Polymorphic Variants

Polymorphic variants are a feature of the OCaml type system that provides a flexible, open-ended form of algebraic data types. Unlike ordinary variant types, which require a fixed, closed set of constructors defined at the point of declaration, polymorphic variants allow constructors to be used without prior declaration and permit the type system to infer the set of valid constructors from context. This makes them particularly suited to protocols, extensible systems, and interface evolution.

The key semantic property of polymorphic variants is that they are structural rather than nominal. An ordinary variant type is identified by its name: two variant types with the same constructors are considered distinct if they have different names. A polymorphic variant type is identified by its structure: the same constructor can appear in multiple types, and the type system tracks which constructors are valid in each context. This enables a form of subtyping that is otherwise unavailable in OCaml's rigid type system.

Polymorphic variants are not merely a convenience for protocol design. They represent a deep structural insight: that the boundaries of a type need not be fixed at definition time, but can evolve as the type is used. This is the type-theoretic analogue of open systems in biology and sociology — systems whose boundaries are not given but negotiated. The polymorphic variant is not a hack to work around OCaml's static typing; it is a demonstration that static typing can be as flexible as dynamic typing when the type system is designed with the right primitives.

The absence of polymorphic variants from every mainstream language except OCaml is not a technical limitation but a design blindness. Languages that add "union types" as a feature years after their introduction are rediscovering what OCaml already proved: that the static-dynamic typing divide is a false dichotomy, and that the right type system can be both rigid and flexible, both safe and expressive.