ghfick wrote:Andrew Stuart's solver [sudokuwiki.org] detects the symmetry of givens. He refers to this technique as 'Gurth's Theorem'.

May I suggest that YZF_Sudoku should have this technique added?

Cenoman wrote:Andrew's solver detects this one, but if the puzzle is morphed a bit more (e.g; by circular swap of bands 1,2,3 to 2,3,1), it detects no longer the "symmetry", actually the automorphism, (nor would I have detected it manually). It doesn't detect the "stick symmetry" either (whether morphed or not).

What would be useful in a solver, would be detecting automorphic puzzles, for any automorphism.

As Cenoman said, to support automorphic puzzles, it is necessary to enumerate various isomorphisms of the puzzle, which will take a long time, and I am not sure if it is enumerating the various possible combinations of rows and columns,6^8=1679616 types in total.