I'm glad I'm not using Haskell lol it would be hell to try to optimize IntMaps or whatever

I wonder why vector works faster than hashtables
I didn't think hashing smth would be 5x slower but I guess it is

wot is fixnum
I'm guessing fixed-point numbers but I don't see why a specialized vector for it would be useful
Must have some obscure low-level applications I don't consider in my lofty tower of abstraction

Oh because racket has arbitrary-size numbers and using fixnums would force it to use the more space-efficient number
I suppose I could try fixnum vectors

"A fxvector is like a vector, but it holds only fixnums. The only advantage of a fxvector over a vector is that a shared version can be created with functions like shared-fxvector."

Never mind lmao

protagonist is HACKING AIRPORT SECURITY CAMERA

I've determined the indices of the fields that could belong to the first six ranges. They are:

'(0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19)

so, all of them,

Oh I accidentally did 1-6 instead of 0-5. But here's the list:

'(1 3 4 5 6 7 8 9 10 11 12 14) :/

Here's the list of the valid indices for each field:

'((14 10 8 6) (16 14 10 8 6 4 2 17 0 15 13 9 7 5 3 1) (6) (16 14 10 8 6 4 2 17 0 15 13 11 9 7 5 3 18 1) (16 14 10 8 6 4 2 17 9 7 5 1) (16 14 10 8 6 4 2 17 0 13 9 7 5 3 1) (16 14 10 8 6 4 2 17 0 15 13 9 7 5 3 18 1) (16 14 10 8 6 17 9 7 5 1) (16 14 12 10 8 6 4 2 17 0 15 13 11 9 7 5 3 18 1) (16 14 10 8 6 4 2 17 0 9 7 5 3 1) (16 14 10 8 6 2 17 9 7 5 1) (16 14 10 8 6 4 2 17 9 7 5 3 1) (16 14 12 10 8 6 4 19 2 17 0 15 13 11 9 7 5 3 18 1) (14 10 8 6 17 7) (16 14 10 8 6 17 9 7 5) (16 14 10 8 6 17 9 7) (14 10 8 6 7) (10 6) (10 8 6) (16 14 10 8 6 17 7))

I guess I'll have to do some constraint-solving...

I'm currently big struggle to make my solving correkt

it works. but not good

I'm gonna quickly refactor to not use mutable sets before I push

wait omfg (for ([i n])) iterates through (range n)?? I didn't know this omg

https://github.com/ionathanch/adventofcode-2020/blob/main/src/16.rkt
Gross. Disgusting. Absolutely abhorrent

okay now that I'm done it's time to get into bed and read haskal's solution

I swear I keep forgetting for loop keywords
I did an if-then-0 in my for/sum instead of simply using #:when, like a fool

Great it's another graph day
I didn't see the graph this time

I've taken the 4th-year algorithms course and yet I remember nothing

Oh my god it's so obvious. Of course it's maximum matching, you have edges between fields and the ranges they could belong to and you need a mapping from one to one exactly

I'm so mad
I'm resisting the temptation to immediately get up and go fix my solution to use a graph

I've decided not to use graphing
The part of the code the max bipartite matching would replace is only seven lines anyway so I don't think the solution would be any more concise
It's already a short enough solution really and even switching from using mutable vectors and sets to lists only it still runs pretty quick
I don't think it would be worth it

NO not ANOTHER GRID PROBLEM

what Do you MEAN FOUR DIMENTIOSNalsj AHASPACE

How do I generalize this lmao I've got... I've copypasted my 3D sol and added a dimension lmao

I was thinking I could tighten up vthe range of locations I check but looking at the visualizations on Reddit it seems like it's pretty monotonic and dense so I don't think any optimizations there would help

YOOOOO

I wasted so much time looking up how to read a string as an s-expression LMAO
It's (read (open-input-string str))... jsyk...

I think this one would be pretty easy in Haskell bc you could probably redefine - and + with different infixities
or even define new infix operators and string-replace all of your - and +

I am so pleased with this one lol

oh man compared to yesterday this was a heckin' stroll in teh park

I'm big struggle lmao

whew
https://github.com/ionathanch/adventofcode-2020/blob/main/src/19.rkt
it is done
I'm so tired

I'm pleased with my solution... now that I've found it lmao
I was so sure in the beginning that I was doing it the very bad no good way and should've used a graph or smth but as it turns out I can indeed just generate all of the strings that the rules produce (at least, for rules 42 and 31, which were the only ones I needed)

I enjoy that all of my solutions so far are fewer than 100 lines of code (with reasonable column widths ofc), but I do also have a 300-line custom library lol

spread out those 300 lines over 15 days, that's 20 extra lines per day, and my days are at most 70 lines, so it's still under 100 "amortized"

Sometimes I feel like I need to add like 50 lines of comments to explain what I'm doing
`ruleset` in my solution for today is particularly unfathomable

it contains an apply append of a map of a map of an apply string-append of an apply cartesian-product of a map
very cursed

I think the only days that felt like cheating were when I used `solve-chinese` and the one with the bags because I just used a graph lib

I think I could do it without the graph lib, I just need to implement bfs
...hmmm...