Andres Löh

I am a partner and Haskell Consultant at Well-Typed LLP. I am also a member of the Haskell Foundation Board, and a co-host on the Haskell Interlude podcast and the Haskell Unfolder YouTube series.

Posting mostly about Haskell, but occasionally also about books I'm reading, video games I'm playing, and other things that interest me.

Public Key

npub1q8tuxwdxkr54fv08qrs2yja0e3u0tyju69qje6ht9fldlagre8dqwymqnn

NIP-05 Address

kosmikus@functional-cafe.mostr.pub

Profile Code

nprofile1qqsqr47r8xntp625k8nspc9zfwhuc784jfwdzsfvat4j5lkl75punkspzemhxue69uhhyetvv9ujumn0wd68ytnzv9hxg8aj3j3

Publishing to

relay.nostr.band

Author Public Key

npub1q8tuxwdxkr54fv08qrs2yja0e3u0tyju69qje6ht9fldlagre8dqwymqnn

Show more details

Last Notes

2025-05-14 14:40:47

If you want to hear from Edsko and me how to test stateful APIs using quickcheck(-lockstep), join us today, 2025-05-14, at 1830 UTC, live on YouTube for a new episode of the #Haskell #Unfolder: https://www.youtube.com/watch?v=RPQVsaZ46Dk&list=PLD8gywOEY4HaG5VSrKVnHxCptlJv2GAn7&index=44

2025-05-07 21:52:27

- reply

@nprofile…qlzx At ZuriHac, you can just talk to people all day, or sit near the lake and enjoy the surroundings, or listen to talks, or ... There's no need to do any hacking, but of course the option to. Hope you decide to go!

2025-05-07 13:14:47

Today, there will be a new #Haskell #Unfolder episode, where Edsko and I will talk about the monomorphism restriction and defaulting. Join us live on YouTube 2025-05-07 at 1830 UTC (or watch later) via https://www.youtube.com/watch?v=nD93M2qlwMk&list=PLD8gywOEY4HaG5VSrKVnHxCptlJv2GAn7&index=43

2025-04-15 14:46:08

- reply

@nprofile…rs4v This is tricky because it mixes behaviour specific to GHCi with the Haskell IO semantics. Typing `x <- action` in GHCi executes the action and binds its result to `x` for further use. As the type of `putStrLn "hello"` is `IO ()`, the result of that action is always `()`.
The GHCi syntax is reminiscent of the syntax used in do blocks, where `do x <- action1; somethingElse x` gets desugared to `action1 >>= \ x -> somethingElse x`.

2025-04-15 14:43:09

The #Haskell #Unfolder is two years old! Join us for a new episode tomorrow, 2025-04-16, at 1830 UTC, live on YouTube. This time, the topic is logic programming with typedKanren, which is a typed version of #miniKanren embedded into Haskell.
https://www.youtube.com/watch?v=ukk5JDRYQSc&list=PLD8gywOEY4HaG5VSrKVnHxCptlJv2GAn7&index=42

2025-03-26 15:13:33

Tonight, 2025-03-26, at 1930 UTC (still an unusual time for US/Canada) there'll be a new episode of the #Haskell #Unfolder, and after all this time, we'll finally talk about generics! https://www.youtube.com/watch?v=db-KjKOBMeA&list=PLD8gywOEY4HaG5VSrKVnHxCptlJv2GAn7&index=41

2025-03-26 08:55:53

One of the best things about Xenoblade Chronicles X getting re-published is that there's so much of the X soundtrack on the internet again, which is honestly one of my favourite video game soundtracks of all time.

