dependency to work)
(mg/generate
  [:and {:gen/fmap '(partial str "kikka_")} string?]
  {:seed 10, :size 10})
;; => "kikka_nWT3K0ya7"

;; :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)

:and generation

Generators for :and schemas work by generating values from the first child, and then filtering out any values that do not pass the overall :and schema.

For the most reliable results, place the schema that is most likely to generate valid values for the entire schema as the first child of an :and schema.

;; BAD: :string is unlikely to generate values satisfying the schema
(mg/generate [:and :string [:enum "a" "b" "c"]] {:seed 42})
; Execution error
; Couldn't satisfy such-that predicate after 100 tries.

;; GOOD: every value generated by the `:enum` is a string
(mg/generate [:and [:enum "a" "b" "c"] :string] {:seed 42})
; => "a"

You might need to customize the generator for the first :and child to improve the chances of it generating valid values.

For example, a schema for non-empty heterogeneous vectors can validate values by combining :cat and vector?, but since :cat generates sequences we need to use :gen/fmap to make it generate vectors:

;; generate a non-empty vector starting with a keyword
(mg/generate [:and [:cat {:gen/fmap vec}
                    :keyword [:* :any]]
                   vector?]
             {:size 1
              :seed 2})
;=> [:.+ [1]]

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 :int]
;   [: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 better performance, use mp/provider:

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

;; 5ms
(p/bench (mp/provide samples))

;; 500µs (10x)
(let [provider (mp/provider)]
  (p/bench (provider samples)))

:map-of inferring

By default, :map-of is not inferred:

(mp/provide
 [{"1" [1]}
  {"2" [1 2]}
  {"3" [1 2 3]}])
;[:map
; ["1" {:optional true} [:vector :int]]
; ["2" {:optional true} [:vector :int]]
; ["3" {:optional true} [:vector :int]]]

With ::mp/map-of-threshold option:

(mp/provide
 [{"1" [1]}
  {"2" [1 2]}
  {"3" [1 2 3]}]
 {::mp/map-of-threshold 3})
; [:map-of :string [:vector :int]]

Sample-data can be type-hinted with ::mp/hint:

(mp/provide
  [^{::mp/hint :map-of}
   {:a {:b 1, :c 2}
    :b {:b 2, :c 1}
    :c {:b 3}
    :d nil}])
;[:map-of 
; :keyword 
; [:maybe [:map 
;          [:b :int] 
;          [:c {:optional true} :int]]]]

:tuple inferring

By default, tuples are not inferred:

(mp/provide
  [[1 "kikka" true]
   [2 "kukka" true]
   [3 "kakka" true]])
; [:vector :some]

With ::mp/tuple-threshold option:

(mp/provide
  [[1 "kikka" true]
   [2 "kukka" true]
   [3 "kakka" false]]
  {::mp/tuple-threshold 3})
; [:tuple :int :string :boolean]

Sample-data can be type-hinted with ::mp/hint:

(mp/provide
  [^{::mp/hint :tuple}
   [1 "kikka" true]
   ["2" "kukka" true]])
; [:tuple :some string? boolean?]

value decoding in inferring

By default, no decoding is applied for (leaf) values:

(mp/provide
 [{:id "caa71a26-5fe1-11ec-bf63-0242ac130002"}
  {:id "8aadbf5e-5fe3-11ec-bf63-0242ac130002"}])
; => [:map [:id string?]]

Adding custom decoding via ::mp/value-decoders option:

(mp/provide
 [{:id "caa71a26-5fe1-11ec-bf63-0242ac130002"
   :time "2021-01-01T00:00:00Z"}
  {:id "8aadbf5e-5fe3-11ec-bf63-0242ac130002"
   :time "2022-01-01T00:00:00Z"}]
 {::mp/value-decoders {:string {:uuid mt/-string->uuid
                                'inst? mt/-string->date}}})
; => [:map [:id :uuid] [:time inst?]

Destructuring

Schemas can also be inferred from Clojure Destructuring Syntax.

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

(def infer (comp :schema md/parse))

(infer '[a b & cs])
; => [:cat :any :any [:* :any]]

Malli also supports adding type hints as an extension to the normal Clojure syntax (enabled by default), inspired by Plumatic Schema.

