comonad

A comonad is a concept from category theory that is dual to a monad. Formally, it consists of a functor W: C → C together with two natural transformations: delta: W ⇒ W∘W (a comultiplication) and epsilon: W ⇒ Id (a counit). These must satisfy laws that are dual to the monad laws: coassociativity (delta followed by W(delta) equals delta followed by delta) and counit laws (epsilon after delta is the identity, and W(epsilon) after delta is also the identity). In essence, a comonad represents contexts that can be observed and duplicated.

In programming, comonads provide a way to model computations that are context-dependent and can be extended with additional context. A common presentation uses two operations on a type constructor w: extract, which retrieves a value from a cocontext, and extend (or duplicate), which expands a context with more structure. These operations must satisfy laws dual to those of monads, such as extract . duplicate = id and extend id = id, ensuring coherent behavior of context access and extension.

Common examples include:

- The Store comonad, parameterized by a set s. A value is represented as Store s a = Store (s → a) s. Here extract (Store f s) = f s, and duplicate (Store f s) = Store (\s' → Store f s') s.

- The Environment (or pair) comonad, W a = (e, a) for a fixed environment e. extract (e, a) = a, and duplicate (e, a) = (e, (e, a)) = extend id (e, a).

Comonads arise in contexts such as dataflow modeling, context-aware computations, and structures like zippers or comonadic representations of environments. They provide a formal framework for computations that read from and manipulate surrounding context.

---