This BHR serves as the entire collection of all the Haskell packages I've written over the years. There are some libraries (definitive-*), some serious executables (curly), and some random unfinished wankery (grow/woosh).

You're welcome to use any of the libraries therein in your own code, if you can make sense of them. If you just want to test out one of the programs, you can also check out the release page to find a compiled and ready-to-use version.

Here is a short description of the different packages in this BHR.

Those are basic libraries, designed to minimize external dependencies while providing most of the functionality found in modern Haskell programs. Most notably, optics (Lenses and the like) are defined very early on in the definitive-base library, and serve as the basis for many standard abstractions.

The definitive-base library defines all the standard Monad transformer combinators, from StateT to LogicT, along with a generalization of the containers library that simplifies the use of Sets and Maps, and introduces the Bimap, Relation and Equiv containers under the same interface.

The definitive-parser library, as it name indicates, defines the usual parser combinators (using the LogicT transformer for non-determinism), and is used for both binary and textual parsing in most other packages. It also defines Serializable and Format typeclasses, as well as their instances for most basic data types and a mechanism for automatic derivation of those classes for types that are instances of the Generic class.

The definitive-graphics library provides high-level wrappers for the GTK windowing framework, allowing one to describe GUIs and their behaviour in the perfect mix of declarative layout and imperative event handling.

The other two (definitive-network and definitive-filesystem) are much less interesting, and should definitely not be included in any serious project.

The packages whose name start with curly are part of the Curly compiler infrastructure (on which more information can be found at curly-lang.org).

The curly package contains the library that describes how the compiler interacts with its environment (reads configuration files, locates packages, host interactive sessions, ...). Using that library, it also implements the curly executable, aka the compiler itself. The curly library is also used by curly-gui, a GUI version of the compiler, to handle all its context-related operations.

At a lower level, the curly-system library contains the implementation of all of Curly's backends, from x86 assembly to JavaScript. You could write a basic batch compiler using only this library.

All of the above packages depend on the curly-core library, which defines all the aspects of the core language, in an architecture-independent way. This is where you can find a description of the object format (in the Library type), the type system, and the syntax and semantic of the Curly language itself.

This BHR also serves as a testing ground for potentially interesting ideas to develop, mainly because if I want to import stuff from another library in here, all I have to do is declare it in the dependencies of the corresponding .cabal file. I blame this all squarely on Stack, for being such a great tool.

CaPriCon is the answer to the question "what would an assembly language for mathematical proofs look like ?". It's a low-level, stack-based language capable of manipulating terms in the Calculus of Prismatic Constructions. More information is available here.

Since I already had a monad for concatenative languages, why not use it for fun, too ?

Logos is a language that makes it easy to define 3D scenes, using the modern OpenGL rendering pipeline for efficiency. It also provide the basic event handling capability necessary for implenting a full-fledged game.

One caveat : you can't yet draw any text. You can load font textures, and manually map their contents onto a quad, but there's no abstraction for it yet (and there will be, in the far future).

After spending months wrestling with libncursesw version issues on different distros, I came to the conclusion that writing my own would actually save me some time. So I did. And it did.

It's not a clone of the "real" ReadLine, not even close. But it handles simple text editing, programmable autocompletion, and command history. Plus, it fits in one ~200LoC file, so it's easy to change something and understand what you did.