Oh, that's fun! (And looks like an easy way to lose track of a few hours as well...)

In case anybody hasn't seen it, the relevant Oglaf (NSFW)

Most people would use "word", "half-word", "quarter-word" etc, but the Anglophiles insist on "tuppit", "ternary piece", "span" and "chunk" (that's 5 bits, or 12 old bits).

Maybe it was due to attempting the puzzles in real-time for the first time, but it felt like there was quite a spike in difficulty this year. Day 5 (If You Give A Seed A Fertilizer) in particular was pretty tough for an early puzzle.

Day 8 (Haunted Wasteland), Day 20 (Pulse Propagation) and Day 21 (Step Counter) were (I felt) a bit mean due to hidden properties of the input data.

I particularly liked Day 6 (Wait For It), Day 14 (Parabolic Reflector Dish) and Day 24 (Never Tell Me The Odds), although that one made my brain hurt.

Day 25 (Snowverload) had me reading research papers, although in the end I stumbled across Karger's algorithm. That's the first time I've used a probabilistic approach. This solution in particular was very clever.

I learned the Shoelace formula and Pick's theorem this year, which will be very helpful to remember.

Perhaps I'll try using Prolog or J next year :)

Oh, just like day 11! I hadn't thought of that. I was initially about to try something similar by separating into rectangular regions, as in ear-clipping triangulation. But that would require a lot of iterating, and something about "polygon" and "walking the edges" went ping in my memory...

Haskell

Wasn't able to start on time today, but this was a fun one! Got to apply the two theorems I learned from somebody else's solution to Day 10.

import Data.Char
import Data.List

readInput :: String -> (Char, Int, String)
readInput s =
  let [d, n, c] = words s
   in (head d, read n, drop 2 $ init c)

boundary :: [(Char, Int)] -> [(Int, Int)]
boundary = scanl' step (0, 0)
  where
    step (x, y) (d, n) =
      let (dx, dy) = case d of
            'U' -> (0, 1)
            'D' -> (0, -1)
            'L' -> (-1, 0)
            'R' -> (1, 0)
       in (x + n * dx, y + n * dy)

area :: [(Char, Int)] -> Int
area steps =
  let a = -- shoelace formula
        (abs . (`quot` 2) . sum)
          . (zipWith (\(x, y) (x', y') -> x * y' - x' * y) <*> tail)
          $ boundary steps
   in a + 1 + sum (map snd steps) `quot` 2 -- Pick's theorem

part1, part2 :: [(Char, Int, String)] -> Int
part1 = area . map (\(d, n, _) -> (d, n))
part2 = area . map (\(_, _, c) -> decode c)
  where
    decode s = ("RDLU" !! digitToInt (last s), read $ "0x" ++ init s)

main = do
  input <- map readInput . lines <$> readFile "input18"
  print $ part1 input
  print $ part2 input

Clever! And removing constraints doesn't increase the path cost, so it won't be an overestimate.

Some (not very insightful or helpful) observations:

• The shortest path is likely to be mostly monotonic (it's quite hard for the "long way round" to be cost-effective), so the Manhattan distance is probably a good metric.
• The center of the puzzle is expensive, so the straight-line distance is probably not a good metric
• I'm pretty sure that the shortest route (for part one at least) can't self-intersect. Implementing this constraint is probably not going to speed things up, and there might be some pathological case where it's not true.

Not an optimization, but I suspect that a heuristic-based "reasonably good" path such as a human would take will be fairly close to optimal.

Yeah, finding a good way to represent the "last three moves" constraint was a really interesting twist. You beat me to it, anyway!

Haskell

Wowee, I took some wrong turns solving today's puzzle! After fixing some really inefficient pruning I ended up with a Dijkstra search that runs in 2.971s (for a less-than-impressive 124.782 l-s).

import Control.Monad
import Data.Array.Unboxed (UArray)
import qualified Data.Array.Unboxed as Array
import Data.Char
import qualified Data.HashSet as Set
import qualified Data.PQueue.Prio.Min as PQ

readInput :: String -> UArray (Int, Int) Int
readInput s =
  let rows = lines s
   in Array.amap digitToInt
        . Array.listArray ((1, 1), (length rows, length $ head rows))
        $ concat rows

walk :: (Int, Int) -> UArray (Int, Int) Int -> Int
walk (minStraight, maxStraight) grid = go Set.empty initPaths
  where
    initPaths = PQ.fromList [(0, ((1, 1), (d, 0))) | d <- [(0, 1), (1, 0)]]
    goal = snd $ Array.bounds grid
    go done paths =
      case PQ.minViewWithKey paths of
        Nothing -> error "no route"
        Just ((n, (p@(y, x), hist@((dy, dx), k))), rest)
          | p == goal && k >= minStraight -> n
          | (p, hist) `Set.member` done -> go done rest
          | otherwise ->
              let next = do
                    h'@((dy', dx'), _) <-
                      join
                        [ guard (k >= minStraight) >> [((dx, dy), 1), ((-dx, -dy), 1)],
                          guard (k < maxStraight) >> [((dy, dx), k + 1)]
                        ]
                    let p' = (y + dy', x + dx')
                    guard $ Array.inRange (Array.bounds grid) p'
                    return (n + grid Array.! p', (p', h'))
               in go (Set.insert (p, hist) done) $
                    (PQ.union rest . PQ.fromList) next

main = do
  input <- readInput <$> readFile "input17"
  print $ walk (0, 3) input
  print $ walk (4, 10) input

