Re: Incremental Data Structures

[Matt Gokey <mgokey@xxxxxxxxxx>]

David Mechner wrote:
> ...
> We specifically wanted to avoid treating each data structure and
> algorithm individually, because writing and debugging backtracking
> routines is tricky and time consuming.
> ...
> I think there's a trap for all of us in computer go, which is the
> impulse to put lots of time and effort into things that we know how go
> about doing - like interface, or optimizing for memory and cpu time -
> instead of the things that are hard and confusing - making the
> computer play go.
>
> Matt Gokey wrote:
>> IMHO, I feel an incremental system using algorithms implemented in
>> concert with data structures designed to be revertable/incremental
>> using deltas only would work more efficiently especially if paired
>> with custom allocation/deallocation using a combination of
>> pools/dynamic allocation by object type to prevent fragmentation
>> problems.

I understand your position and agree with the points you've made entirely,
but since I plan to experiment and use a kind of searching technique
requiring visiting thousands upon thousands of positions, I think
efficiency and optimizing will be a necessary evil.
I've read many threads of discussion in the mail list archives meandering
through many of the same issues over and over rather than discussing new
ideas about how to play go.  That's the exciting discussion, in my opinion
also.  You inquired about my thoughts on your revertible system, and my
comments relate to it in light of the kind of research I want to do.
That said, I'll probably start with a simple system, yours or something
similar, until experience reveals its inadequacies. :-)

Matt