Reified
Repository F# setup
open Reified
open Reified.Refinements
open Reified.Result

Choosing an API

Five operations read a declaration. They differ in the input they accept, the validation they run, and the error they return. Pick by where the input came from and whether the producer is trusted.

Operation Input Validation run Constructor Error type Trust the payload?
Schema.parse Data Field decoding, constraints, refinements, and constructor invariants Runs only after every field succeeds SchemaErrors (accumulated, path-aware) No — the untrusted boundary
Schema.check an existing 'model Same field rules and constructor as Schema.parse, read back from the value Re-invoked with the checked fields; returns the original value on success SchemaErrors (accumulated, path-aware) No — the value's history is uncertain
Json.deserialize JSON string Wire shape and required fields only; constraint metadata is skipped Runs; a checked constructor failure raises raises JsonCodecException Yes — storage, internal service, cache
Json.tryDeserialize JSON string Same as Json.deserialize Same as Json.deserialize Result<'model, string> with the first structural failure Yes — same trusted path, non-throwing
Contract.parse Data Version selection, that version's schema parse, migrations, then a Schema.check of the migrated value Via each version's parse and the final recheck ContractError No — old wire shapes must stay readable

Json.parseData is the bridge between the first and third rows: it turns JSON text into Data with no schema applied. Schema.parseRetainingInput wraps Schema.parse and keeps the original Data for redisplay and auditing.

Verified
open Reified
open Reified.SchemaDSL
open Reified.ConstraintDSL

type Signup = { Email: string; Age: int }

let signupSchema =
    schema<Signup> {
        field _.Email { constraints [ present; email ] }
        field _.Age { constrain (atLeast 13) }
        construct (fun email age -> { Email = email; Age = age })
    }

let data = Data.ofNameValues [ "email", "ada"; "age", "11" ]

Schema.parse signupSchema data
// Result<Signup, SchemaErrors>

Schema.check signupSchema { Email = "ada"; Age = 11 }
// Result<Signup, SchemaErrors>

let codec = Json.compile signupSchema

Json.tryDeserialize codec """{"email":"ada@example.org","age":36}"""
// Ok signup — constraint metadata is not run on the trusted path

Json.tryDeserialize codec """{"email":"ada@example.org","age":"thirty"}"""
// Error "JSON decode failed at $.age: expected digit"

When the same payload needs different treatment

  • A request body from a client you do not control goes through Json.parseData (or Json.tryParseData) into Data, then Schema.parse — never through Json.deserialize, which would skip the rules.
  • A value read back from your own database or another service you trust can use Json.deserialize/ Json.tryDeserialize. If the payload carries constraint attributes that matter, remember the trusted path does not read them.
  • A value produced by a serializer, mapper, or import goes through Schema.check, which re-runs the rules on a value that already exists.
  • A payload whose shape has shipped in several versions goes through Contract.parse, which selects the version, parses, migrates, and rechecks the current model.

The same trust distinction is stated once in JSON codecs and in the Schema or Generated DTO decision.