AdventOfCode2020

Solutions for Advent of Code 2020 https://adventofcode.com/2020. Solutions are built on a Windows machine. Some workarounds are required to get certain languages to run. These are documented sparingly, but should give you enough hint to know how to approach it.

Install Python through Visual Studio or download from https://www.python.org. Alternatively, run it in the Windows Subsystem for Linux (WSL).

1. python aoc1.py
   Addends = 2 -> 942 + 1078 = 2020, 942 * 1078 = 1015476
   Addends = 3 -> 956 + 802 + 262 = 2020, 956 * 802 * 262 = 200878544

I ran this using node, but it should also be runnable on any modern browser.

2. node aoc2.js
   Valid passwords by counting: 418
   Valid passwords by positioning: 616

Install dotnet.script with choco install -y dotnet.script.

3. dotnet script aoc3.csx
   Number of trees for right 3, down 1 is 178
   Product of trees in the given sizes is 3492520200

Run it in WSL.

4. perl aoc4.pl
   Number of passports with required fields: 190
   Number of passports with validated fields: 121

Create a container using an image with Swift docker run --interactive --tty --name swift-latest swift:latest /bin/bash, then invoke the swift interpretor.

5. swift aoc5.swift
   Largest ID is 826, my seat ID is 678

Rust can be run like a script with cargo-script. Install with cargo install cargo-script. It needs to be compiled to a temporary directory before running. I'll probably only use this if I'm really CPU-bound or I needed to use a Rust library.

6. cargo script aoc6.rs
       Updating crates.io index
      Compiling aoc6 v0.1.0 (C:\???\Profit!\AppData\Local\Cargo\script-cache\file-aoc6-4f48aadee8a27792)
       Finished release [optimized] target(s) in 2.04s
   At-least-one responses: 7027, All-or-none responses: 3579

I ran this using SBT. It's awfully slow, 13s compared to the 2s for Rust (and I was already complaining). I really enjoyed the expressiveness of the Scala language though. Talk about a love-hate relationship.

7. sbt run
   [info] welcome to sbt 1.4.3 (Oracle Corporation Java 14.0.2)
   [info] loading project definition from C:???\Profit!\project
   [info] loading settings for project aoc7 from build.sbt ...
   [info] set current project to aoc7 (in build file:/C:/???/Profit!/)
   [info] compiling 1 Scala source to C:???\Profit!\target\scala-2.13\classes ...
   [info] running aoc7
   148 bags can contain a shiny gold bag.
   shiny gold bag contains 24867 other bags.
   [success] Total time: 13 s, completed Dec 8, 2020, 11:34:51 PM

Java has changed a lot since I last used it. During my time at Amazon we were stuck at Java 8. The first thing I realized though is the number of imports compared to other languages I've used so far. I've also used some of the newer Java features, such as (1) jShell REPL (JEP 222) from Java 9, (2) Local-variable Type Inference (JEP 286) from Java 10, (3) Switch Expressions (JEP 325) from Java 12, and my favorite (4) Single File Execution (JEP 330) from Java 11.

8. java aoc8.java
   Loop detected, breaking out! Accumulated value: 1446
   Repaired line 253 to nop -128
   Accumulated value: 1403

I haven't found a way to run Visual Basic.NET source code without compiling. When I started coding I was confused by namespaces like My.Computer.FileSystem. I ended up referencing aoc3.csx for usage of the .NET Core libraries.

9. dotnet run --project aoc9
   Found invalid value 22477624.
   Sum of smallest and largest of contiguous values is 2980044

I used WSL for Ruby. I spent the most time relative to other puzzles on this so far for the second star. I knew that I could generate the full combination and then count it, but that's going to be a O(n!) solution and take forever. I knew there must be a way to count some how, and is probably a dynamic programming solution. It wasn't until I pasted the first 10 combinations in Excel manually that I see the pattern.

10. ruby aoc10.rb
    product of 1-jolt diff and 3-jolt diff is 2664
    Number of combinations is 148098383347712

This took me the longest time to setup so far. I used Ubuntu and GCC. I had wanted to use LLVM/Clang to use the latest syntax, but it turned out that Ubuntu Bionic contained outdated versions. I tried compiling Clang but ran out of memory on my 8GB Microsoft Surface Book. I compiled on a bigger machine but when I tried to copy back to my Surface Book I ran out of space. To build on Ubuntu I only needed to install sudo apt-get install gnustep-devel.

