Tomas Rokicki

I think the first step to getting it into the Rust code would be to figure out how to distinguish "moves" from "rotations". Right now the C++ code does it...

In this context, I'm referring to the first set (as in Set) of pieces in the puzzle definition (as in, perhaps, corners for the 3x3x3) and of that set, the...

Simple case: https://alpha.twizzle.net/edit/?alg=%28U+F+R+D+B+L%29%27 Clicking on a move actually moves the highlight to the move at the other end. -tom On Fri, Feb 14, 2025 at 2:02 PM DougCube ***@***.***> wrote:...

It's unlikely to matter for CPP; twsearch avoids allocations on any hot path. For Rust, it's definitely worth checking out.

I agree this is better in some ways, but not all. Primarily, at the moment, a .tws file encompasses not only a puzzle definition, but also implicitly a metric (since...

Yes, this is true, and I agree they are less than useful for this purpose. New files generated by puzzlegeometry in cubing.js mitigate this somewhat by, for each piece, indicating...

Can't we just plug in a trait? -tom On Fri, Mar 14, 2025 at 6:40 PM Lucas Garron ***@***.***> wrote: > I have a draft at > https://github.com/cubing/twsearch/compare/lgarron/TargetPatternCache but >...

Sounds very nice! Compare the -A option in the C++ twsearch which focuses on searching for algorithms that affect only a few pieces, using three distinct strategies. If you are...

This is a good point. For all puzzles, the rotations produced are not redundant on an axis; for every axis, we produce exactly one rotation move. So for instance on...

Right. What we do is pretty basic IDFS. We do have a heuristic, but we do not order the children by that heuristic and "explore most promising children first" so...