Re: [gnugo-devel] readconnect

[Tristan Cazenave]

Daniel Bump a écrit :
> 
> Hi Tristan,
> 
> A couple of things about readconnect.c.
> 
> I think it would be good if we try to get this functional
> as soon as possible.  At the moment you are the only one
> that really understands the code. I think as soon as
> Gunnar and I understand the code well enough to work on it
> it will improve quickly.
> 
> We are studying your papers (now linked from devel.html)
> but there is also the matter of documenting the code. Perhaps
> this is what is most urgently needed.

ok, I will add more comments today.

> 
> If you were to give us ascii documentation we would take
> responsibility for converting it into Texinfo and putting
> it into the doc.
> 
> In addition to the Texinfo documentation, nearly all GNU
> Go functions are prefaced by comments describing what
> they do. These comments are missing from readconnect.c.
> I think it would be helpful if you could add them.
> 
> We'd like to be able to detect situations when connection
> depends on ko. The usual reading algorithms have 4 return
> values, which in numerical order are WIN > KO_A > KO_B > 0
> (=LOSE). I presume that the Abstract Proof search
> algorithm, like alpha/beta, is not restricted to AND/OR
> trees but can take more general truth values. Is this
> correct? How hard will it be to retrofit ko awareness into
> your algorithms?

For ko values, it is doable.
For integer values, i am working on it...


Tristan.