Container (type theory) Source: en.wikipedia.org/wiki/Container_(type_theory)

In type theory, a discipline within mathematical logic, containers are abstractions which permit various "collection types", such as lists and trees, to be represented in a uniform way. A (unary) container is defined by a type of shapes S and a type family of positions P, indexed by S. The extension of a container is a family of dependent pairs consisting of a shape (of type S) and a function from positions of that shape to the element type. Containers can be seen as canonical forms for collection types.^[1]

For lists, the shape type is the natural numbers (including zero). The corresponding position types are the types of natural numbers less than the shape, for each shape.

For trees, the shape type is the type of trees of units (that is, trees with no information in them, just structure). The corresponding position types are isomorphic to the types of valid paths from the root to particular nodes on the shape, for each shape.

Note that the natural numbers are isomorphic to lists of units. In general the shape type will always be isomorphic to the original non-generic container type family (list, tree etc.) applied to unit.

One of the main motivations for introducing the notion of containers is to support generic programming in a dependently typed setting.^[1]

Containers in type theory correspond to polynomial functors in a categorical setting. A container defined by a set of shapes ${\displaystyle S}$ and a family of positions ${\displaystyle P(s)}$ induces an endofunctor ${\displaystyle T:\mathbf {Set} \to \mathbf {Set} }$ defined by:

${\displaystyle T(X)=\sum _{s\in S}P(s)\to X}$

For a function ${\displaystyle f:X\to Y}$, the functor action is:

${\displaystyle T(f)(s,g)=(s,f\circ g)}$

This abstractly models the familiar `map` operation, ensuring functoriality and strength (i.e., compatibility with product structures). Containers form a full subcategory of endofunctors over Set, closed under initial algebras and final coalgebras.

Moreover, containers correspond to polynomial functors internally representable in any locally cartesian closed category (LCCC), supporting a wide range of categorical constructions.^[2][3]

Indexed containers (also called dependent polynomial functors) extend the idea of containers to families of types indexed over a base type ${\displaystyle I}$. An indexed container consists of:

• A family of shapes ${\displaystyle S(i)}$ indexed by ${\displaystyle i\in I}$,
• A set of positions ${\displaystyle P(i,s)}$ for each shape ${\displaystyle s\in S(i)}$.

Its extension is a functor:

${\displaystyle T(X)_{i}=\sum _{s\in S(i)}\prod _{p\in P(i,s)}X_{j(p)}}$

Here, ${\displaystyle j(p)}$ maps positions to result indices. This construction can model data types with dependent structure like vectors (lists indexed by length) and syntax with binding.^[4]

Categorically, indexed containers correspond to dependent polynomial functors in a locally cartesian closed category. They are closed under constructions such as reindexing, initial algebras, and dependent sum/product types.^[5]

Every indexed container can be represented using small inductive-recursive definitions, providing a unifying view of data families in dependent type theory.^[6]

• Container (abstract data type)
• Polynomial functor (type theory)

• Container Types blog

${\displaystyle \Gamma \!\vdash }$

This programming language theory or type theory-related article is a stub. You can help Wikipedia by expanding it.

• v
• t
• e