Advent of Code 2024

Well, here we Go!

... Sorry. I'll get my coat.

Day 1

Another year, another set of puzzles, another ludicrous README, another new programming language.

Typical Day 1 puzzle, easing you in oh so gently, the calm before the storm. Spent most of my time setting up a framework for running the code. Go is a curious beast. Considerably less painful than Rust to learn, yet missing a lot of its pretty features. Syntax somewhere between C and Python. Error handling seems very verbose (i.e. if err != nil everywhere). For this puzzle I wanted to zip() a couple of lists together a la Python, but a zip equivalent doesn't exist in the standard library, so I've got a C-style for loop over indices, which feels a bit old-skool, but whatever. Reading lines from a file was a bit funky. Open a file, handle failure, create a scanner, scan the lines. I dunno, it all feels a bit like teaching my parents how the internet works. A lot of words, seemingly unnecessary repetition, not really sure if they (or I) have really understood it all.

Still, 2 stars is 2 stars and I quite enjoyed mucking about with cobra-cli. Onwards and upwards. Roll on Day 2!

Day 2

Dear oh dear oh dear. Weak. This was extremely painful to watch for the omnipresent, obnoxiously critical, entirely imaginary observer leering over my shoulder (imagine some dusty old Professor Snape figure sneering away). I finished ages ago and he's still tutting away, whispering snide comments and disparaging remarks in my ear. Perhaps staying up to 2am last night wasn't the brightest of ideas, or perhaps my mind is simply dimming as the cobwebs encroach, marking my steady progress into the depths of middle age. No! I'm not dead yet, dammit! I CAN STILL DO THIS!

Anyway, I got to the answer eventually by way of a few coffees and some confused meanderings around the misty peripheries of reason and logic, haphazardly stumbling onto a successful solution like one of those infinitely many monkeys with a typewriter. Turns out I was trying to be a bit too "clever" (admittedly, based on how this puzzle went, I'm taking quite the liberties with the word), going for some Leetcode-style highly-optimised single-pass solution and getting bewildered by edge cases. It's not a job interview you doofus. Just use brute force like you always do! Ugh.

After solving it myself, I had a look on YouTube at Jonathan Paulson's Python solution, which took a different approach. After I stopped crying (he solved it in 3 minutes 55 seconds despite having some annoying connectivity issues with the server that probably cost him around a minute, whereas it took me ... well, never you mind how long it took me. It's not a competition. Jeez, just leave it, will you?), I re-implemented his algorithm in Go (the "Alt" functions), which was clunky but educational - got to look at some iter and slices package stuff.

Tomorrow is another day.

Day 3

Still not exactly showering myself in glory. Easier than yesterday, but instead of using a simple regex for part 2, I went down a rabbit hole of finding start and stop indices manually and then slicing and dicing on those instead of letting the regex engine doing it for me. But I got the answer both ways and used more Go features than I otherwise would.

Regex in Go feels a bit incomplete, particularly the handling of named capturing groups (unless I'm missing something, you need to map the names to the result values manually to be able to refer to them by name, which kind of defeats the point).

Day 4

Out all day marshalling school kids around public transport and a busy museum (and yes, I was supposed to be marshalling the kids. It was sanctioned by the school. I didn't just decide to round up some random children and shepherd them around the city all day for kicks). And so I didn't get started till after dinner. And I'm exhausted since herding cats is tiring. So it didn't go that well, OK?

Anyway, I had some idea of organising the input into a 2D character array, creating 1D arrays of string "views" at the different angles (diagonal is quite fun), and doing a string count of XMAS and SAMX for each string in the views for the 4 different orientations. I'm quite glad I started down this track because I came across "runes" in the process, which is a cool name for unicode codepoints. But I got fed up because it turned out to be a bit of a faff and, like I said, I'm tired. And more importantly, I sensed it was unlikely that that approach would work for part 2. How right I was. So I came up with a very dull, workmanlike implementation that might as well have been written in C. Bo-ring!

Day 5

Go, why do you have no set in your standard collections? Yes, I know I can just use a map, but sometimes I want a set. Is it too much to ask? It's mildly irritating but also quite fun getting used to Go's quirks. As I mentioned above, the syntax feels to me like a hybrid of C-ish languages that I know (C, C++, Java, Rust) and Python, and I sometimes assume Go will behave like one or the other, before smacking me in the chops because Go is Go and it's a bit different from any of these.

