Intro 🛣️
Flamingo is compiler and runtime for ALM[1], a DSL for describing state machines in an ultra-compact way, emphasizing simplicity, modularity, and correctness.
Flamingo is built with web apps in mind and compiles to efficient, portable
Javascript solved by a WASM interpreter[2].
Flamingo programs handle state management - you can think of Flamingo as DSL for writing
Redux reducers. Hook up browser events to Flamingo's runtime, bind state changes
to the DOM as views with, e.g., React, and voilà! you have a web app!
The ALM language is designed for commonsense reasoning about real-world problems. Using a few simple but powerful constructs from logic, ALM allows us to express pure business logic, free from the incidental complexity of algorithms. As a logic programming system, Flamingo trades performance for expressiveness. Flamingo programs can be much slower than their imperative counterparts; however, they are often orders of magnitude less overall code. At the same time, ALM is carefully limited in its expressiveness compared to other logic programming systems, always favoring simplicity and maintainability over terseness and higher-orderedness.
Flamingo is also built for correctness. Its semantics as a state machine language make it amenable to automatic verification: describe both your positive and negative constraints in the same expressive language, and Flamingo will automatically verify they hold on every possible sequence of inputs[3]. It's like Alloy or TLA+, but easy.
Flamingo benefits from decades of optimization research in the fields of Datalog and ASP. Combined with Flamingos' automatic verification, these characteristics make Flamingo ideal for the rapid design and implementation of complex, practical systems.
In terms of language theory, ALM is a logic programming language, specifically an extension of Datalog that adds an explicit notion of action and change, negative constraints, and a powerful, object-oriented module system[4].
A Little Example
Here's a small example that models menus that can open and close.
Obviously, ALM's syntax is quite different from imperative or functional languages; however, it's ultimately much simpler - even this small example covers the majority of it. We'll cover the syntax in depth in the coming sections. But first, let's discuss the motivation behind the language: the tar pit of modern software development.
[1]: ALM is the "Action Language with Modules", first described in this paper by Daniela Inclezan and Michael Gelfond. It is the successor to AL, which comes from a long line of research on reasoning about actions and change beginning with John McCarthy's situation calculus in 1975.
[2]: The Flamingo-to-JS is a work in progress. For now, Flamingo is translated into ASP and solved by the wonderful Clingo solver.
[3]: This is currently a bit hand-wavy in that we only check a fixed number of objects up to a fixed horizon of actions. In the short term, this will be configurable. In the long term, a major focus of my grad research is figuring out a way to verify infinite domains via induction and SMT. If this is possible, Flamingo will be the easiest proof language in existence by a long shot.
[4]: Flamingo is a mostly-faithful Javascript implementation of the Modular Action Language ALM, a specification language with a rich history beginning with John McCarthy's situation calculus.