The New Sudoku Players' Forum

[champagne]

Assuming the virtual catalog tested and cleaned, I am back to the topic "high density files..."

Although I see low interest currently, I am convinced that one day, the problem will come back.

Having in mind the last data base extensions in the field of the tridagon pattern, I think more and more that a canonical form

rank of the solution
81 bit field of the minimal sudoku pattern in this solution grid

is a good target.

This is in text mode something as

10 characters for the rank,
81/6 14 character for the pattern expressed in text mode, 6 bits per printable character

with then a potential for a .zip compressor.

Such a pattern is very easy to decompress using the virtual catalog.
Compression is a little harder with several steps:

Code: Select all

find the solution of the puzzle
morphs it to min lexical (and solve auto morphs if any)
Get the rank.

All tools are there now , but my brute force solver is not in DLL mode. I intend to build this DLL as next step, with all entries used in my programs.

[coloin]

81 bit field of the minimal sudoku pattern in this solution grid ....
81/6 14 character for the pattern expressed in text mode, 6 bits per printable character....

thats a big leap again !!! but maybe it can be done to go from minlex solution grid and minlex pattern -> actual puzzle.

starting from a min lex pattern [of the puzzle] ... what would be the best way to describe and then define the 6^8 x 2 permutes ....[3,359,232] ?

identify a box to be box1 [1 - 9]
+/- transpose - 0 or 1
row swap 1-3 [012345]
col swap 1-3 [012345]
identify box to be box5 [0123]
row swap 4-6 [012345]
col swap 4-6 [012345]
row swap 7-9 [012345]
col swap 7-9 [012345]

9x2x6x6x4x6x6x6x6 = 3,359,232

I guess which ever is easiest to program efficiently ....

[champagne]

81 bit field of the minimal sudoku pattern in this solution grid ....
81/6 14 character for the pattern expressed in text mode, 6 bits per printable character....

thats a big leap again !!! but maybe it can be done to go from minlex solution grid and minlex pattern -> actual puzzle.

starting from a min lex pattern [of the puzzle] ... what would be the best way to describe and then define the 6^8 x 2 permutes ....[3,359,232] ?

identify a box to be box1 [1 - 9]
+/- transpose - 0 or 1
row swap 1-3 [012345]
col swap 1-3 [012345]
identify box to be box5 [0123]
row swap 4-6 [012345]
col swap 4-6 [012345]
row swap 7-9 [012345]
col swap 7-9 [012345]

9x2x6x6x4x6x6x6x6 = 3,359,232

I guess which ever is easiest to program efficiently ....

In my DLL minlexing a given, the permutation is defined in the following way

Code: Select all

Transpose first or not
reorder rows
reorder columns
1296 * 1296 * 2 = 3 359 232 permutations


And the first rows gives the digit map.

The minlexing operation delivers also the list of auto morphs.

So where I am, I don't see this as a big leap. The only new point is to apply grid auto morphs to get the canonical view.

But a DLL for the brute force would be welcome, to have a relatively simple process.

Another issue is to have it accepted as the canonical form of any puzzle;