Haskell

A little slow (1.106s on my machine), but list operations made this really easy to write. I expect somebody more familiar with Haskell than me will be able to come up with a more elegant solution.

Nevertheless, 59th on the global leaderboard today! Woo!

import Data.List
import qualified Data.Map.Strict as Map
import Data.Semigroup

rotateL, rotateR, tiltW :: Endo [[Char]]
rotateL = Endo $ reverse . transpose
rotateR = Endo $ map reverse . transpose
tiltW = Endo $ map tiltRow
  where
    tiltRow xs =
      let (a, b) = break (== '#') xs
          (os, ds) = partition (== 'O') a
          rest = case b of
            ('#' : b') -> '#' : tiltRow b'
            [] -> []
       in os ++ ds ++ rest

load rows = sum $ map rowLoad rows
  where
    rowLoad = sum . map (length rows -) . elemIndices 'O'

lookupCycle xs i =
  let (o, p) = findCycle 0 Map.empty xs
   in xs !! if i < o then i else (i - o) `rem` p + o
  where
    findCycle i seen (x : xs) =
      case seen Map.!? x of
        Just j -> (j, i - j)
        Nothing -> findCycle (i + 1) (Map.insert x i seen) xs

main = do
  input <- lines <$> readFile "input14"
  print . load . appEndo (tiltW <> rotateL) $ input
  print $
    load $
      lookupCycle
        (iterate (appEndo $ stimes 4 (rotateR <> tiltW)) $ appEndo rotateL input)
        1000000000

42.028 line-seconds

Haskell

Phew! I struggled with this one. A lot of the code here is from my original approach, which cuts down the search space to plausible positions for each group. Unfortunately, that was still way too slow...

It took an embarrassingly long time to try memoizing the search (which made precomputing valid points far less important). Anyway, here it is!

{-# LANGUAGE LambdaCase #-}

import Control.Monad
import Control.Monad.State
import Data.List
import Data.List.Split
import Data.Map (Map)
import qualified Data.Map as Map
import Data.Maybe

readInput :: String -> ([Maybe Bool], [Int])
readInput s =
  let [a, b] = words s
   in ( map (\case '#' -> Just True; '.' -> Just False; '?' -> Nothing) a,
        map read $ splitOn "," b
      )

arrangements :: ([Maybe Bool], [Int]) -> Int
arrangements (pat, gs) = evalState (searchMemo 0 groups) Map.empty
  where
    len = length pat
    groups = zipWith startPoints gs $ zip minStarts maxStarts
      where
        minStarts = scanl (\a g -> a + g + 1) 0 $ init gs
        maxStarts = map (len -) $ scanr1 (\g a -> a + g + 1) gs
        startPoints g (a, b) =
          let ps = do
                (i, pat') <- zip [a .. b] $ tails $ drop a pat
                guard $
                  all (\(p, x) -> maybe True (== x) p) $
                    zip pat' $
                      replicate g True ++ [False]
                return i
           in (g, ps)
    clearableFrom i =
      fmap snd $
        listToMaybe $
          takeWhile ((<= i) . fst) $
            dropWhile ((< i) . snd) clearableRegions
      where
        clearableRegions =
          let go i [] = []
              go i pat =
                let (a, a') = span (/= Just True) pat
                    (b, c) = span (== Just True) a'
                 in (i, i + length a - 1) : go (i + length a + length b) c
           in go 0 pat
    searchMemo :: Int -> [(Int, [Int])] -> State (Map (Int, Int) Int) Int
    searchMemo i gs = do
      let k = (i, length gs)
      cached <- gets (Map.!? k)
      case cached of
        Just x -> return x
        Nothing -> do
          x <- search i gs
          modify (Map.insert k x)
          return x
    search i gs | i >= len = return $ if null gs then 1 else 0
    search i [] = return $
      case clearableFrom i of
        Just b | b == len - 1 -> 1
        _ -> 0
    search i ((g, ps) : gs) = do
      let maxP = maybe i (1 +) $ clearableFrom i
          ps' = takeWhile (<= maxP) $ dropWhile (< i) ps
      sum <$> mapM (\p -> let i' = p + g + 1 in searchMemo i' gs) ps'

expand (pat, gs) =
  (intercalate [Nothing] $ replicate 5 pat, concat $ replicate 5 gs)

main = do
  input <- map readInput . lines <$> readFile "input12"
  print $ sum $ map arrangements input
  print $ sum $ map (arrangements . expand) input

Haskell

This problem has a nice closed form solution, but brute force also works.

