Advent of Nickel (2022)
This is my attempt to solve each of the advent of code challenges in nickel for the 2022 advent of code.
General notes
nickel doesn't have file IO, so each input is checked in as a string const
nickel also is very slow, so most of these execute slowly.
nickel's interpreter uses a large amount of memory too.
Every solution checked in completed, on my machine, in under 64GiB of memory, but several of them used quite a bit of that.
Solutions are in default.ncl files in their given day. It's often wise to
comment out part1 / part2 and only compute 1 part at a time since sometimes
both parts together take significantly longer or more memory, and I only verified one part at a time completed for me.
Specific days
Some days have specific notes, see below
day 10
Part 2 of day 10 requires specific output formatting.
You can achieve this with jq
$ nickel export -f day10/default.ncl | jq '.part2' -cr
Closing notes
This was my attempt to solve the advent of code in nickel, and it was a failed attempt.
I did 2021's in nix, and that was already difficult at times due to nix's quite poor performance... But nickel is exponentially worse.
For day 11, part 2, I wrote nickel code to solve it, and based on some back-of-the-envelope math, it will require more memory than exists on the planet, and multiple decades, to actually complete. It certainly doesn't complete in 64GiB of ram and 4 hours.
The same solution, when ported to nix line-by-line (surprisingly easy to do! They really are similar languages), executes in 7 seconds and a mere 900MiB of memory.
At this point, I think the only way to make any of the later days execute will be to start improving the nickel interpreter's performance rather than to improve my nickel code or stumble upon new algorithmic insights.
I spent some time trying to do just that, but it's not happening anytime too soon.
So, in closing, I'm throwing in the towel early on this one. I'll do the reamining days in nix instead (though those will also be done a bit late, I've got some life-stuff to deal with first!), and call that good enough.
I might come back to this, either to try and improve nickel's performance myself and make progress, or when the upstream nickel development progresses such that these become tractable.
That's all for now. Merry Christmas :)