11. gcc $(gnustep-config --objc-flags) -std=c11 -o aoc11 aoc11.m
    Number of occupied seats using immediate seats: 2310
    Number of occupied seats using nearest seats: 2074

I referenced the Kotlin documentation a little more than expected. It seemed to have taken a lot of hints from Scala.

12. kotlin aoc12.kts
    Manhattan distance for Direction navigation: 1838
    Manhattan distance for Waypoint navigation: 89936

I spent several days on this and probably wouldn't be able to solve this without spending several weeks studying the Chinese remainder theory. On top of this, I chose to practise this in Julia. I hate Julia. It's the worst language I've ever came across so far. It's super unintuitive.

i. 1-index arrays. Please read EWD831 by Edsger Dijkstra https://www.cs.utexas.edu/users/EWD/ewd08xx/EWD831.PDF

ii. Imperative syntax e.g. split(str, delim) instead of str.split(delim). This is due to lack of classes in Julia and it makes method chaining very clumsy. It also directly conflicts with their style guide to pass functions directly instead of lambdas: map(x -> parse(Int, x), value) cannot be written as map(parse, value). As of 1.5.3 there isn't a way to do line continuation! Imagine this: sort(map(x -> parse(Int, x), filter(x -> x != "x", split(str, delim))), rev=true)

iii. Methods like split(::String, ::String), keys(::Dict) don't return Array{String} or Set{String} as expected. Instead, they return a view to the underlying type i.e. Array{SubString{String}}, KeySet{Dict}, and and is incompatible with the regular data structures. This affects your method signature, and you can't call methods like sort(split(str, delim)).

iv. Integer division symbol: ÷. Bitwise XOR is ⊻. Checking for strict subset is ⊊. I can't even type these characters on my laptop keyboard. Some of these characters aren't even ASCII.

v. % is NOT the modulus operator. -37 % 35 returns -2, not 33. The modulus function is mod(::Int, ::Int)

vi. Variables declared in a local scope within a method is NOT visible to the rest of the method.

julia> x = 1
1
julia> for i in 1:10; x = i; end
julia> x
10 # Expected

julia> for i in 1:10; y = i; end
julia> y
ERROR: UndefVarError: y not defined # wtf!

I really hope there'll be a major rework in the language. It's super frustrating to work with such a badly design language. Credits to https://dev.to/qviper/advent-of-code-2020-python-solution-day-13-24k4 for referencing the algorithm.

13. julia aoc13.jl
    ID for earliest bus is 3246
    Time where schedules align: 1010182346291467

Finally got this out. It wasn't very difficult, I got the general design out on my first attempt. I did read the question wrong though and had to amend my code. Between Christmas celebration and debugging an issue at work, I was trying to squeeze this one in. I love writing in Go. It's concise and Rust-like. Not too many surprises, except I always forgot to declare and assign a new variable using :=, and that variables are passed by value by default. This manifested as strange bugs when I write to one of my []strings.Builders and it doesn't seemed to have any effect. I love to be able to use pointers when needed and the ability to use switch statements to replace if-elses.

14. go run aoc14.go
    Sum of masked values: 15403588588538
    Sum of mask-generated values: 3260587250457

My initial solution was printing values for each iteration and took a few hours, caused an OOM in Visual Studio Code. I thought this was one of the mathy kind of question, but I couldn't see the pattern. I read other people's solution and found that it's almost exactly like mine. I turned off printing and the code now runs in 2 seconds. It reinforces the fact that printing is an expensive operation.

15. dart run aoc15.dart
    2020th value for [0, 3, 6]: 436
    2020th value for [14, 8, 16, 0, 1, 17]: 240
    30000000th value for [14, 8, 16, 0, 1, 17]: 505

I struggled with nim, almost as much as I struggled with Julia. The toolchain is buggy. I did install choosenim to make it easier, but there seemed to be a problem with the compilation cache. Even though my syntax was correct, the compiler failed with weird errors such as "undeclared identifier" and "indentation error". I never found the root cause, but I worked around it by using choosenim to install a different version of nim. It was a bad experience and I don't think it's ready to be used for heavy production use at this time.

16. nim c -r aoc16.nim
    Ticket Scanning Error Rate: 29759
    Product of departure fields: 1307550234719