(infer '[a :- :int, b :- :string & cs :- [:* :boolean]])
; => [:cat :int :string [:* :boolean]]

Pulling out function argument schemas from Vars:

(defn kikka
  ([a] [a])
  ([a b & cs] [a b cs]))

(md/infer #'kikka)
;[:function
; [:=> [:cat :any] :any]
; [:=> [:cat :any :any [:* :any]] :any]]

md/parse uses the following options:

key	description
::md/inline-schemas	support plumatic-style inline schemas (true)
::md/sequential-maps	support sequential maps in non-rest position (true)
::md/references	qualified schema references used (true)
::md/required-keys	destructured keys are required (false)
::md/closed-maps	destructured maps are closed (false)

A more complete example:

(infer '[a [b c & rest :as bc]
         & {:keys [d e]
            :demo/keys [f]
            g :demo/g
            [h] :h
            :or {d 0}
            :as opts}])
;[:cat
; :any
; [:maybe [:cat
;          [:? :any]
;          [:? :any]
;          [:* :any]]]
; [:altn
;  [:map
;   [:map
;    [:d {:optional true} :any]
;    [:e {:optional true} :any]
;    [:demo/f {:optional true}]
;    [:demo/g {:optional true}]
;    [:h {:optional true} [:maybe [:cat
;                                  [:? :any]
;                                  [:* :any]]]]]]
;  [:args
;   [:*
;    [:alt
;     [:cat [:= :d] :any]
;     [:cat [:= :e] :any]
;     [:cat [:= :demo/f] :demo/f]
;     [:cat [:= :demo/g] :demo/g]
;     [:cat [:= :h] [:maybe [:cat
;                            [:? :any]
;                            [:* :any]]]]
;     [:cat :any :any]]]]]]

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 or cherry (for client side) as external dependency. If it is not present, the malli function evaluator throws :sci-not-available exception.

For ClojureScript, you need to require sci.core or malli.cherry manually.

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

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

NOTE: For now, the AST syntax in considered 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 :compile of type properties children options -> opts:

(def Between
  (m/-simple-schema
   {:type `Between
    :compile (fn [_properties [min max] _options]
               (when-not (and (int? min) (int? max))
                 (m/-fail! ::invalid-children {:min min, :max max}))
               {:pred #(and (int? %) (<= min % max))
                :min 2 ;; at least 1 child
                :max 2 ;; at most 1 child
                :type-properties {:error/fn (fn [error _] (str "should be between " min " and " max ", was " (:value error)))
                                  :decode/string mt/-string->long
                                  :json-schema {:type "integer"
                                                :format "int64"
                                                :minimum min
                                                :maximum max}
                                  :gen/gen (gen/large-integer* {:min (inc min), :max max})}})}))

(m/form [Between 10 20])
; => [user/Between 10 20]

(-> [Between 10 20]
    (m/explain 8)
    (me/humanize))
; => ["should be between 10 and 20, was 8"]

(mg/sample [Between -10 10])
; => (-1 0 -2 -4 -4 0 -2 7 1 0)

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

See also Recursive Schemas.

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.registry :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 implementations

Malli supports multiple type of registries.

Immutable registry

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

Var registry

Var is a valid reference type in Malli. To support auto-resolving Var references to Vars, mr/var-registry is needed. It is enabled by default.

(def UserId :string)

(def User
  [:map
   [:id #'UserId]
   [:friends {:optional true} [:set [:ref #'User]]]])

(mg/sample User {:seed 0})
;({:id ""}
; {:id "6", :friends #{{:id ""}}}
; {:id ""}
; {:id "4", :friends #{}}
; {:id "24b7"}
; {:id "Uo"}
; {:id "8"}
; {:id "z5b"}
; {:id "R9f"}
; {:id "lUm6Wj9gR"})

Mutable registry

clojure.spec introduces a mutable global registry for specs. The mutable registry in malli forces 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!)

In action:

Static type checking via Typed Clojure

Typed Clojure is an optional type system for Clojure.

typed.malli can consume a subset of malli schema syntax to statically type check and infer Clojure code.

See this in action in the malli-type-providers example project.

