Re: [gnugo-devel] Owl code & shoulders

[Gunnar Farneback]

Trevor wrote:
> Does the OWL code have any support for identifying shoulders of eyes?
> The following points labelled A, B, and C are what I mean by shoulders:
> 
> 
> AOA   BOB  CO|
> O.O   O.O  O.|
> AOA   ---  --+
> 
> 
> In the first case, O must control (either with a stone there, or by
> being able to capture an X stone put there) 3 or the 4 shoulders
> (labeled A) for the '.' to be an eye.  (There are, of course, very
> strange exceptions, but this is almost always true.)
> 
> In the second case, O must control both shoulders (labeled B).
> 
> In the third case, O must control the single should C for '.' to 
> be an eye.

Yes, this is called "Topological Eye Analysis" in GNU Go and is
documented in doc/eyes.texi. It's implemented in the functions
topological_eye() and evaluate_diagonal_intersection() in optics.c.
The owl code calls topological_eye() from within owl_determine_life().

> Problem trevor.tst 450 (in the attached patch tar.gz) is an example 
> where such a concept would be very useful, either in the core OWL 
> code, or as helper functions to OWL patterns, I think.  I'm not
> sure quite how to attack this one.

I'll take a look at this later, but you can rest assured that without
knowing this concept at all, the life and death reading of GNU Go
would surely be a magnitude weaker. :-)

/Gunnar