rDrama Advent of Code Day 10: Pipe ( | ) edition

i unfortunately missed this ine because i was getting drunk with my friends... oopsies. anyways still super ineberated and havent attwmpted it, i rlly ahte how this is scored because u have to be a nowlifer to get on the leaderboard :/

!codecels join the AoC hate train

wasted an ungodly amount of time reading stack traces with printed ascii grids and debugging because I zoned out and implemented "north, south, east, west" in one place as "up, down, left, right"

shan't doing part two, as I will be taking my own life, thank you for your understanding

I thought I had a solution to part 2 by just checking every cell and traversing until I hit an edge, but I didn't read the part that squeezing between pipes also counts and now I want to kill my elf

Ended up using pick's Theorem.

>finish puzzle

>look outside

>surprised the sun is up

Time flies when you're having fun

m←↑⊃⎕NGET'10.in'1 ⋄ s←⍸'S'=m
u←'|F7S' '-J7S' '|JLS' '-FLS' ⋄ v←u⌷⍨⊂3 4 1 2
j←(¯1 0)(0 1)(1 0)(0 ¯1) ⋄ k←(1 2)(2 3)(3 2)(2 1)
t←{⊂⍵}⌺3 3⊢m ⋄ w←(m∘∊¨v)∧¨u{∨/¨⍺∘∊¨(⊂⍵)∘⊃¨t}¨k
x←{l←1⊃⍵ ⋄ b←1↓2↑⍵ ⋄ ,∘⍵¨⊂¨b~⍨(⊂l)∘.+j/⍨l∘⌷¨w}
f←{⊃p⌷⍨1⍳⍨∧/¨s=⊃¨p←{⊃¨x¨⍵}⍣{s∊⊃¨⍺}⊣x s}
⎕←⌊2÷⍨≢p←f s

n←((¯1 0)(0 ¯1)) ((¯1 0)(0 1)) ((0 1)(1 0)) ((0 ¯1)(1 0)) ((0 ¯1)(0 1)) ((¯1 0)(1 0))
m←'.'@(c/⍨~p∊⍨c←,⍳⍴m)⊢('F7JL-|'⌷⍨n⍳⊂(⊂∘⍋⌷⊣)s-⊃¨(1∘↑,⍥⊂¯1∘↑)¯1↓1↓p)@s⊢m
⎕←+/∊('.'=m)∧{2|≢¨'L7|FJ'⎕R'|'¨'F7|LJ'⎕R''¨~∘'.-'¨,\⍵}⍣1⊢m

My part 1 was broken and couldn't figure it out, so i copy pasted like 10 solutions from the reddit solution thread that all gave me a wrong solution. And eventually I found the issue myself.

At the end the issue was that my F, moved exactly like the L. Otherwise it is just a simple dfs(and hoping that the longest contiguous part is the loop. But it would be easy to check if you reached S again. Also surrounded my entire pipe maze with one '.' in every direction, to not have to do bounds checking.

Part 2, whenever I find time later

Part 1 was straight forward path finding, just annoying to keep track of indexes. I need to take a break and think about part 2

I had to do a Point In Polygon algorithm for computer vision class once and frankly I don't like it and just looked it up again. Rederiving algorithms is for suckers when half the pythoncels are just importing and using it directly.

Probably should clean it up and write a more general function for up, down, left, right movement but I don't care it's sunday and nice out I'm going outside.

Good morning, I hate grids

I wasted maybe 8 hours on this today before starting from scratch and solving it in 10 minutes. My main mistakes:

1. Not appreciating that there's only one path to take out of each pipe.

2. Trying to do BFS for part 2 to fill the middle, writing some horrendous code for the tunnel parts.

I got up to 170 lines of code, realised "something's not right here" and cheated a little by reading a help thread on the subreddit. Once I understood the idea that you're just looking for the area inside a shape then it was fine. I used shapely which definitely felt like cheating but I just don't care.

I probably won't continue if the difficulty keeps ramping up. I have too much work to spend that much time on it every day, plus Nim doesn't have a great library ecosystem to solve these problems.

Part 1 was alright:

Part 2 was an abomination, half nim, half python, half Gimp.

1. Found the path in Nim and create a new input file with all non-path tiles zeroed out.

2. Used a python script to draw an image upscaled 3 times (140x3, 140x3), where each 3x3 pixels corresponds to a pipe pattern or a 3x3 red dot for non-pipes.

3. In Gimp bucket fill the outer wall, bucket fill the inner wall , select all pixels in inner wall, remove holes in selection (captures the non-pipes pixels), subtract a selection of the inner wall.

4. Read number of pixels in selection and divide by 9.

Prettier colours:

part 2 was written by the devil.

I feel like Yandev with this terrible code, but what I did was trace along the path while marking every orthogonal unoccupied square on the grid with either an L or R depending on whether it's on my relative left or right while I travel along the loop. Then I make the Ls and Rs spread out along orthogonal squares using BFS until there's no more unoccupied squares. From there, I can tell where L or R is the enclosed area because the other one will be touching the edge of the grid, and I can just count them to get the answer.

Part 2 was a pain in the arse today, but there are two ways I could imagine solving it:

1) Double the resolution of the input grid, so that the "squeezing" behaviour becomes visible and allows you to flood fill.

2) Consider the path as a polygon and each pixel as a point, then check which points are in the polygon. If you're on Python shapely is a fairly good library for this.

I went for the latter, which is a bit slow but far less memory intensive.

frick, my part 2 works only with the first example, not the next example where it's doubly enclosed. gonna kms later, bye girls.

I did not enjoy trying to figure out what constitutes an edge or a vertex in Part 2. This was a clever way on the puzzle designer's part to turn something that should have been relatively simple into a total god darn nightmare.

Code for both parts:

Your sponsor code doesn't work

mann I tried the 2nd part but my caveman method of just marking from the outside-in didn't work & I'm partial about making a maze algorithm rn

Snapshots:

https://carbon.vercel.app:

• ghostarchive.org

• archive.org

• archive.ph (click to archive)

https://adventofcode.com/2023:

• ghostarchive.org

• archive.org

• archive.ph (click to archive)