![](https://private-user-images.githubusercontent.com/106390/298580220-af04daa4-6d5f-49ee-ba33-bc971b7355fc.png?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJpc3MiOiJnaXRodWIuY29tIiwiYXVkIjoicmF3LmdpdGh1YnVzZXJjb250ZW50LmNvbSIsImtleSI6ImtleTUiLCJleHAiOjE3MjUyNDkzOTYsIm5iZiI6MTcyNTI0OTA5NiwicGF0aCI6Ii8xMDYzOTAvMjk4NTgwMjIwLWFmMDRkYWE0LTZkNWYtNDllZS1iYTMzLWJjOTcxYjczNTVmYy5wbmc_WC1BbXotQWxnb3JpdGhtPUFXUzQtSE1BQy1TSEEyNTYmWC1BbXotQ3JlZGVudGlhbD1BS0lBVkNPRFlMU0E1M1BRSzRaQSUyRjIwMjQwOTAyJTJGdXMtZWFzdC0xJTJGczMlMkZhd3M0X3JlcXVlc3QmWC1BbXotRGF0ZT0yMDI0MDkwMlQwMzUxMzZaJlgtQW16LUV4cGlyZXM9MzAwJlgtQW16LVNpZ25hdHVyZT0zY2JmM2E3MzUyZjY2MmIyNWZmY2IwMmQzNjRmZGZkZWJlOTAwN2ExMTAwYjUwOGEwZWM5NzNjMGExYWQ0MTgwJlgtQW16LVNpZ25lZEhlYWRlcnM9aG9zdCZhY3Rvcl9pZD0wJmtleV9pZD0wJnJlcG9faWQ9MCJ9.F0gsyF-soTa-CBCn9HMkAY7nUsY4xWt0ujdhZvusEL4)
/!\ WIP - everything is subject to change
Tiny library allowing you to generate json-schema from clojure specs.
Specs allow us to define contracts, why not re-using these to generate json-schema. Spec forms are just "data", should be easy right?
Well, there are a few challenges:
- specs can be defined as arbitrary predicates
- specs can also be aliases to other specs
- specs can be the result of composition of other specs via s/and s/merge s/multi-spec etc
- specs can be parameterized (kind of, think
s/coll-of
,s/int-in
& co) - specs have no metadata, that makes adding features json-schema supports but spec doesn't a bit complicated
How pact
attempts to handle these:
-
Spec forms/composition of specs: schemas can be inferred for all clojure.spec forms (s/and & co). In the cases where we cannot infer the schema we provide ways for you to specify what to generate.
-
Spec alias chains: we ensure that alias chains are understood and walk them up trying to find the first spec key that will allow json-schema generation. For instance if you have a spec ::foo that is an alias to ::bar that is itself an alias to ::baz that is a
string?
, trying to generate json-schema for ::foo will check them in order until it finds enough information to do so (from ::baz definition). -
Metadata: we have a few helpers that allow you to specify/override
title
,description
,format
,pattern
,$id
on top of existing specs, that will later show up in the json-schema for these. They also understand spec aliases and pick up the first walking back the spec alias chain. -
Register new generators, either globally or on per-call basis.
-
Arbitrary predicates: Predicate forms parsing is done using spec , with conform. We destructure predicate forms via conform, so that you grab any value from them (anything in the body of the function really) in order to then generate the appropriate openapi data via a supplied function that take these bindings as argument.
pact
comes with a number of useful predicate parsers that allow to generate correct schemas for common cases (numercic comparaisons, length bounds and so on). This is really an escape hatch, generally you don't have to go that far and can get by registering generators foridents
orforms
(more on this later).
By default pact is strict, it will throw at generation time if it cannot infer the json-schema for a spec, but it will allow you to specify the missing bits.
It can also function in non strict mode where unknowns generate whatever you specify by default, or just skip what it can't infer in some cases.
You can also tune the interpretation of some forms to be less strict, for
instance having only the first component of a s/and
to be taken into account
and a few others like this. But by default we try to cover the full spec.
We do not provide an openapi generator, if you want to generate openapi
using pact
it's very easy to do so, there's no need for an extra lib layer to
do so. That also gives you more control over the way you manage $refs, paths and
other openapi details.
(require '[s-exp.pact :as p])
(require '[clojure.spec.alpha :as s])
;; simple specs
(p/json-schema `(s/and number? int?))
=> {:allOf [{:type "number"} {:type "integer" :format "int64"}]}
(p/json-schema `(s/or number? int?))
