buster @axionbuster.com · 27/12/2024TIL a purely functional double-ended queue with worst-case constant time access to either end is possible and that listing monads explicitly in such a data structure can improve the time complexity of certain operations (such as list concatenation) by changing assoc. of (>>=) 130
buster @axionbuster.com · 23/12/2024Backpropagation is a functor arxiv.org/abs/1711.10455arxiv.orgBackprop as Functor: A compositional perspective on supervised learningA supervised learning algorithm searches over a set of functions $A \to B$ parametrised by a space $P$ to find the best approximation to some ideal function $f\colon A \to B$. It does this by taking e... 020
Reposted by busterfluffy_xenia @fluffyxenia.bsky.social · 21/12/2024@wunidz.bsky.social here ya go. Not the easiest on the eyes but it's what I could make drive.google.com/file/d/1udxh...drive.google.comresult.mp4 4135
buster @axionbuster.com · 22/12/2024i feel so sad when a fellow man says “there are no femcels,” because that means he’s not often had the experience of girls-you’ve-just-met trying to cling onto him and do weird “cutesy” things hoping it’ll actually make them look cute. i always wonder if/how they get bf at all 000
buster @axionbuster.com · 21/12/2024gist.github.com/axionbuster/...gist.github.comGrid ray marching with floating-point error compensationGrid ray marching with floating-point error compensation - Lib.hs 000
buster @axionbuster.com · 20/12/2024DDA line drawing algorithm, dimension-independent, lazy, infinite original algorithm is from Wikipedia (for 2D, in C) image: magnified image output, (0, 0) to (10, 5) 100
buster @axionbuster.com · 20/12/2024to build a list in the correct order in Haskell, choose one of three ways: 1. use a writer monad 2. build it in the reverse order, and then reverse it back 3. instead of using a list as an accumulator, use a function of type (a -> [a]) as the accumulator. 110
buster @axionbuster.com · 20/12/2024the usefulness of MultiWayIf in Haskell programming cannot be overstated (but you could hack it using case () of _ | condition1 -> expression1; _ | condition2 -> expression2; ...) 031
buster @axionbuster.com · 20/12/2024if you have an array with four (4) elements -- a product type -- then its index could be one of four (4) choices -- a sum type indexing is to logarithms as functions are to exponentials 000
Reposted by busterEcho @isthere.an.echo.inhere.orisitjust.me · 18/12/2024at santa-scale it probably has to be done by separate nodes in a distributed system 2211
buster @axionbuster.com · 20/12/2024there's a whole community on the internet accumulating knowledge about teaching, metathinking and problem solving amazing 100
buster @axionbuster.com · 20/12/2024when i was 22, i was rotting in bed when Twan van Laarhoven was 22, he published in a blog post an innovation that would be later named after himself ... van Laarhoven lenses also, scary to think that he's nearly 40 now time flies like a banana fruit flies like an arrow 010
Reposted by busterArnaud @pankzsoft.net · 18/12/2024If one has to use CPP in Haskell, doing it in the cabal file as a way to conditionally include modules instead of sprinkling #ifdef/#endif across the code-base feels much cleaner to me. A small amount of code duplication is better than slicing and dicing code to the point it becomes unreadable. 021
Reposted by busterHacker News Top 100 Bot (Unofficial) @hn100.bsky.social · 19/12/2024Haskell vs. Ada vs. C++ vs. an Experiment in Prototyping Productivity (1994) [pdf] Discussion 011
buster @axionbuster.com · 19/12/2024CRACKED fucking project github.com/hyperion-mc/...github.comGitHub - hyperion-mc/hyperion: 170,000 player capable Minecraft game engine built in Rust.170,000 player capable Minecraft game engine built in Rust. - hyperion-mc/hyperion 000
buster @axionbuster.com · 19/12/2024can you believe that this guy's Haskell implementation of QOI codec beats the C implementation? github.com/0xd34df00d/h...github.comGitHub - 0xd34df00d/hsqoi: Haskell implementation of the QOI image formatHaskell implementation of the QOI image format. Contribute to 0xd34df00d/hsqoi development by creating an account on GitHub. 020