2025-03-11 19:09:16

Tomorrow, 2025-03-12, at 1930 UTC (note: one hour later than usual in Canada / US due to asynchronous DST shifts), there'll be a new episode of the #Haskell #Unfolder, titled "understanding through a model". We'll talk about QuickCheck and convolutions ...
https://www.youtube.com/watch?v=0QTt2W7CVnA&list=PLD8gywOEY4HaG5VSrKVnHxCptlJv2GAn7&index=40

2025-02-25 17:10:01

Please consider nominating yourself to be a member of the #Haskell Foundation Board: https://discourse.haskell.org/t/2025-call-for-nominations-for-the-haskell-foundation/11373

2025-02-25 17:10:01

Please consider nominating yourself to be a member of the #Haskell Foundation Board: https://discourse.haskell.org/t/2025-call-for-nominations-for-the-haskell-foundation/11373

2025-02-11 20:41:49

Tomorrow, 2025-02-11, we'll live-stream a new episode of the #Haskell #Unfolder. Edsko and I will discuss the various strategies for deriving instances that GHC offers and their relative strengths and weaknesses. Hope to see you all there! https://www.youtube.com/watch?v=NEUbp2CsuNg&list=PLD8gywOEY4HaG5VSrKVnHxCptlJv2GAn7&index=39

2025-02-11 20:41:49

2025-01-14 15:58:45

The #Haskell #Unfolder is back! In tomorrow's episode, we'll take a look at CUDA and write some simple functions in it. We'll also show how to interface from Haskell to CUDA and how to use QuickCheck to test CUDA functions from Haskell.
https://www.youtube.com/watch?v=ww4C3jlsesw&list=PLD8gywOEY4HaG5VSrKVnHxCptlJv2GAn7&index=38

2024-12-14 07:12:07

I wonder if today's (14) #AdventOfCode task was specifically designed to be difficult for AI to solve due to vague specification, or if that's actually a problem that AI is particularly good at.
I was first slightly annoyed at having to guess, but then I got it with my first solution attempt (in #haskell), so I'm happy enough.

2024-12-12 14:06:52

- reply

@nprofile…mclj I think `iterate` and then `(!!)` should be perfectly fine. I don't see where any performance problem should come from. I would avoid `take` and `last` though.

2024-12-04 11:00:48

- reply

@nprofile…fy0y Yes, Elm is certainly great when it comes to error messages.
I'd really like if we made a push for significantly improving the GHC error messages.
Helium has comparatively good errors for Haskell (https://github.com/Helium4Haskell/helium), but it's hardly alive and never was usable for anything but teaching due to various constraints in the language and no library support.

2024-12-04 10:16:53

- reply

@nprofile…fy0y I wouldn't say they're *totally* irrelevant. But extracting the important pieces of information out of it is a bit of an art that one has to learn, definitely.
I think it's widely agreed that the error message format isn't great, but it's one of those topics where as soon as you're trying to change anything, there's endless bikeshedding.
Despite all this, recent versions of GHC have some improvements.
(1) There's error codes now, and an error message index at https://errors.haskell.org/ where error messages are explained, together with examples and some advice on how to read them.
(2) There's an ongoing move to a more structured error message format *within* GHC, which should make it easier to systematically change the output format via flags or postprocessors etc once this is complete.

2024-12-03 19:55:09

Tomorrow, 2024-12-04, at 1930 UTC, there'll be a new episode of the #Haskell #Unfolder live on YouTube. This time, we will try to solve Day 4 of #AdventOfCode 2024.
https://www.youtube.com/watch?v=KkKxHApYSfc&list=PLD8gywOEY4HaG5VSrKVnHxCptlJv2GAn7&index=37

2024-11-25 22:14:49

- reply

@nprofile…7q8n Hmm. Strange. I think column numbers should usually match the editor (but it is possible I am missing something, I haven't been thinking about this in years). Do you have an example where this is wrong?

2024-11-25 21:21:38

- reply

@nprofile…7q8n Regarding the `aligncolumn` question, the answer is that the default column type isn't just `l`, so the replaced one isn't just `c` either. This could really use better abstractions, but e.g. with the default `poly` settings, something like the following should work:
```
\aligncolumn{2}{@@{}>{\hspre}c<{\hspost}@@{}}
```
(The double @s are only needed if you're doing this within a file processed by lhs2TeX, to escape its @-processing.) If you need this often, it's probably better to define a `\centercolumn` macro that expands to this.

