MuniHac

The classical planning branch of AI - one of the oldest branches of AI - is primarily interested in determining a sequence of actions to move from a known state to another state that satisfies some conditions required by an operator. Traditionally, this work has happened in the very physical space of moving robots, but it's so much more general! In this talk, Ollie will demonstrate his work on a Haskell DSL to interact to the state of the art general purpose solver "Fast Downward", but in the context of automating a hypothetical smart home.

In this talk, we will look at a seemingly simple programming problem – checking whether the elements in a binary tree are in order – and not only explore many different approaches to solving this problem, but also show how they are all connected. As we do that, we will touch upon equational reasoning, defunctionalization, the worker-wrapper transformation, and more. We learn about the limits of type-driven development and finally appreciate the deep significance of type class laws.

This talk only assume elementary knowledge of Haskell (or functional programming in general): Algebraic data types and recursive functions.

I've always wanted a Haskell IDE. The absence of a simple and robust solution led me to build Ghcid, which I've been using for many years. However, I have moved on, now using a real Haskell IDE based on technology we developed at Digital Asset for the DAML programming language. It turns out building an IDE is harder than I expected, so in this talk I'll cover three topics:

1. The theory behind building an IDE. We use an in-memory dependency graph, with some interesting tweaks.
2. What's there today and how you can use it. I'll also cover the relationship to haskell-ide-engine, DAML and other related projects.
3. What's missing and how you can help. I'll be hacking on the IDE at MuniHac, and everyone is welcome to join in.