Hi newtton!
Thanks for your E-mail. See below for my response.
On Sunday 18 Oct 2009 22:06:37 newtton wrote:
> I ran across a Freecell position that gives your solver (and others)
> difficulty. (The moves leading to this position are given below.) Here's
> what happened: I was playing with Freecell Pro and got into this
> position. When I checked the solvability, I got Yes, but I couldn't see
> how myself. I fumbled around a bit and then checked the solvability
> again, and it said No. This would not have been a surprise except the
> only difference from the first position was a reversible move (the three
> of clubs can be moved freely between two columns.) What a shock! I
> thought the solver always told the truth (or came to no conclusion at
> all.) I reported the problem to Michael Keller and he suggested the newer
> version of FC Pro that has 3 solvers, including yours. Yours reports this
> position impossible. But the 3rd solver, Patsolve, reported it solvable
> -- whereever the 3 of clubs was placed. Using the solution generated by
> Patsolve I was able to see how to get out of the difficulty. This is just
> my opinion, but it seems to me that if you are going to release a program
> to the public and call it a solver, then it is understandable if it can't
> come to a conclusion about certain positions, but it should at least come
> to the correct conclusion if it does.
>
Freecell Solver in some of it solver configuration uses heuristics that may
report some solvable board layouts as unsolvable. This is due to the fact that
it uses meta-moves (= sequences of several atomic moves and meta moves)
instead of atomic moves, which move only one cards. However, there is a preset
called "good-intentions" that first runs a good meta-moves heuristic and then
a good (for some values of good) atomic moves one to yield a usually fast and
absolutely accurate verdict. For more information, consult the Freecell Solver
documentation here, or alternatively the Freecell Pro documentation:
http://fc-solve.berlios.de/docs/#distributed-docs
I should note that Freecell Pro, as great as it is, is relatively lacking and
buggy and suffers from several major design and internals issues. It has also
incorporated an incredibly old, under-optimised and buggy version of Freecell
Solver, which I refuse to support (due to the fact that the current version of
the fc-solve library should be backwards compatible with it, but greatly
enhanced). I am now much happier with the integration of Freecell Solver into
PySolFC:
*
http://pysolfc.sourceforge.net/
*
http://cards.wikia.com/wiki/PySol
Like FreeCell Pro, PySolFC is open-source, but as opposed to it, it is cross-
platform, running mostly natively on Windows, Linux, Mac OS X and other
distributions, has much more Solitaire variants (many of which also have a
Freecell Solver-supported solver and help system, which despite its historical
name can now solve several other Solitaire variants besides Freecell). PySolFC
is also arguably the most powerful and featureful open-source (and free-of-
charge) Solitaire suite.
> #715037 Attempt: 5 NumFcs=4 (FCPro)
> 8h 5d 5c 5b 5a d5 65 65 68 65
> 6d 76 72 a2 6a 76 74 a7 6a 83
> 72 c6 38 7c a7 83 38 6a 83 c6
> 56 7c 38 83
>
Well, using Freecell Pro would be a bit hard because I'm based on Linux (which
luckily for you is still x86) and would need to use wine or Virtual Box. I'll
try but you can ease my job by giving the board layout in the Freecell Solver
notation:
http://fc-solve.berlios.de/docs/distro/README.html
Regards,
Shlomi Fish
--
-----------------------------------------------------------------
Shlomi Fish http://www.shlomifish.org/
Why I Love Perl - http://shlom.in/joy-of-perl
Chuck Norris read the entire English Wikipedia in 24 hours. Twice.
Received on Sun Oct 18 2009 - 14:17:40 IST