Re: [gnugo-devel] fuseki db

[Arend Bayer]

Gunnar wrote:

> Arend wrote:
> > The easiest solution that comes to my mind would be to patch
> > extract_fuseki to generate an .sgf-file instead,
> 
> I hope you're not thinking of making an sgf file where the fuseki
> patterns are represented as a "proper" branching move tree. As far as
> I know that would be far from easy to generate from extract_fuseki. An
> sgf file with lots of parallel variations, where only the leaf nodes
> represent patterns, should be feasible though.

I don't understand where the difficulty is. Of course there are
transpositions. But they can just be ignored, i.e. if one position can
be reached in multiple ways in the .sgf-file, it is enough if the
continuation is added to one of these branches.

> Quick solution:
> What about the Makefile patch below? A drawback is that cvs update
> will tend to refetch the file every time it thinks it's missing but
> that should be possible to avoid with some cvs flag or by wrapping cvs
> update in a script which first uncompresses fuseki19.db.

Woudldn't it be much simpler to remove fuseki19.db from CVS and use

>  fuseki19.c : Blah...
>       zcat fuseki19.db.gz | ./mkpat -b -f fuseki19 -i $(srcdir)/fuseki19.db 
> -o fuseki19.c

in the Makefile? Only drawback is that we would have to generate diffs
to fuseki19.db by hand.

Arend