sml-schema

JSON Schema (draft 2020-12) parser and validator for Standard ML. Provides a typed SML abstract syntax tree mirroring the JSON Schema keywords, a parser that converts a Json.json value into a schema, and a validator that checks an instance against a schema, producing a list of structural errors.

Part of the sjqtentacles monorepo of SML libraries. It depends on sml-json (vendored, for the JSON AST and parser) and sml-rederiv (vendored, for the pattern keyword's regex matching).

Features

• Schema ADT -- SObject, SArray, SString, SNumber, SBool, SNull, SRef, SAllOf, SAnyOf, SOneOf, SNot, SDefs, STrue, SConst, SEnum mirroring the draft 2020-12 keywords.
• Parser -- fromJson : Json.json -> schema result walks a JSON value and builds the typed schema tree. Unknown keywords are ignored (draft 2020-12 permits vendor extensions); malformed keyword values cause an Err.
• Validator -- validate : schema -> Json.json -> error list checks an instance against a schema. Each error carries a JSON-Pointer-style path and a message. An empty list means the instance is valid.
• Keyword coverage -- type, properties, required, additionalProperties, items, prefixItems, minItems, maxItems, minimum, maximum, exclusiveMinimum, exclusiveMaximum, minLength, maxLength, pattern (via sml-rederiv), enum, const, allOf, anyOf, oneOf, not, $ref, $defs.

Status

Working and tested. The parser and validator cover all the keywords listed above, exercised by a test suite spanning type validation, object/array constraints, numeric bounds, pattern matching, enum/const, combinators, and local $ref/$defs resolution.

Portability

Pure Standard ML using only the Basis library (plus the vendored sml-json and sml-rederiv) -- no FFI, no threads. Verified on MLton and Poly/ML, with identical, deterministic output across both.

Building and testing

make test        # build + run the suite under MLton (default)
make test-poly   # run the suite under Poly/ML
make all-tests   # run under both
make clean

Usage

(* Parse a schema from a JSON string. *)
val schemaJson = Json.parseJson
  "{\"type\":\"object\",\"properties\":{\"name\":{\"type\":\"string\"}},\"required\":[\"name\"]}"
val schema = case schemaJson of
    CharParsec.Ok v => (case Schema.fromJson v of
                            Schema.Ok s => s
                          | Schema.Err e => raise Fail e)
  | CharParsec.Err e => raise Fail (CharParsec.errorToString e)

(* Validate an instance. *)
val instance = Json.parseJson "{\"name\":\"Alice\"}"
val errors = case instance of
    CharParsec.Ok v => Schema.validate (schema, v)
  | CharParsec.Err e => raise Fail (CharParsec.errorToString e)
(* errors = [] -- valid *)

val badInstance = Json.parseJson "{\"age\":30}"
val badErrors = case badInstance of
    CharParsec.Ok v => Schema.validate (schema, v)
  | CharParsec.Err e => raise Fail (CharParsec.errorToString e)
(* badErrors = [{path = "", message = "missing required property: name"}] *)

API summary

Function	Description
Schema.schema	The JSON Schema ADT (SObject, SArray, SString, ...).
Schema.fromJson : Json.json -> schema result	Parse a JSON value as a schema.
Schema.validate : schema -> Json.json -> error list	Validate an instance; empty list = valid.
Schema.error	{path: string, message: string}.

Dependencies

• sml-json (vendored, which vendors sml-parsec) -- JSON AST and parser.
• sml-rederiv (vendored) -- Brzozowski-derivative regex engine for the pattern keyword.

Installing with smlpkg

smlpkg add github.com/sjqtentacles/sml-schema
smlpkg sync

Then reference the library basis from your own .mlb:

lib/github.com/sjqtentacles/sml-schema/sml-schema.mlb

For Poly/ML, use the sources listed in sources.mlb in order (the vendored sml-json and sml-rederiv first, then schema.sig and schema.sml).

License

MIT. See LICENSE.