BFS VS advent-of-code

For day 1 and day 2, i used my custom transpiler. But day 6 was an interesting challenge: there was basically no parsing required! So, instead, i manually crafted this piece of Brainf*ck nonsense. It solves part 1! Since all non-brainf*ck characters are ignored, i've manually commented it to make it as readable as Brainf*ck can be.

For the purpose of writing more interesting programs with it, i ended up making a Forth-like custom language on top of Brainf*ck that allows me to give a name to snippets like the one above, like addi, and use them instead of having to copy and paste everything; plus it does some rudimentary type-checking, which ensures i'm not messing up my memory layout too much. Thanks to it, i can write things such as this snippet, which i used to read numbers from the input:

After that, it was a fairly simple loop, nothing too complicated. The full code is here, and the generated raw brainf*ck is alongside it: 975 lines of nonsense. And that's it! My last brainf*ck solution of the year; i didn't expect to be able to solve that many days: 1, 2, 4, 6, 9, 10, and now 25... I've learned a lot, expended my library of macros quite significantly, and now i know what to clean for next year!

I don't have such functions: i have time-travelling hash-maps. :P

full code: https://github.com/nicuveo/advent-of-code/blob/main/2022/haskell/src/Day17.hs

part 1, original, transpiled

code

code on Github

For part 2, i used the transpiler. I duplicate the buffer of 14 cells, sort it using a bubble sort, and then only do 13 comparisons.

As usual, code on Github, recording on Twitch.