RE: [computer-go] Core data structures: modify or copy?

["David Fotland" <fotland@xxxxxxxxxxxxxxx>]

I'll add that in my chess and arimaa programs, the board state is only a few
hundred bytes,
so I copy the state when making a move.  That way there is no incremental
do/undo code, and its simpler.
I still maintain a lot of state incrementally for speed though, like the
zobrist hash, toal material, etc.

David

> -----Original Message-----
> From: computer-go-bounces@xxxxxxxxxxxxxxxxx 
> [mailto:computer-go-bounces@xxxxxxxxxxxxxxxxx] On Behalf Of Don Dailey
> Sent: Monday, May 16, 2005 10:04 AM
> To: computer-go@xxxxxxxxxxxxxxxxx
> Cc: computer-go@xxxxxxxxxxxxxxxxx
> Subject: Re: [computer-go] Core data structures: modify or copy?
> 
> 
> 
> Hi Peter,
> 
> If the amount of work your program does making a move is 
> significant (you have a fairly complex data structure), then 
> you should seriously consider making your life simple, just 
> copy state and apply a move to the next state.  It will save 
> you a lot of time debugging and free you up greatly to experiment.
> 
> If your program is of the style where it needs to very 
> rapidly traverse the game tree, then you will probably have 
> to have a very simple data structure that is incrementally 
> updated and downdated to a global board.
> 
> Apparently Dave Fotland (from his post) uses both methods.
> 
> Stuart Yeates suggested a hybrid approach which is 
> interesting, but will no doubt cause some debugging 
> nightmares if you are not careful. If you have good 
> bookeeping skills, you can have great fun with pointers to 
> the parts of the data structure that do not change.
> 
> - Don
> 
> 
> 
> 
>    From: Peter Drake <drake@xxxxxxxxxxxxxxxxx>
>    Date: Sat, 14 May 2005 20:10:42 -0700
> 
>    I'm once again playing with the core data structures in my 
> Go program.
> 
>    There must, of course, be a data structure to represent 
> the board.   
>    If there is to be any kind of search, there are two options:
> 
>    1) Maintain only one copy of the board, making and undoing 
> all moves  
>    here.
>    2) Make a new copy at each node.
> 
>    #1 might require some extensive winding and unwinding, 
> especially in  
>    a best-first search such as proof-number search.  #2 uses 
> a lot of  
>    time copying (or rederiving) information on blocks, liberties, etc.
> 
>    Any arguments for one over the other?
> 
>    Peter Drake
>    Assistant Professor of Computer Science
>    Lewis & Clark College
>    http://www.lclark.edu/~drake/
> 
> 
>    _______________________________________________
>    computer-go mailing list
>    computer-go@xxxxxxxxxxxxxxxxx
>    http://www.computer-go.org/mailman/listinfo/computer-go/
> 
> _______________________________________________
> computer-go mailing list
> computer-go@xxxxxxxxxxxxxxxxx 
> http://www.computer-go.org/mailman/listinfo/computer-go/
> 


_______________________________________________
computer-go mailing list
computer-go@xxxxxxxxxxxxxxxxx
http://www.computer-go.org/mailman/listinfo/computer-go/