Re: [Bug-gnubg] Match and Position ID question

[Øystein Johansen]

Thomas A. Moulton wrote:
> Am I correct in the assumption that the Match ID and Position ID pair
> uniquely identify the state of a match in progress?

Yes!

> Is it documented how they are computed (or better yet are there
> functions that can compute them)?

Yes, see the links Michael sent. Is basically 10 bytes (80 sits) which
send through a base 64 ascii subroutines.

BTW: The base 64 subroutines is also implemented in Glib, if you use the
set in glib, which is just as convenient, just remove the training '=='.

> What I would like build is a command line interface as a demo of
> evaluating a position and determining the proper actions.
> (Cube, Move, etc)

Something like:

$ evaluate MID:PID --level=worldclass

and then it finds the correct cube action etc. Or:

$ evaluate MID:PID --dice=31 --level=worldclass

finds the best moves with a 31 roll.

I remember I isolated the evaluator code some years ago, compiled it to
a DLL on my windows system, (or .so on a linux system). That wasn't
really hard, so I guess your task should be solvable.

> I am also hoping it can be stateless so that each request can provide a
> new MID/PID pair and still remain fairly efficient.

I think it's possible to have it stateless.

However: A different solution.

Have an external player running as a deamon, and then make your command
line tool send requests to the deamon and get evaluations back.

.. or maybe the python interface .... I can think of many ways to solve
this right now...

-Øystein