buster @axionbuster.com · 18/12/2024github.com/axionbuster/...github.comGitHub - axionbuster/varcomp: Variadic function composition (without hacks)Variadic function composition (without hacks). Contribute to axionbuster/varcomp development by creating an account on GitHub. 000
buster @axionbuster.com · 17/12/2024Variadic Composition (uses the Text.Printf trick) (language: GHC2021, no other extensions) 010
buster @axionbuster.com · 17/12/2024can you find a solution to this problem? (or decide that there is not) 000
buster @axionbuster.com · 12/12/2024is floating point division by 0 defined in Haskell? i know it’s unconditionally UB on C/C++ 200
buster @axionbuster.com · 12/12/2024fun fact i learned. a box-box collision problem can be reduced to a ray-box collision problem 100
buster @axionbuster.com · 12/12/2024afraid of formatting and reinstalling my Windows box because uBlock Origin will be blocked on Chrome 000
Reposted by busterGraydon Hoare @graydon-pub.bsky.social · 02/12/2024Having real windows support out of the gate was a key strategic choice for Rust. It helped that Mozilla was shipping a browser there and knew the relative platform sizes in its bones. I don't like windows especially, but seriously buy a windows dev box and make your stuff work on windows, always. 514318
Reposted by busterbusiness goose🪿 @goose.art · 01/12/2024at my old job doing pentesting we had one client who we tested before and reported a finding about base64 not being secure which they claimed to have patched when we did a retest, we found that they just base64ed 3 times instead 🤦♂️ 5583
buster @axionbuster.com · 02/12/2024If only Haskell allowed ? as the last character of an identifier... Well now that would be a huge and confusing change 000
Reposted by busterJosiah @josiah.rs · 01/12/2024👨🏻💻🎄Day 1 Solution of #AdventOfCode in base R. Some may even call it the best R. 6613
buster @axionbuster.com · 01/12/2024#Haskell made many mistakes but also has some Godsend takes IO being a monad is one Transactions are computation that is never partially performed. Optimistic transactions try first and if it fails rolls back. But arbitrary IO can’t be rolled back. So they live in a distinct monad (STM) 020
buster @axionbuster.com · 01/12/2024I just completed "Historian Hysteria" - Day 1 - Advent of Code 2024 #AdventOfCode adventofcode.com/2024/day/1 220
buster @axionbuster.com · 01/12/2024Some brand names on Amazon look like randomly generated and yet manage to deliver good stuff 200
buster @axionbuster.com · 30/11/2024youtu.be/jkqCCSGjadk?...youtu.beUnknown Song "You're Going Down"YouTube video by x 200
buster @axionbuster.com · 30/11/2024Haskell Generics take such an ungodly amount of time to compile I'm going wild 000
Reposted by busterChris Penner @chrispenner.ca · 20/11/2024Did other #Haskell devs make it waaay too far into their career wondering what the #ActorModel fuss was about only to discover you'd been casually doing it in Haskell for years? It's so easy with STM and async that you don't need actor frameworks like other langs, so I never noticed I was doing it. 2142
Reposted by busterLambdaDuck @anka213.bsky.social · 20/11/2024stock deriving can’t derive the proper foldTree, but there are libraries that can. e.g. hackage.haskell.org/package/cata... and hackage.haskell.org/package/recu...hackage.haskell.orgcatamorphismExposes a Template Haskell function for generating catamorphisms. 241
buster @axionbuster.com · 30/11/2024Appalled at the number of people who think (entry-level) MacBooks are overpriced. Maybe they’re non-Americans so economics are somehow different, or all they do is surf the net (and literally nothing else), or they just drop $3k on premium laptops, or must run things like SolidWorks…(?) 010
![-- | the DDA line drawing algorithm
-- (dimension-independent, lazy, infinite)
--
-- you can generate an infinite ray by
-- ignoring the number of steps
--
-- algorithm is due to Wikipedia
dda ::
forall f i s.
