Algebraic data types and patterns

An algebraic data type describes a fixed set of alternatives. Each constructor identifies one alternative and may carry values specific to that case. This makes the possible shapes of a value explicit in its type.

This checked program defines a generic result type and matches both branches:

data Result e a = Err e | Ok a.
unwrapOr :: Int -> Result(Int, Int) -> Int.
unwrapOr = \(fallback, result) -> case result {
  | Err _ -> fallback
  | Ok item -> item
}.
unwrapOr 0 (Ok 41).

Expected output:

41

Patterns test a value's shape and bind the parts needed by a branch. They can match literals, constructors, tuples, lists, alternatives, or a value together with one of those shapes. _ ignores a part that the branch does not need.

Pattern alternatives are available at the top level of a case arm or pattern lambda parameter. Every alternative must bind the same names with the same types. Nested alternatives and guards on pattern-lambda parameters are not currently supported; case arms support one guard after the complete pattern.

The compiler checks each pattern against the value being matched. Constructors must belong to its type and carry the expected number of fields. Tuple and list shapes, duplicate binders, alternative binders, guard types, and branch result types are checked before evaluation.

Case arms are tried in source order and must cover the complete resolved input type. Unguarded constructor arms can exhaust a closed ADT; guarded arms do not contribute coverage. The compiler reports an example missing pattern for an incomplete match and rejects an arm whose whole pattern space was covered by earlier unguarded arms. See control flow for guards, ordering, and diagnostic codes.