🚀 go-pugleaf

christos@cco.caltech.edu (Christos S. Christoforou) writes:

>Hi!
>I have questions concerning the use of the
>gnuan feature of Gnuchess.
>If one wants to have gnuan solve chess problems
>(like for instance, white to play and win etc),
>in what sort of format should the input file be?

>And another question: Has anyone tried to have gnuan
>run in the background of a Unix environment? How would
>a batch file look? I tried it once and crashed our servers
>and I don't think I should be doing it again unless
>someone who knows more explains a bit about it to me.

>Thanks for all replies/suggestions/hints.

>christos

For making a list of moves, Henry Thomas' (sp?) "notation"
program is pretty nice.  It takes algebraic notation and
will yield gnuan format with the "-d gnu" option.

I use the following script for running gnuan in the background:

--------------cut snip cut snip cut snip----------------------
#/bin/csh run-gnuan
# Run gnuan against a list of gnuan input files
#   given as arguments to run-gnuan.
# Stores results in ${1}.any, ${2}.any, etc.
#   (use distinct basenames for input files!)

# Set the max number of moves and time per ply in minutes, here:
set moves = 29; set time = 10   # for thorough analysis
#set moves = 2; set time = 1   # speed check for legal moves

foreach i ( $* )

gnuan > ${i:r}.any << !
$i
$moves
$time
!

end
--------------cut snip cut snip cut snip----------------------

You will need to delete all the leading <tabs> on each line,
and change the other <tabs> to spaces in the "notation"
output so that "gnuan" will accept it, though.

Also, some illegal moves slip through "notation" -- it wasn't
equipped for that.  So I run the files through gnuan with
moves=1 or 2 to check legality.
Oh yeah, sometimes tournament games *have* illegal moves.
Please don't mention this to the solvability people.
Anyway, these programs shouldn't just die when this happens.

If you want to run the script in the background, put a "&"
at the end of the command line, e.g.:

run-gnuan file1 file2 &

Be careful with the time and moves on a heavily-used machine!

I don't think "gnuan" will solve a *position*.  This would be
a nice improvement (hint hint :).  Most of the games I analyze
have only a few moves that I'm puzzled over (big grin here)
and I hate to wait a day for "gnuan" just to get to the move.

It does give the best line from every position that occurs
during a game.  The "score" is the backed up evaluation from
the best line.
I've been wondering exactly what the "ply" column is, and why
the best line is sometimes shorter than the "ply" number.

"gnuan" wasn't happy with my queening a pawn, or knighting it,
etc.  Fortunately, I could analyze most of those continuations
(:-).

Mike White
michael-j-white@cecer.army.mil

mwhite@osiris.cso.uiuc.edu (Dr. Michael J. White) writes:
>I don't think "gnuan" will solve a *position*.  This would be
>a nice improvement (hint hint :).  Most of the games I analyze
>have only a few moves that I'm puzzled over (big grin here)
>and I hate to wait a day for "gnuan" just to get to the move.

I hated this too, so I made a small "quick and dirty" update to gunan.c.
The three lines of code:

       flag.force = 0;
       SelectMove (computer, 1);
       Undo ();

appears twice in gnuan.c (from line 825 and 871 in the distribution I have).

In both cases I have replaced them with:

       if( s[ strlen( s ) - 1 ] == '#' )
         s[ strlen( s ) - 1 ] = 0;
       else {
         flag.force = 0;
         SelectMove (computer, 1);
         Undo ();
       }

This has the effect that moves suffixed with # in the input file aren't
analyzed, they are just entered.

Jan