It's a bit like that time I annoyed a bunch of Italians. I already spoke Spanish and a some French, so I spent a couple of weeks learning some basic Italian, and then went on holiday to Italy, possessed of the hideously arrogant misconception that I was fluent in the language. You know, if you don't know the word it's bound to be some Italianisation of whatever I'd say in Spanish/French, right? Am I right??? So yeah, I just swanned around Rome, happily spamming the locals with my bastard pidgin until being humbled by a bunch of confused looks and the occasional rude gesticulation and accompanying stream of fruity invective. AoC is definitely good for learning the idioms of a new coding language (particularly if that language happens to be Rust: I'm looking at you, Madam Compiler) in the same way an that an angry Roman flipping his fingers off his chin and telling you where to go (inside your own bottom, apparently) is quite motivating for actually bothering to get off your lazy arse and learn the true idioms of his spoken language. (I should probably say I'm exaggerating (er ... lying) for comic effect. In reality, all the Italians I met on that holiday were really friendly and usually found my "Italian" hilarious, if occasionally mystifying. At least I appeared to be trying to speak it instead of shouting English at them slowly.)

Anyhow, Day 5 was straightforward enough. A basic dependency graph problem. I just used a map of pages to their set (I mean - grrrr! - map) of prerequisites and it fell out reasonably quickly (once I'd fixed the sloppy bugs I introduced along the way of course).

Day 6

Brute forced this one.

Day 7

Brute forced this one too.

Yeah.

Day 8

Bunch of maps. Counting antinodes. Little picture on paper of adding/subtracting vectors so I don't get too befuddled. Fine. Whatever Can't really be bothered looking for more optimal solutions or making my code nice or anything. Got the answer. It's a Sunday. Happy. Yeah ...

... Well, a lot happier than I am about the water pipe leaking in my bathroom. The pipe is copper and some imcomprehensible clown of a contractor (or perhaps it was some recklessly over-ambitious previous home owner somewhere low down on the DIY Dunning-Kruger curve; there are some now-disabled DIY electrics - "interesting" was how the electrician put it - in the attic as well as some highly dubious carpentry in the kitchen so I wouldn't be surprised) went and partially embedded it in the concrete underflooring when they were laying the lovely faux wood floor panels (who the hell puts a "wooden" floor in their bathroom anyway. It's a bathroom. Use tiles you idiots!) Of course, the concrete has reacted with the copper over the years and has corroded it, resulting in a burst pipe, a puddle on my floor, who knows what behind the wooden panelling, and a day of stress on hold to the insurance company who are currently overwhelmed by the volume of claims arising from Storm Darragh. Of course they need to cut the wall away and possibly get a massive chisel out to hack at the concrete and it's costing a minor fortune. I'm paying one way or another. Either I pay for it all, or the insurance company does and I pay the excess and an increased premium the rest of my life. Anyway, rant over. Might actually talk about something actually relevant to advent of code tomorrow. Yeah, right!

Day 9

Let's talk about plumbing. Actually, let's not. And let's definitely not talk about the terrifying, crumbling, decaying mortar and brickwork currently holding up my house that we discovered behind the wall panels.

Right, Day9. Not to put too fine a point on it: total shitshow. I didn't have any particular problem coming up with a reasonable approach (if not optimal, at least workable - ultimately 200ms for both parts). 2-pointers solution for part1, single pass, fine. That didn't extend to part2, but ok, I just created maps of where the spaces and blocks were and corresponding sizes and iterated over them slotting the blocks in. But I made stupid error after stupid error and took an eternity to fix them all. Not sure why, but I didn't use the example data until the end, which would have sped things up considerably. I'm putting it down to the stress of having a house that's falling apart and the endless demands on my wallet from the plumber as he finds problem after problem. Yeah, that's what it was. But like I said, I don't want to talk about that.

Day 10

I had a really good day today. An excellent one in fact. It really was the best day I've had for quite some time ...

... so it's a good thing my abject performance on today's Advent of Code puzzle didn't spoil it.

Day 11

Day 2 took me ages, sad I know. I went from linked list implementation for part 1 (obviously borked for part 2, since it's hundreds of millions of elements by the time you get to iteration 40ish) to trying to come up with some massively overengineered hallucinogenic recursive thing because I noticed that you tend to get repeats. So a zero will expand to a massive list 20 odd iterations later, but some of these elements will be zeros so you know what they'll end up being in 20 iterations time, etc. And ... inhales deeply on spiff ... woah, man, it's like totally self-similar spiraling into infinity, patterns within patterns within patterns, dude.

So I totally missed the dead simple map-of-stone-to-count-of-occurrences solution that I ended up with, which only takes 12ms to run.

From foam slipper to crack a nut to broken-and-unfixable-Rube-Goldberg-machine-complete-with-flamethrower-and-chainsaw to crack a nut, totally oblivious to the nutcracker that was in the kitchen drawer all along. Ahhhh ... duh, why didn't I see that?

Takes nutcracker.

Crack.

