I have an off by one in my data, but the example i...

# advent-of-code

[Gerard Klijs]

I have an off by one in my data, but the example is fine. Earlier I had 43 in the example, since I didn't use the different geologic index to calculate the points 'behind' it.

[karelpeeters]

Where is the off-by-one? Part 1 or part 2?

[Gerard Klijs]

Only part 2, only in the real data. When I filter options more aggressively it does give the 'right' answer.

[karelpeeters]

What options? What filtering?

[Gerard Klijs]

How I handled the problem was kind of brute force, going to any direction, when needed changing tools. to prune it back I check the location and how fast that location has already been reached. When the current time is 7 or more slower the option can be removed (since whatever equipment is used, it could be changed and have the same time). If I set the time to check to the current fastest minus 3 I get the 'right' answer.

[karelpeeters]

Do you also handle cases where multiple points downwards can be reached faster by changing tools?

[Gerard Klijs]

I handle any case, once I have a time of a path reaching the target, I start trimming any path taking longer.

[karelpeeters]

Not sure whether that approach works in general, it's not one of the standard pathfinding algorithms. I'd just add "tool" as a 3rd axis and pathfind on that.

[Gerard Klijs]

But how did you account for the 3rd axis being more expensive?

[karelpeeters]

Keep a priority queue with states (x,y,tool,time) that are sorted by the time it took to reach them.

[karelpeeters]

(while running something like Dijkstra or even A*)

[Gerard Klijs]

Thanks, solution for part 2 runs correctly now, and under a second.