Re: [gnugo-devel] trevor_1_23.4.tar.gz

[Trevor Morris]

At 07:46 PM 2/6/2002 +0100, Gunnar Farneback wrote:
>Trevor wrote:
>> >> Pattern RA007
>> >> 
>> >> X*   snapback
>> >> X.
>> >> 
>> >> :-,A,value(75)
>> >> 
>> >> aA   snapback
>> >> ab
>> >> 
>> >> ; rgoal[a] == 1 
>> >> ; && NORTH(A) != color
>> >> ; && SOUTH(A) != color
>> >> ; && EAST(A) != color
>> >> ; && WEST(A) != color
>> >> ; && olib(A) == 1 && xlib(b)==1
>> >
>> >Why this complex constraint? Wouldn't it be more straightforward and
>> >efficient to modify the pattern to the one below?
>> >
>> >?x?
>> >X*x
>> >X.?
>> >
>> >/Gunnar
>> 
>> The problem with this is that it doesn't match correctly at the edge,
>> which it should. 
>
>Right, wasn't thinking of the edge behavior.
>
>> There is a complexity throughout the code, of having special edge
>> patterns, corner patterns, and center patterns. This is a good
>> example.
>
>Usually the center, edges, and corners are dissimilar enough to need
>distinct patterns, or are you disagreeing about that?

I'm not sure if I disagree with that or not.  I often add
an edge pattern without thinking very carefully about whether it
would be better as without the edge, and vice-versa.  I'm not sure
if this is a real problem or not.

BTW, in this particular case, I've simply re-labelled the pattern as
"snapback or better" :

Pattern RA007

Y*   snapback or better
X.

:8,A,value(75)

aA
Xb

; rgoal[a] == 1 && xlib(b)==1

Turns out that I don't really care if it's captured or not;  it's
still a fine move.