On Wed, Aug 28, 2019 at 11:23 AM Skip Cave <s...@caveconsulting.com> wrote: > Looks like Pete had issues with the efficiency of his J solution to the > Advent of Code: Marble Mania problem: > http://www.petecorey.com/blog/2018/12/14/advent-of-code-marble-mania/
That problem needs to work with three independent values: 1) The "board" of placed marbles 2) The number of the next marble to be placed 3) The scores of each of the elves (which implies knowing how many elves are playing) For something like this, I'd be strongly tempted to use OOP: https://www.jsoftware.com/help/learning/25.htm for the first draft. It might be possible to simplify things analytically, but I'd like to have working examples in front of me, to test against, before I went there. Simplifications: (1) creating an instance of the initial version of the game only needs to know how many are playing (and, in a sense, how long the game will continue - though using that number implies playing that game to completion). (2) We can represent the "current marble" by rotating the board so it's at position 0. [There may be faster approaches, but I'd want to be able to compare the speed of the possibly faster approach to this approach to know if it's actually faster. Approaches taking advantage of mathematical equivalences be faster.] I'd actually start out by coding for the case where the board has 3 or 4 marbles on it. Then I'd start dealing with the other possibilities. This is close to how he approached it, so it's likely to run into some of the same speed issues. Note that it's also possible to use an approach based on using modulo arithmetic on indices and #!.n inv (or complex #!.n ) but by itself that won't be faster. To be faster, I think I might try working out how to do "batches of 23" as "one step". I suspect that would get me something close to a 10x speedup. But, anyways, my experience with those "advent of code" problems has been that they are designed to tend to take hours to solve, and often reqire several code versions. Thanks, -- Raul ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
