Solution: Optimizing Aldous-Broder

# This is a modified---and biased!---version of the Aldous- Broder
# algorithm. If you thought the result was unbiased, consider that
# this version now *prefers* a least-visited neighbor over other
# neighbors; the result may or may not exhibit a tell-tale texture, but
# it's still biased!
class ModifiedAldousBroder
# Write your code here
  def self.on(grid)
    cell = grid.random_cell 
    unvisited = grid.size - 1 
    visits = Hash.new(0)
    while unvisited > 0
      visits[cell]+=1
      # find the set of least-visited neighbors, and pick one randomly
      neighbors_by_visit = cell.neighbors.group_by { |n| visits[n] }
      min = neighbors_by_visit.keys.min 
      neighbor = neighbors_by_visit[min].sample 
      if neighbor.links.empty?
        cell.link(neighbor)
        unvisited -= 1
      end
      cell = neighbor
    end
    grid
  end
end