Micah Cantor

Haskell’s error messages are horrific. Their language is mostly aimed at a very academic audience and is one of the main reasons the language is so inaccessible.

Racket's slideshow package is fun

docs.racket-lang.org/slideshow/

Analogous in a way to cursive handwriting

Everyone gets bidirectional BFS wrong People really need to stop blindly copying code from the Internet.

What happens when LLMs train on code that's popular but wrong? zdimension.fr/everyone-get... explores this, with interactive diagrams exploring the error

Abstract syntax trees but they’re actually made out of…published abstracts?

Probably you should use vector (hackage.haskell.org/package/vector) or Text.

And if you need efficient add/delete then maybe IntMap (hackage.haskell.org/package/cont...)

hackage.haskell.org/package/hash...

Emacs lisp would be another example, though that language is uglier than scheme

GNU guix basically takes this to the logical extreme. It’s a whole OS that is configurable and scriptable in Scheme.

There is mutation via set! in Scheme but it’s easily avoidable and not idiomatic.

But otherwise it seems to meet your criteria. Easy to parse, human-readable, embeddable, interpretable.

Is this just Scheme?

(Or replace unit with the fields of the type you’re creating)

It’s annoying, but can be alleviated by following the pattern of wrapping *all* types in a module, and then exposing a function create : unit -> t from that module.

Particularly important for record types because otherwise type inference sucks ass

For all the excitement about RAG in the LLM space, enriching the context of code generation with static analysis (whether it be something like control-flow or typechecking) oddly seems to be relatively unexplored so far.

I didn't realize until recently that iCloud lets you use a custom domain for email for $1/month. Good stuff. Compare that to $6/month for 1 user on Google workspace for a custom domain on gmail.

Yikes. Somehow one of the only language tooling this makes more sense for 🙃

I’m assuming Haskell? Are you using different ghc versions for the LSP and cabal/stack?

Day 5 #AdventOfCode

I'm a little behind now, but here's day 4! #AdventOfCode

I’ve thought a bit about syntax sugar that turned method syntax into module qualified syntax. In OCaml terms, suppose it was required that every module exposed a type t.

Then you could have lst : t and turn lst.map(f) into List.map(lst, f)

Basically UFCS but with a standard module structure

No it doesn’t. My comment wasn’t very clear but I meant that since you don’t need to create a new file to created a namespaced modules in OCaml, it’s easier to avoid circular imports by just putting code that uses the same types in the same file (but still different modules)

I admit it’s ugly but I have done in Haskell to avoid cyclic imports, mainly when I wanted to annotate functions using the same types but otherwise separate functionality into separate files. It’s part of why I much prefer OCamls modules.

Merge wasm_of_ocaml by OlivierNicole · Pull Request #1724 · ocsigen/js_of_ocaml This aims to merge wasm_of_ocaml, currently hosted at https://github.com/ocaml-wasm/wasm_of_ocaml, back into this repo. This is pending on Dune support ocaml/dune#11029. For now, Dune is pinned for...

wasm_of_ocaml (a compiler of OCaml to WASM, and a fork of the JS compiler js_of_ocaml) has been merged. Congrats to all that contributed!

github.com/ocsigen/js_o...

The real tragedy of missing the Lowe Post is I haven’t gotten to hear the Thibs impression

Day 3 #AdventOfCode

Bicameral, Not Homoiconic Parenthetically Speaking: Articles by Shriram Krishnamurthi

Sometimes it takes me 22 years (+ one evening) to write a blog post. Here are my thoughts on "homoiconicity" and, as an alternative, "bicameral syntax". (Warning: 4000 words.)
parentheticallyspeaking.org/articles/bic...

I've been waiting for this since @lexi-lambda.bsky.social launched her own campaign against homoiconic a while ago

Day 2 #AdventOfCode

Racket should have reduce in racket/list !

#lang racket/base (require racket/port) (require racket/string) (require racket/list) (require racket/hash) (define (parse filename) (define lines (port->lines (open-input-file filename) #:close? #t)) (define input-pairs (for/list ([line lines]) (map string->number (string-split line " ")))) (define left (map car input-pairs)) (define right (map cadr input-pairs)) (values left right)) (define (solve-part-1 left right) (define left-sorted (sort left <=)) (define right-sorted (sort right <=)) (apply + (map abs (map - left-sorted right-sorted)))) (define (solve-part-2 left right) (define right-appearances (make-hash)) (for ([x right]) (hash-update! right-appearances x add1 0)) (define similarity-scores (for/list ([x left]) (* x (hash-ref right-appearances x 0)))) (apply + similarity-scores)) (define-values (left right) (parse "input.txt")) (solve-part-1 left right) (solve-part-2 left right)

Day 1 #AdventOfCode

GitHub - micahcantor/advent-of-code-2024 Contribute to micahcantor/advent-of-code-2024 development by creating an account on GitHub.

Doing Advent of Code this year in Racket. I may not finish them all, but should be fun!

github.com/micahcantor/...