(ns typed-example.malli-type-providers
  (:require [typed.clojure :as t]
            [malli.core :as m]))

;; just use malli instrumentation normally
(m/=> foo [:=> [:cat :int] :int])
;; Typed Clojure will statically check `foo` against its schema (after converting it to a type)
(defn foo [t] (inc t))
;; Typed Clojure will automatically infer `foo`s type from its schema
(foo 1)

(comment (t/check-ns-clj)) ;; check this ns

Visualizing schemas

DOT

Transforming Schemas into DOT 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:

Lite

Simple syntax sugar, like data-specs, but for malli.

As the namespace suggests, it's experimental, built for reitit.

(require '[malli.experimental.lite :as l])

(l/schema
 {:map1 {:x int?
         :y [:maybe string?]
         :z (l/maybe keyword?)}
  :map2 {:min-max [:int {:min 0 :max 10}]
         :tuples (l/vector (l/tuple int? string?))
         :optional (l/optional (l/maybe :boolean))
         :set-of-maps (l/set {:e int?
                              :f string?})
         :map-of-int (l/map-of int? {:s string?})}})
;[:map
; [:map1
;  [:map
;   [:x int?]
;   [:y [:maybe string?]]
;   [:z [:maybe keyword?]]]]
; [:map2
;  [:map
;   [:min-max [:int {:min 0, :max 10}]]
;   [:tuples [:vector [:tuple int? string?]]]
;   [:optional {:optional true} [:maybe :boolean]]

Options can be used by binding a dynamic l/*options* Var:

(binding [l/*options* {:registry (merge
                                  (m/default-schemas)
                                  {:user/id :int})}]
  (l/schema {:id (l/maybe :user/id)
             :child {:id :user/id}}))
;[:map
; [:id [:maybe :user/id]]
; [:child [:map [:id :user/id]]]]

Performance

Malli tries to be really, really fast.

Validation performance

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 expected {: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 (= expected (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 (= expected (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 (19000ns)
(let [spec (spec/keys :req-un [::x ::z] :opt-un [::y])
      transform #(st/coerce spec % st/string-transformer)]
  (assert (= expected (transform data)))
  (cc/quick-bench (transform data)))

;; plumatic schema (2200ns)
(let [schema {:x schema/Bool
              (schema/optional-key :y) schema/Int
              :z schema/Str}
      transform (sc/coercer schema sc/string-coercion-matcher)]
  (assert (= expected (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?, pos?, neg?, 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, :some, :nil, :string, :int, :double, :boolean, :keyword, :qualified-keyword, :symbol, :qualified-symbol, 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.

malli.experimental.time

The time namespace adds support for time formats as defined by ISO 8601 - Date and time — Representations for information interchange.

Currently supported platform and providing implementations:

• JVM: via the java.time package.
• JS: via the js-joda package

The following schemas and their respective types are provided:

Schema	Example	JVM/js-joda Type (java.time)
:time/duration	PT0.01S	Duration
:time/period	P-1Y100D	Period
:time/instant	2022-12-18T12:00:25.840823567Z	Instant
:time/local-date	2020-01-01	LocalDate
:time/local-date-time	2020-01-01T12:00:00	LocalDateTime
:time/local-time	12:00:00	LocalTime
:time/offset-date-time	2022-12-18T06:00:25.840823567-06:00	OffsetDateTime
:time/offset-time	12:00:00+00:00	OffsetTime
:time/zone-id	UTC	ZoneId
:time/zone-offset	+15:00	ZoneOffset
:time/zoned-date-time	2022-12-18T06:00:25.840823567-06:00[America/Chicago]	ZonedDateTime

To use these schemas, add the schemas provided by (malli.experimental.time/schemas) to your registry.

Using time-schemas to default registry:

(require '[malli.experimental.time :as met])

(mr/set-default-registry!
  (mr/composite-registry
    (m/default-schemas)
    (met/schemas)))

To use these schemas in ClojureScript you will need to install the npm packages @js-joda/core and @js-joda/timezone.

npm install @js-joda/core @js-joda/timezone

Because historical timezone data can add ~500kb to your ClojureScript build malli does not require the @js-joda/timezone package directly. You must require timezone data before requiring the malli.experimental.time namespace if you want to make use of zone related time objects.

