Schema

The Schema module contains primitive schemas, collection and recursion combinators, union constructors, and the operations that parse or check a completed declaration.

Start from the task:

• Declare a record schema and parse untrusted input: Schema quickstart.
• Fields, field blocks, and constructors: SchemaDSL.
• Reusable value rules: Constraints.
• Invariant-carrying field types: Refined Schemas.
• JSON in and out: JSON Codecs.
• Published shape: JsonSchema.
• Older wire versions: Versioned Contracts.

Use Schema.parse for untrusted Data. It decodes fields, applies constraints and refinements, accumulates independent failures as SchemaErrors, and invokes the model constructor only after the fields succeed.

Use Schema.check when a typed value came from another constructor, an import, or a database mapper. It reads the declared fields back from the value and checks them.

The same model constructor runs again, so cross-field invariants are checked too. Success returns the checked value itself.

Both operations return Result<_, SchemaErrors>. The difference is only the input: Schema.parse reads untrusted Data and decodes each field, while Schema.check reads fields back from a value that already exists. The field rules, constructor, error cases, and paths are the same.

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

Schema.parse signupSchema input
// Result<Signup, SchemaErrors>: untrusted Data, every field checked and accumulated

Schema.check signupSchema { Email = "ada"; Age = 11 }
// Result<Signup, SchemaErrors>: an existing value, the same field rules and constructor

The Schema quickstart introduces the workflow.

Constraint equivalents

These are the handwritten operations emitted for derivation attributes. Use them inside a field block with constrain, except for the metadata operations shown directly:

Purpose	Schema DSL
Pattern	constrain (pattern expression)
Minimum, maximum, exact, or bounded natural length	constrain (minLength n), maxLength, length, lengthBetween
Present value or supplied input key	constrain present, mustSupply
Inclusive/exclusive numeric bounds	constrain (atLeast n), greaterThan, atMost, lessThan
Numeric multiple	constrain (multipleOf n)
Distinct list elements	constrain distinct
Email text	constrain email
Open format metadata	format (SchemaFormat.create name)
Omitted-input default	defaultValue value

See Derivation Attributes for the complete attribute mapping.