If this wasn't a public repo I'd be using some fairly ripe and fruity vocabulary right now. But it is, so dang and blast it all to heck.

Day 12

Bleurgh! Is that a bit of sick in my mouth? Yep. What a disaster. That'll teach me to solve AoC problems with a hangover. Sleepy eyes, sluggish solving, cruddy code. The less said the better. Onto day 13 ...

Day 13

... Is it a crime to just use maths here? It's just a system of linear equations with a unique solution (number of unknowns = number of equations and the equations are not linearly dependent). Feels ... well ... cheaty.

But I'm still hungover, so 2 stars is 2 stars.

Day 14

See what you can do without a hangover? Perfectly sensible solution, bit of a manual thing to get part 2, but I did it a bit more formally later, choosing the iteration with the minimum variance of the robot positions.

Problem is I'm going to have a hangover tomorrow. The downside of Advent of Code is that it tends to coincide with Christmas parties.

Day 15

Graaahaaaaaaarrrraaarrrr ...

Day 16

...rrrraaaaahhhhhhgggh.

Dear God.

Day 17

Really happy with my part 1 code. Clear and simple.

But part 2? ... Goddamit, Eric! That's half my Tuesday gone.

Part 2 required spotting a really gnarly pattern in the output when you look at the register in octal. Goodness knows how to do show it analytically, but the following works:

• The rightmost digit of the output is fixed by the leftmost octal digit of the register.
• The exact mapping of register digit to output digit changes at each place value, but once the top digit is fixed, changing digits in other place values of the register has no effect.
• Then the second-bottom in the output is only affected by the second top octal digit in the input and so on.
• So with a bit of bitshifting of the register, and indexing into the corresponding output/program digits, find the matching digits at each of the 16 powers of 8 (my program length is 16 digits) and Bob's your uncle.

Did I explain that well? Not really. Can I be bothered to do better? No.

It's not the full story either. Bob was not my uncle.

To make things harder:

• There are sometimes multiple solutions to a given digit and sometimes none.
• So I had to store a bunch of alternative candidate solutions at each place value and chuck 'em out if they weren't working.
• At the end I took the smallest surviving candidate.

Clear? Thought not. It worked though. 2 stars. And it all runs in under 1ms.

As usual, I'll probably have made this waaaaay more complicated than it needed to be. Some cunning bastard will have a 2-line solution: "Oh, obviously you just do <insert cunning thing>". Yeah yeah, smug it up. Bet you're fun at parties.

Day 18

Relatively easy day, but what is it with me and off-by-one errors? I always find it disproportionally hard to reason about / visualise off-by-ones compared to other software concepts and this was a classic example, both with identifying the blocking corrupted byte (I got the timestep off by one) and with my initial attempt at a binary search, which went into an infinite spin because I failed to add one to the lower bound. It extends into real life too. Five friends coming round for pizza? I buy four pizzas and go hungry. Third house on the left? I knock on the second one and annoy a stranger.

Talking of right and left, I also suffer from an off-by-one error in the space of only those two values. Sad but true. I only nail it around 50% of the time, which is maximally useless. One time I left a house party in the Spanish city of Valencia at 2am, took a quick look at the map, and started leading a group of friends and acquaintances "towards home" along the street. Only I'd turned left instead of right and, given the slightly inebriated state of our group, the mistake wasn't spotted until well over an hour later when we reached the city limits and became alarmed by the sudden absence of buildings. By the time we got home at 5am some of the friends had become acquaintances and some of the acquaintances had become enemies. At least it was warm and it wasn't raining. Had it been either cold or raining my life would probably have turned out very different because one of the girls in the group, who I only vaguely knew at the time and who takes to cold and rain like your average housecat, is now my wife. True story.

Classic Ramage.

Day 19

Another Christmas party another hangover. Luckily for me today was easy. A bit of recursion here, a bit of memoisation there, and when I finally managed to count things properly, even my wine-damaged brain managed to lurch to victory in a relatively short time. Terrified of what this simplicity means for tomorrow and the weekend though.

Day 20

Not that bad really which probably means tomorrow will be hell on earth. Today was another grid/maze, but not one that needed any kind of complicated traversal. Crass brute force was good enough, especially since I unexpectedly ended up pouring more wine down my gullet last night and don't have the wit or energy to do any better.

Day 21

Part 2 took me forever. Forever.

Day 22

Easy. Next!

Day 23

Next!

Which means Day 24 is going to be ...

Day 24

... brutal. My code is a hellscape of ugly brute-forcery, heuristics and manual gubbins. Can't be bothered to make any improvements. At least I now know how to add two numbers using logic gates. Which is nice.

Day 25

Easy peasy as it should be. Still managed to make a meal of it.

And that's it! Till next year, suckas!