For example, to include only timezone data for +/- 5 years from the time the library was released, use:

(ns com.my-co.my-app
  (:require ["@js-joda/timezone/dist/js-joda-timezone-10-year-range"]))

For more info see:

https://github.com/js-joda/js-joda/tree/main/packages/timezone

min/max

Time schemas respect min/max predicates for their respective types:

(import (java.time LocalTime))

[:time/local-time {:min (LocalTime/parse "12:00:00") :max (LocalTime/parse "13:00:00")}]

Will be valid only for local times between 12:00 and 13:00.

For the comparison of Periods, units are compared to corresponding units and never between.

For example a Period of 1 year will always compare greater than a period of 13 months; that is, conceptually (< P13M P1Y)

If you want to add further constraints you can transform your Periods before being used in min and max per your use-case or combine the schema with :and and :fn for example.

Transformation - malli.experimental.time.transform

The malli.experimental.time.transform namespace provides a time-transformer from string to the correct type.

Formats can be configured by providing a formatter or a pattern property

• pattern: should be a string
• formatter: should be a DateTimeFormatter

(require '[malli.experimental.time.transform :as mett])

(as-> "20200101" $
  (m/decode [:time/local-date {:pattern "yyyyMMdd"}] $ (mett/time-transformer))
  (m/encode [:time/local-date {:pattern "yyyy_MM_dd"}] $ (mett/time-transformer))
  (= "2020_01_01" $))
; => true

Generators - malli.experimental.time.generator

Require malli.experimental.time.generator to add support for time schema generators.

Generated data also respects min/max properties.

When generating Periods there is no way distinguish between nil values and zero for each unit, so zero units will not constrain the generator, if you need some of the units to be zero in generated Periods you can always gen/fmap the data:

[:time/period {:gen/fmap #(. % withMonths 0) :min (. Period of -10 0 1)}]

This would generate Periods with a minimum years unit of -10, minimum days unit of 1 and months unit always equal to zero. Without the fmap the months unit could be any negative or positive integer.

JSON Schema - malli.experimental.time.json-schema

Require malli.experimental.time.json-schema to add support for json schema time formats.

Json schema formats map to the following string formats:

• time/local-date: date
• time/offset-time: time
• time/offset-date-time: date-time
• time/duration: duration

Description

You can call describe on a schema to get its description in english:

(require '[malli.experimental.describe :as med])

(med/describe [:map {:closed true}
               [:x {:optional true} int?]
               [:y :boolean]])
;; => "map where {:x (optional) -> <integer>, :y -> <boolean>} with no other 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
• 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
• experimental: stuff in malli.experimental ns, code might change be moved under a separate support library, but you can always copy the old implementation to your project, so ok to use.
• private API: private vars and malli.impl namespaces, all bets are off.

Development

Malli is open for contributions. Before contributing with a PR, please open an issue for it.

Adding new schema types

To add a new schema type, e.g. :float, you should adding the following:

• schema definition to malli.core + tests
• default encoder/decoder mappings into malli.transform + tests
• JSON Schema mappings into malli.json-schema + tests
• Generators into malli.generator + tests
• OPTIONALLY adding inferrers into malli.provider + tests
• update README.md

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 cherry
npx shadow-cljs run shadow.cljs.build-report app-cherry /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

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

Formatting the code

clojure-lsp format
clojure-lsp clean-ns

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'

Babashka

Since version 0.8.9 malli is compatible with babashka, a native, fast starting Clojure interpreter for scripting.

You can add malli to bb.edn:

{:deps {metosin/malli {:mvn/version "0.9.0"}}}

or directly in a babashka script:

(ns bb-malli
  (:require [babashka.deps :as deps]))

(deps/add-deps '{:deps {metosin/malli {:mvn/version "0.9.0"}}})

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

(prn (malli/validate [:map [:a [:int]]] {:a 1}))
(prn (malli/explain [:map [:a [:int]]] {:a "foo"}))

3rd party libraries

• 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
• malapropism - malli-backed configuration library
• muotti - a graph based value transformer library with malli-support
• malli-select - spec2 selection for Malli (for when you only need part of the herd 🐑)

License

Copyright © 2019-2022 Metosin Oy and contributors.

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