[seqfan] Re: Knight's tour on infinite board

[M. F. Hasler]

On Sun, May 5, 2019, 17:08 Hugo Pfoertner wrote:

> @Maximilian: One has to be careful with the path encoding, because the
> direction changes against the spine of the previous path segment are not
> multiples of 360/8 degrees, but are dependent on the orientation of the
> previous move.


Yes of course I know that it's not exactly that value but switches between
two nearby values. I meant to make it simple to explain what I intended to
code by 0,±1, ±2, ±3, -4.
I should have added ~ or ≈ in front of 360°/8.
The main idea I wanted to suggest was to use turtle style relative encoding
that can be stored in one single 3 bit signed integer, rather than absolute
coordinates using 2 signed integers of ever-growing unbounded size for each
move, with the additional (main) drawback of being dependent on the
orientation of the first move and permuted+sign flipped in nontrivial ways
under the different possible mirror symmeties.
I hope everyone got the idea in spite of this inaccuracy.

- Maximilian