This year I'm attempting AoC in Idris2, which I don't really know. A couple of days I fell back to Python for expediency or did Python first so I could see the full puzzle first, and work on the Idris solution at my leisure. The challenge here is that I don't know the language, don't know the libraries, and I've got a couple of decades of imperative programming behind me.
(backfilled on day 11)
Day 1
- Learned how to read files
Day 2
Day 3
- Bailed for python
- Went back and redid in Idris
Day 4
- Looks like I started to switch to python (beginnings of a parser) then switched back
Day 5
- I did this in python between parts 1/2
- Decided to take the dx/dy approach while in python land
- I changed the code for part two and left instructions to make the part 2 solution work for part 1
Day 7
- I think I did python first
Day 8
- Looks like I did it in python first and then Idris
- This was the day I had issues trying to figure out Bitset.bit and ended up with a case statement instead.
Day 9
Day 10
Day 11
Day 12
Day 13
Had a weird issue I don't understand where (Int,Int) was a (Type,Type) ratherthan a Type.
Day 15
Initial version is a little slow, but works. I'll probably revisit and get rid of the sorted maps. There also was
a choice to be
./build/exec/day15 2127.84s user 22.29s system 99% cpu 35:59.38 total
Using a second SortedMap for the unvisited nodes: (score,Point) -> ()
./build/exec/day15 2.77s user 0.05s system 99% cpu 2.836 total
Day 16
Spent way too much time tracking down < 16 vs <= 16. Other than that pretty straightforward. For fun, I decided to write a little combinator parser instead of using Data.String.Parser. The "Show Packet" was a little bit of a pain because it had to be total, so I had to work in a Nat for depth.
Day 17
I spent some time rearranging stuff to make it perform better, turns out the performance issue might have been the nub I was doing at the end.
Day 18
Initially I parsed into a tree, but it was much easier to implement the normalization rules on a list of tokens.
Day 19
This one was a bit of a slog. Dunno if I have the best solution, but it's fast, if complicated. I decided that an invariant would be the vectors between points, with some normalization that is described at the head of day19/Main.idr.
Day 20
The trick here was that an infinite number of cells change in problem given in input, but not the example. The problem statement hints at this possibility.
Day 21
This is slower than I'd like and takes a bunch of memory, so I'm probably missing something.
build/exec/day21 162.74s user 14.75s system 99% cpu 2:58.17 total
Yeah, I convolved the two players states and didn't need to.
build/exec/day21 0.08s user 0.03s system 39% cpu 0.277 total
Day 23
Actually got to use A* for this. I'd wondered if I'd ever have a use for it. On the coding side, a bit of frustration dealing with Fin and Vect. It seems like you can't do much with Fin (like add 1 or realize a Fin 4 is also a Fin 7). I ended up just casting my problems away, but makes me wonder about the utility of Vect.
Day 24
Delayed because I was busy with Christmas prep. This one took some effort because I went completely in a wrong direction. I thought I was being asked to analyze the program, so I wrote code to translate into single static assignment form, and then a bunch of optimization passes. Eventually I retried with looking at all inputs (at each stage), winnowing duplicate states to the best key, and choosing true for all branch points that we had control over.
Day 25
This one was fairly easy. It could be made faster, maybe by using Buffer, but good enough.