Re: A Sudoku solver in Tcl

Bernard Desgraupes <bdesgraupes@easyconnect.fr> wrote:
> In version 0.2 (not released yet), I have now implemented the Block to
> Row/Col reduction method and am currently working on the Naked and
> Hidden subsets reduction. I envision a "-clue" or "-hint" option which
> would return only a suggestion about the next move while solving a
> sudoku.

Yes, yes, yes :-)
Also an option to disable the dancing-stuff.
While I understand its value of just getting a solution,
my logical self grossly refutes any trial&backtrack
attempt for sudoku. Its just playing unfair :-)

I've found a trick for sudoku (general, not your prog):
+-------+-------+-------+
| 1 2 3 | 4 6 7 | 5 8 9 |
| 4 . 6 | 1 . 9 | 7 2 3 |
| 7 . 9 | 2 . . | 4 6 1 |
+-------+-------+-------+
| . . . | . . 8 | . . . |

the digits 5 and 8 are not yet determined in the
first two of these blocks, and in the second block
there is another possible place for just one of them.

My trick now, for sudokus found in magazines (which
are supposed to have only one solution) is that
in the middle block, the "5" must be placed bottomright
because if "3" was there, then the sudoku would
have two solutions.

I'd like to hear opinions on whether this is cheating,
or if this is a legal deduction. (perhaps it even has
already a name by which it is known to experts)

In most cases, I also found other ways around this trick,
but in one (from a book) that appeared to be the only way
to continue at some point.

> I have just installed the project on SOurceForge now:
> https://sourceforge.net/projects/sudokut/
This one doesn't work yet.
The page is empty (except for the topmost sourcforge header-line),
and the page title says: "SourceForge.net: Exiting with Error"

> Thanks for the feedback,
Thanks for the program in the first place!
Received on Sat Dec 3 03:28:41 2005