(My keyboard broke during part two. Yet another day off the bottom of the leaderboard...)

import Control.Monad
import Data.Bifunctor
import Data.List

readInput :: String -> [(Int, Int)]
readInput = map (\[t, d] -> (read t, read d)) . tail . transpose . map words . lines

-- Quadratic formula
wins :: (Int, Int) -> Int
wins (t, d) =
  let c = fromIntegral t / 2 :: Double
      h = sqrt (fromIntegral $ t * t - 4 * d) / 2
   in ceiling (c + h) - floor (c - h) - 1

main = do
  input <- readInput <$> readFile "input06"
  print $ product . map wins $ input
  print $ wins . join bimap (read . concatMap show) . unzip $ input

Torn between doing the problem backwards and implementing a more general case -- glad to know both approaches work out in the end!

Ah, nice! Dealing with each range individually makes things much simpler.

Haskell

Not hugely proud of this one; part one would have been easier if I'd spend more time reading the question and not started on an overly-general solution, and I lost a lot of time on part two to a missing a +. More haste, less speed, eh?

import Data.List
import Data.List.Split

readInput :: String -> ([Int], [(String, [(Int, Int, Int)])])
readInput s =
  let (seedsChunk : mapChunks) = splitOn [""] $ lines s
      seeds = map read $ tail $ words $ head seedsChunk
      maps = map readMapChunk mapChunks
   in (seeds, maps)
  where
    readMapChunk (title : rows) =
      let name = head $ words title
          entries = map ((\[a, b, c] -> (a, b, c)) . map read . words) rows
       in (name, entries)

part1 (seeds, maps) =
  let f = foldl1' (flip (.)) $ map (ref . snd) maps
   in minimum $ map f seeds
  where
    ref [] x = x
    ref ((a, b, c) : rest) x =
      let i = x - b
       in if i >= 0 && i < c
            then a + i
            else ref rest x

mapRange :: [(Int, Int, Int)] -> (Int, Int) -> [(Int, Int)]
mapRange entries (start, end) =
  go start $ sortOn (\(_, b, _) -> b) entries
  where
    go i [] = [(i, end)]
    go i es@((a, b, c) : rest)
      | i > end = []
      | b > end = go i []
      | b + c <= i = go i rest
      | i < b = (i, b - 1) : go b es
      | otherwise =
          let d = min (b + c - 1) end
           in (a + i - b, a + d - b) : go (d + 1) rest

part2 (seeds, maps) =
  let seedRanges = map (\[a, b] -> (a, a + b - 1)) $ chunksOf 2 seeds
   in minimum $ map fst $ foldl' (flip mapRanges) seedRanges $ map snd maps
  where
    mapRanges m = concatMap (mapRange m)

main = do
  input <- readInput <$> readFile "input05"
  print $ part1 input
  print $ part2 input

Not familiar with Lean4 but it looks like the same approach. High five!

Puzzles on the weekend are usually a bit more involved than weekdays. 23 is probably going to be a monster this year...

Haskell

11:39 -- I spent most of the time reading the scoring rules and (as usual) writing a parser...

import Control.Monad
import Data.Bifunctor
import Data.List

readCard :: String -> ([Int], [Int])
readCard =
  join bimap (map read) . second tail . break (== "|") . words . tail . dropWhile (/= ':')

countShared = length . uncurry intersect

part1 = sum . map ((\n -> if n > 0 then 2 ^ (n - 1) else 0) . countShared)

part2 = sum . foldr ((\n a -> 1 + sum (take n a) : a) . countShared) []

main = do
  input <- map readCard . lines <$> readFile "input04"
  print $ part1 input
  print $ part2 input

Haskell

A rather opaque parser, but much shorter than I could manage with Parsec.

import Data.Bifunctor
import Data.List.Split
import Data.Map.Strict (Map)
import qualified Data.Map.Strict as Map
import Data.Tuple

readGame :: String -> (Int, [Map String Int])
readGame = bimap (read . drop 5) (map readPull . splitOn "; " . drop 2) . break (== ':')
  where
    readPull = Map.fromList . map (swap . bimap read tail . break (== ' ')) . splitOn ", "

possibleWith limit = and . Map.intersectionWith (>=) limit

main = do
  games <- map (fmap (Map.unionsWith max) . readGame) . lines <$> readFile "input02"
  let limit = Map.fromList [("red", 12), ("green", 13), ("blue", 14)]
  print $ sum $ map fst $ filter (possibleWith limit . snd) games
  print $ sum $ map (product . snd) games