2024-11-25 21:18:25

- reply

@nprofile…7q8n You can generally not expect to keep your code to be valid Haskell directly. The way to go is to extract the Haskell code using `--newcode` mode just as you're extracting LaTeX code using `--poly` mode. If you've never done this before, this will require wrapping lots of your `%format` directives in conditionals, but e.g. `spacing.fmt` does the right thing, and `^` in `--newcode` mode formats as nothing.

2024-11-25 21:07:45

- reply

@nprofile…7q8n I'd normally do something like
```
%include spacing.fmt
...
> ( a -> b ) -> a -> c
> bar x -> biz y ^
```

2024-11-19 15:04:15

Tomorrow, 2024-11-20, at 1930 UTC, there'll be a new episode of the #Haskell #Unfolder, on concurrency and the FFI. Edsko will explain to me what exactly the difference is between "safe" and "unsafe" FFI calls and whether safe calls are actually safe ...
https://www.youtube.com/watch?v=IMrBTx7aYjs&list=PLD8gywOEY4HaG5VSrKVnHxCptlJv2GAn7&index=36

2024-11-07 18:18:43

- reply

@npub1ky2…ghvc I've certainly come around on this and use `pure` mostly now. I don't think `pure` is ideal, because the notion of purity is subjective and local. But `return` was called this way to mimic imperative code, and often creates entirely the wrong intuition in particular with beginners coming from other languages, in that it actually returns from something and affects control flow. The name `pure` doesn't come with this baggage.

2024-11-05 20:38:26

Tomorow, 2024-11-06, at 1930 UTC, Edsko and I will stream a new episode of the #Haskell #Unfolder. This time, we're going to take a look at two type classes in the Haskell world that are perhaps less widely known: Distributive and Representable.
Join us live on YouTube: https://www.youtube.com/watch?v=g_vKOg0LdlI&list=PLD8gywOEY4HaG5VSrKVnHxCptlJv2GAn7&index=35

2024-11-05 20:36:41

In the latest episode of the #Haskell #Interlude podcast, @npub1zw4…j72m and I are talking to Gabriele Keller. I've known Gabriele for a long time, mostly from afar, because for the longest time I've been in the Haskell world she was in Australia. But now she's professor at the group at Utrecht University where I did my PhD, so this brings back many fond memories.
https://haskell.foundation/podcast/57/

2024-11-05 07:00:40

- reply

@npub1j34…njzs Heh, I guess you're talking about NPlusKPatterns? I would phrase it a bit differently. It makes sense from the compiler perspective.They did not want / couldn't just remove the feature, so it had to become optional in the same way other extensions are optional, so it made sense to make it into an extension that's then "discouraged". And of course, there was no clear order between the standards effort and the compiler changes, so the extension terminology may have ended up elsewhere.

2024-10-14 19:26:27

On Wednesday, 2024-10-16, at 1830 UTC, it's time for the #Haskell #Unfolder once again: Edsko and I will have a closer look at monads and discuss so-called "Kleisli composition", i.e., composition of monadic functions, and see how it relates to "normal" function composition, and also to the monad laws. https://www.youtube.com/watch?v=n5ZtsHrYWq0&list=PLD8gywOEY4HaG5VSrKVnHxCptlJv2GAn7&index=34

2024-10-09 13:15:36

We've published a retrospective of the first 18 months of making the #Haskell #Unfolder on the Well-Typed blog: https://well-typed.com/blog/2024/10/18-months-of-unfolder/

2024-10-01 14:50:11

In our new #Haskell #Unfolder episode (Wednesday, 2024-10-02, live on YouTube), we will discuss @npub1wrp…97pa 's diagrams package.
https://www.youtube.com/watch?v=iX9ShJG5L0A&list=PLD8gywOEY4HaG5VSrKVnHxCptlJv2GAn7&index=33
As promised in the previous episode, we will show how to visualise the tic-tac-toe game tree in order to produce pictures similar to https://xkcd.com/832/