=> {:anyOf [{:type "number"} {:type "integer" :format "int64"}]}
(p/json-schema `(s/coll-of string? :max-count 3))
=> {:type "array", :items {:type "string"} :maxItems 3}
(s/def ::foo string?)
(s/def ::bar keyword?)
(s/def ::baz int?)
;; maps
(s/def ::thing (s/keys :req-un [::foo ::bar] :opt-un [::baz]))
(p/json-schema ::thing)
=> {:type "object",
:properties
{"foo" {:type "string"},
"bar" {:type "string"},
"baz" {:type "integer", :format "int64"}},
:required ["foo" "bar"]}
;; merged maps
(p/json-schema `(s/merge (s/keys :req-un [::foo]) (s/keys :req-un [::bar]) (s/keys :opt-un [::baz])))
=> {:allOf
[{:type "object",
:properties {"foo" {:type "string"}},
:required ["foo"]}
{:type "object",
:properties {"bar" {:type "string"}},
:required ["bar"]}
{:type "object",
:properties {"baz" {:type "integer", :format "int64"}}}]}
;; multi-specs
(s/def ::tag #{:a :b :c :d})
(s/def ::example-key keyword?)
(s/def ::different-key keyword?)
(defmulti tagmm :tag)
(defmethod tagmm :a [_] (s/keys :req-un [::tag ::example-key]))
(defmethod tagmm :default [_] (s/keys :req-un [::tag ::different-key]))
(s/def ::ms (s/multi-spec tagmm :tag))
(p/json-schema ::ms)
=> {:anyOf
[{:type "object",
:properties {"tag" {:enum #{:c :b :d :a}}, "different-key" {:type "string"}},
:required ["tag" "different-key"]}
{:type "object",
:properties {"tag" {:enum #{:c :b :d :a}}, "example-key" {:type "string"}},
:required ["tag" "example-key"]}]}
;; arbitrary predicates
(s/def ::p (s/and number? (fn [x] (>= x 10))))
(p/json-schema ::p)
=> {:allOf [{:type "number"} {:minimum 10, :type "number"}]}
;; metadata
(-> (s/def ::animal string?)
(p/with-description "An animal")
(p/with-title "Animal")
(p/with-pattern "[a-zA-Z]")
(p/json-schema))
=> {:type "string", :title "Animal", :description "An animal", :pattern "[a-zA-Z]"}
;; overriding output, we make string? return something different
(p/json-schema ::animal {:idents {`string? {:type "string" :foo "bar"}}})
=>
{:type "string",
:foo "bar",
:title "Animal",
:description "An animal",
:pattern "[a-zA-Z]"}
;; also works for parameterized forms
(p/json-schema `(s/coll-of any?) {:forms {`s/coll-of (fn [_coll-of-arg _opts] {:foo :bar})}})
=> {:foo :bar}
;; it understands alias chains
(s/def ::foo string?)
(s/def ::bar ::foo)
(s/def ::baz ::bar)
(p/json-schema ::baz)
=> {:type "string"}
;; also inherits all properties from chains
(p/with-pattern ::bar "[a-zA-Z]")
(p/json-schema ::baz)
{:type "string", :pattern "[a-zA-Z]"}
;; and allow overrides at any level
(with-pattern ::baz "[a-z]")
(p/json-schema ::baz)
=> {:type "string", :pattern "[a-z]"}
You can extend the way pact generates schemas via json-schema
options or by registering schema generators
s-exp.pact/register-ident!
: registers a new generator for anident
(spec key or symbol)
(register-ident! `int? {:type "integer" :format "int64"})
s-exp.pact/register-form!
: registers a generator for aform
, ex:coll-of
parameterized spec forms (think,int-in
,nilable
,coll-of
and so on) ex: how nilable is implemented
(p/register-form!
`s/nilable
(fn [[form] opts]
{:anyOf [{:type "null"}
(json-schema* form opts)]}))
-
s-exp.pact/register-pred!
: allows to set predicate conformer & schema generator for arbitrary predicates found in spec forms, ex:(s/and string? (fn [s] (str/includes? s something)))
It takes 2 arguments a spec key, that would be used to potentially conform (pattern match) a spec form onto bindings to use for a json-schema, and a function that will receive these arguments and return json-schema data matching the form.
There are also matching options you can pass to the json-schema
function to
have overrides be applied per call (without being registered globally):
:idents
- map of idents -> values:forms
- map of forms -> function of form args+opts:preds
- map of pred conformer spec key -> generator function which will take the conformed values and options
see API.md
clj -X:test
Copyright © 2024 Max Penet
Distributed under the Eclipse Public License version 1.0.