Re: [gnugo-devel] fuseki db

[Gunnar Farnebäck]

Doug wrote:
> I'm not sure I understand. You mean for matching the fuseki patterns
> during play, right?

Yes, only compile 64-bit hashes into the binary instead of less
compact full board descriptions (various associated values must remain
too of course).

> In principle, one could do all pattern matching with
> hashes, but I agree it makes particular sense for whole-board patterns.

Well, arbitrarily shaped patterns anywhere on the board and in any
orientation plus lots of wildcards does not seem very attractive for a
hashbased matching. I was thinking specifically of whole-board
patterns without wildcards.

> This did occur to me, but since there's normally some general 
> purpose compression applied along the way, it wasn't clear how big the 
> benefit might be.

That is true, but it should be possible to do better with a custom
compression. For some of us who have lots of cvs working copies and
(unpacked) old versions lying around it could also save quite a bit of
diskspace.

> One other thing I noticed: if pattern 1000 is put back in, regression test
> nngs4:810 ought to be changed as well. I guess I should check regression
> for the whole new fuseki db, in case there are others like that.

I don't think there are any more tests like that. I took out pattern
1000 specifically in response to that testcase.

> So should I make a patch putting those five patterns back in, or just edit
> the db files?

Either is fine with me but it would be good if someone more offered
their opinion on those patterns.

/Gunnar