Attempting this year with rust
⭐️ Spoilers lurk below ⭐️
--- Day 1: Trebuchet?! ---
Read through the first day of 'Comprehensive Rust' and jumped straight into sliding windows of strings... so a much harder opener than other years - punishing several naive approaches that would normally slide for a few days.
Final code is not pretty, with some gross labelled loop...break blocks.
--- Day 2: Cube Conundrum ---
Used today's puzzles to try out some of the functional features of iterators and am fairly happy with the result. The top-level functions are nice and readable. The inner workings of the input parsing: less so.
--- Day 3: Gear Ratios ---
Enjoyed day 3, but didn't get to it until the 4th. Solution is fairly clean, but didn't get a huge amount of reuse between parts 1 and 2.
--- Day 4: Scratchcards ---
Lots of parsing again today - tackled it in one functional chain which wasn't very pretty. Might reach for regexp next time there's something similar. Part 2 sounded like it was drifting towards some nasty recursion, but with the given limits was fairly straightforward.
--- Day 5: If You Give A Seed A Fertilizer ---
Once I stopped trying to store billions of ints in a HashMap, part 1 suddenly became very doable... funny that. The rust compiler is doing all of the heavy lifting for me in part 2. My solution is just a brute force through all seeds. Unoptimised builds wouldn't even churn through the 1st range, but optimised did the whole thing in ~2 minutes. I think I figured out the nicer approach to take, but won't refactor it unless I catch back up with the other days.
--- Day 6: Wait For It ---
Got up a bit earlier for today. Could have been a lot faster by not bothering to parse the input, but did so anyway in the interests of keeping the inputs secret and being a bit too purist...
Realised while doing part 2 that the twist was 64-bit ints and enjoyed barrelling and just changing some u32s to u64s. I imagine there's a more elegant maths solution as well.
--- Day 7: Camel Cards ---
A lot of upfront logic in building the Hand struct and determining HandType derivation. Learnt a bit about #[derive ... ] for the enum and writing a custom PartialOrd function for the ranking. I vey much shoehorned in part 2 by passing a flag all the way through to the core logic but it worked well.
--- Day 8: Haunted Wasteland ---
Nice and easy first part with HashMap. Then let an optimised brute force for part 2 run for fun while trying to prove that the sequences looped nicely and implementing LCM (via GCD with the Euclidean algorithm which I definitely did not remember). It got ~0.02% there.
--- Day 9: Mirage Maintenance ---
Pretty simple one for day 09. Repeatedly called a small function to reduce the sequences to zeroes, and picked up on summing the final elements for part 1 quickly. I don't have an intuitive understanding for why alternately adding and subtracting works for part 2 - found it by quick trial and error.
--- Day 10: Pipe Maze ---
Could have done part 1 a fair bit faster, but ended up fighting a bunch of spaghetti code that couldn't decide if it was using positional offsets or directional enums - made more complicated by a dodgy application of a Tile enum. Came back to part 2 in the afternoon with a fairly good idea of what the heuristic for an enclosed tile would be. Spent a long time debugging why the new example wouldn't even get through the part 1 solver before realising the input wasn't square...
--- Day 11: Cosmic Expansion ---
Thought today was going to be based on cellular automata at first; still enjoyed both parts. The part 2 solution evaded me due to a calculation error with how I was expanding the space. Tidied up and DRYed out before pushing.
--- Day 12: Hot Springs ---
Only done part 1 so far. Solution is well organised and fairly readable, but unoptimised both in terms of generating permutations that are trivially invalid and by doing a lot of Vec cloning where I imagine proper knowledge of Rust would point to a better approach. Will have to think on how to reduce part 2, given that for some of the more unknown records, the number of permutations is just absurdly large (~2^79).
--- Day 13: Point of Incidence ---
This solution isn't very pretty, lots of duplication and a fairly gross bit of logic in part 2 to determine whether the lines of reflection are new or old, instead of just having the initial results to compare against. Learnt a bit about Result<T, Err>. Also the reflection finding is pretty naive - likely a better way to do this.
I had mostly caught up by the 13th, but quickly fell even further behind. Time to try and catch up a bit...
--- Day 14: Parabolic Reflector Dish ---
Enjoyed this one, not too difficult a concept, so just had to go through the motions of being able to 'roll' in each direction. Kept track of seen states for the loop detection needed for part 2 using strings of the entire state which is not going to be space efficient, but hey.
--- Day 15: Lens Library ---
One of my favourites so far. Nothing too taxing, bar a very in-depth problem, but I enjoyed the story and kept a fairly clean implementation. Is it bad that I didn't use a HashMap? 😄
--- Day 16: The Floor Will Be Lava ---
This felt a bit like mapping the cube from last year, but not quite as bad. Lots of manually setting up mappings from beam directions and tile types to new beam directions. Thankfully had considered looping beams pretty early on. Nothing fancy for part 2, just tried each entry point. As each tile & beam direction combo is completely deterministic, it should be possible to memoise matching parts of beam paths from different entry points.
Managed 8 stars today (23rd), not too bad.