Cedille-Core
A minimal (1k LOC) programming language capable of proving theorems about its own terms.
What that means?
There are big and small programming languages out there. C++ and Haskell are big languages. Other languages, such as Brainfuck, are so simple they could be implemented in 317 Python characters. The Lambda Calculus is popular for being a simple language that serves as the foundation of many functional programming languages.
Despite being turing-complete, there is one thing those languages can't do: expressing and proving mathematical theorems about its own terms. The few languages that can do that are rather big: Idris, Agda, Coq and Isabelle are examples. Some languages like the Calculus of Constructions (such as implemented on Haskell-Morte-Library) are small and capable of expressing and proving mathematical theorems about its own terms, but, since their expressivity is very limited, they're not useful for proving useful properties about everyday programs and applications. Until recently, we had no language that was both small and featured practical theorem proving.
Cedille is a language developed by Aaron Stump, aiming to solve that problem, among others. It is capable of proving useful theorems about its own terms, yet can be implemented in a very small amount of code. Cedille-Core is a compressed version of Cedille with less type inference and smaller code size.
Syntax
| name | syntax | description |
|---|---|---|
| type of types | Type |
the type of types |
| kind | Kind |
the type of type of types |
| lambda | [var : type] body |
a function |
| -lambda | [var : type] body |
a computationally irrelevant function |
| forall | {var : type} body |
the type of a function |
| -forall | {var : type} body |
the type of a computationally irrelevant function |
| application | (f x) |
application of lambda f to argument x |
| -application | (f -x) |
application of lambda f to erased argument x |
| intersection | <x : A> B |
type of a term t that has type A and [t/x]B |
| both | @x B a b |
intersection of terms a : A and b : B[a/x] |
| first | .a |
first, erased view of a dependent intersection |
| second | +a |
second, full view of a dependent intersection |
| equality | |a = b| |
proposition that terms a and b of possibly different types are equal |
| reflexivity | $a b |
proof that |a = a|, erasing to b |
| symmetry | ~a |
if a proves |a = b|, then a proves |b = a| |
| rewrite | %x A e a |
if e proves |x = y|, replaces x by y on the type A of term a |
| cast | ^e a b |
if e proves |a = b|, then cast term b to the type of term a |
| definition | def x t u |
replaces ocurrences of x in u by t |
Technical specification
Please check the specification repository.
