Notes on Advanced Programming course from DIKU, 2025

Created, I think, mostly by Troels Henriksen. Seems like also Fritz Henglein.

• Main site: https://github.com/diku-dk/ap-e2025-pub/tree/main
  • you probably want to clone or fork that repo

Each week

• Read that week's course notes, which are like a concise textbook

• Read that week's lecture notes and work through exercises; e.g., week 1

  • To make sure you are set up with haskell, go to, e.g. week1/handout and make sure cabal test works (no errors)

  • The recommended strategy for the exercises is to start editing the starter files in the handout folder. Note that the final exercise solutions are given in the solution folder, but the goal is for you to work out the exercise solutions yourself, with minimal reference to the solution folder sources

• Work on that week's assignment; e.g., week 1 The assignment is in, e.g. "a1.pdf". The "handout" archive contains a project tree with starter files to which you are supposed to add code, including a unit test suite. For week 1, at least, the assignment starting source is just the result of doing the exercises.

Notes from working through week 2

Week 2 is implementing an interpreter for mostly the same language as for week 1, but this time in a monadic style. While I found week 1 easy, I did struggle with the monadic approach. While writing individual eval functions is more concise, the monadic plumbing takes more work. Also, I'm still vaguely uneasy that monads hide both state and control flow.

I did find that struggling with the monad plumbing was very helpful for me, including implementing the monadic type checker.

Notes from working through week 1

My goal for looking into this course was to get comfortable with a monad-centric approach to implementing a compiler, such as is pervasive in the futhark compiler.

Unfortunately, week 1 doesn't touch on monads. Nonetheless, I spent a fair amount of time tediously implementing the exercises, in the order given, and writing test cases. It's pretty standard stuff if you've written an interpreter before. It served to remind me of the basics of Haskell. The assignment for week 1 took less time than the exercises -- I found the tasks straightforward. I didn't spend time removing parentheses from the pretty printer.

If you want to do some work, but not spend all the time working through the exercises, you could jump right to the assignment.

The beginning of Eval_Tests.hs in the week 1 assignment has definition of the Y combinator, which is always baffling, and then uses that definition to implement factorial. Very pretty.

The remaining weeks introduce basic functors, etc. and then monadic parsing. Then there's a digression into "free monads" in week 4. Troels said to ignore the free monad stuff, so I'm not sure what to make of it. Nothing really on stacks of monads for compilation.

Using Nix for Haskell

Install nix following: https://nixos.org/download/#download-nix

To enable flakes, add the following to your ~/.config/nix/nix.conf:

experimental-features = nix-command flakes

Even though the Haskell stack tool has nix integration, my approach is very basic. I just do nix develop at the command line, which runs this flake.nix script, which puts me at a shell with haskell tools available. From there, I start up an emacs, and haskell mode is able to find everything.

If, for example, you've cloned the ap-e2025-pub repo and want to use nix, you first have to copy the flake.nix file to the ap-e2025-pub top level folder, and then do git add flake.nix so that nix "sees" the file. Then you can do nix develop.

Emacs with eglot and haskell language server lsp-haskell

For some reason, I have not been able to get eglot + lsp-haskell working well for me. I can get the server to start, but errors from build are not reported, which is a real pain. I've even tried copying just, e.g. week1/handout to a folder under my ~/Downloads, thinking the issue is that there are too many separate projects (one per week), but that brute force approach did not seem to help.