2024-08-14 18:06:07

The #Haskell #Unfolder episode on the nothunks package starts in about 20 minutes: https://www.youtube.com/watch?v=NTW62s3mrXQ&list=PLD8gywOEY4HaG5VSrKVnHxCptlJv2GAn7&index=31

2024-08-12 21:26:00

In this week's episode of the #Haskell #Unfolder, Edsko and I will discuss the nothunks library, which can be used to systematically test for the presence of (unexpected) thunks and thereby hopefully identify or prevent some space leaks. Join us on YouTube, 2024-08-14 at 1830 UTC: https://www.youtube.com/watch?v=NTW62s3mrXQ&list=PLD8gywOEY4HaG5VSrKVnHxCptlJv2GAn7&index=31

2024-08-12 21:26:00

2024-07-30 14:28:37

Tomorrow, Edsko and I will talk about (run)ST in the 30th episode of the #Haskell #Unfolder. 2024-07-31 at 1830 UTC live on YouTube: https://www.youtube.com/watch?v=KNXi5CrrjKg&list=PLD8gywOEY4HaG5VSrKVnHxCptlJv2GAn7&index=30

2024-07-17 16:40:58

In tonight's episode of the #Haskell #Unfolder, Edsko is going to explain to me how exceptions, and in particular the new annotations of exceptions, work in GHC 9.10. I'm sure I'll learn a lot, and we of course invite you all to join us live on YouTube.
https://www.youtube.com/watch?v=SwOkh9N41Y8&list=PLD8gywOEY4HaG5VSrKVnHxCptlJv2GAn7&index=29

2024-07-05 11:27:33

I've just released the final part of my #Haskell Introduction course, on Monads, at https://teaching.well-typed.com/intro/monads.html
This means the complete course is now available. Hope you enjoy it.

2024-07-02 17:51:07

In tomorrow's episode of the #Haskell #Unfolder, Edsko and I will explain how overlapping instances can be "encoded" using type families. Join us live 2024-07-03 at 1830 UTC.
https://www.youtube.com/watch?v=gN6zIfkdlxE&list=PLD8gywOEY4HaG5VSrKVnHxCptlJv2GAn7&index=28

2024-06-21 15:54:18

Part 4 of my / Well-Typed's video-based #Haskell Introduction course is now available, discussing Parametric Polymorphism as well as Overloading: https://teaching.well-typed.com/intro/polymorphism-and-overloading.html

2024-06-13 14:05:04

More on GHC specialisation by my colleague Finley: https://well-typed.com/blog/2024/06/choreographing-specialization-pt2/ #haskell

2024-06-11 17:03:25

Join us tomorrow, 2024-06-12, at 1830 UTC, for the 27th episode of the #Haskell #Unfolder. Edsko and I will discuss "duality" as a design principle.
Note that this episode will focus on practice, not theory, and should be understandable to experts as well as (relative) beginners.
https://www.youtube.com/watch?v=XXw9DsDzonk&list=PLD8gywOEY4HaG5VSrKVnHxCptlJv2GAn7&index=27

2024-05-15 13:23:31

Later today, 2024-05-15, at 1830 UTC, join Edsko and me for the 25th episode of the #Haskell #Unfolder "from #Java to Haskell" live on #youtube.
https://well-typed.com/blog/2024/05/haskell-unfolder-episode-25-from-java-to-haskell/

2024-04-13 15:30:15

It's finally time for the next episode of the #Haskell #Unfolder. In episode 23, my colleagues Finley, Edsko, and I will discuss the topic of "specialisation", an important GHC optimisation that can help eliminate the cost intrinsic to overloaded functions. This will be streamed live on YouTube, on Tuesday, 2024-04-16, at 1830 UTC. https://www.youtube.com/watch?v=ksW04Cl2dgo&list=PLD8gywOEY4HaG5VSrKVnHxCptlJv2GAn7&index=23