(Traversable f, MonadZip f, Integral i) =>
-- | The starting point
f i ->
-- | The ending point
f i ->
-- | The number of steps (points) to get to the end
-- and the points
ST s (i, [f i])
dda start end = do
let fi = fromIntegral :: i -> Double
difference = liftM2 (-) end start
steps = maximum $ abs <$> difference
step = ((/ fi steps) . fi) <$> difference
v <- newSTRef `traverse` fmap fi start
(succ steps,) <$> fix \loop -> do
v' <- traverse readSTRef v
forM_ (mzip v step) \(r, s) -> modifySTRef r (+ s)
(fmap round v' :) <$> loop -- lazy list](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:yljaphrlqlnwolprgrrkr5yp/bafkreiewnazcxo4362h4nyiattjuhs5bvpkbprlf4fyxh42oerahn5gv7a)
![prog :: IO ()
prog = runEff . runLog . runSpan $ do
span "example program" do
log "foo"
span "greeting" do
log "hello"
log "world"
log "bar"
>>> prog
[Start span 'example program']
[LOG] foo
[Start span 'greeting']
[LOG] hello
[LOG] world
[End span 'greeting']
[LOG] bar
[End span 'example program']](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:yljaphrlqlnwolprgrrkr5yp/bafkreicmxzs7byhahvxobsyn55p73fwf45k6qckhs47x5cbgnp2vj7iiaq)
![-- | a mysterious module that apparently samples points on the face of a coid
module Face (facepoints, dbgcountfacepoints, dbgdesmos) where
import Control.Lens hiding (index)
import Data.Functor.Rep
import Data.Ix (range)
import Data.List (intercalate)
import Linear.V3
import Text.Printf
-- | retrieve the points on the faces of a cuboid
-- normal to a certain direction (sig: {-1, 0, 1}) from said cuboid
facepoints :: V3 Int -> V3 Int -> [V3 Int]
facepoints coid sig =
let (!) = index
-- p, q, r: dimensions
-- ps: list of points to sample (?)
-- h: permutation of a V3
rule p q r ps h
| sig ! p == 0 = []
| otherwise =
[ let a = if i >= coid ! r then coid ! r else i
b = if j >= coid ! q then coid ! q else j
c = if sig ! p < 0 then 0 else coid ! p
in h do V3 a b c
| (i, j) <- ps
]
-- this thing is a total mystery to me
st i -- compute starting / ending points for the face (?) in dimension i
| sig ! i < 0 = (1, 0)
| sig ! i > 0 = (0, 1)
| otherwise = (0, 0)
h0 (V3 a b c) = V3 c b a
h1 (V3 a b c) = V3 b c a
h2 (V3 a b c) = V3 b a c
p0 =
let (sj, ej) = st ey
(si, ei) = st ez
in range ((si, sj), (coid ! ez - ei, coid ! ey - ej))
p1 =
let (sj, ej) = st ez
in range ((sj, 0), (coid ! ez - ej, coid ! ex))
p2 = range ((0, 0), (coid ! ey, coid ! ex))
in concat
[ rule ex ey ez p0 h0,
rule ey ex ez p1 h1,
rule ez ex ey p2 h2
]
-- | print a list of vectors in a format that can be copy-pasted into Desmos
dbgdesmos :: (Show a) => [V3 a] -> String
dbgdesmos vs = "[" ++ intercalate "," (f <$> vs) ++ "]"
where
f (V3 x y z) = printf "(%s,%s,%s)" (show x) (show y) (show z)
-- | predict the number of points sampled by 'facepoints'
dbgcountfacepoints :: V3 Int -> V3 Int -> Int
dbgcountfacepoints ((+ pu…](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:yljaphrlqlnwolprgrrkr5yp/bafkreidovanjhfb3ijpewimigsecbe2kfm5pbcpo2z5ordtu76rixtjnjy)
![extrapoints :: STRef s (Seq IP) <- new Seq.empty -- right-to-left, FIFO
out :: STRef s [IP] <- new []
callCC \break_ -> do
let break = break_ ()
fix \continue -> do
let tell = write out
read extrapoints >>= \case
a :<| points -> do
tell [a]
if sameblock a end
then break
else continue
_ -> pure ()
-- todo...
break](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:yljaphrlqlnwolprgrrkr5yp/bafkreic4a6mgfwt6zcyjvpouwxy3dhsmaz4homfyegem2it2rkmhmqio2q)
