Re: [gnugo-devel] arend_1_18.1: next attempt to bring order to move_reasons (supercedes arend_1_17.1)

[Gunnar Farneback]

Arend wrote:
> This is an extended and corrected version of arend_1_17.1. There is
> actually more new added code than deleted old one, but I still believe
> it makes the main functions estimate_territorial_value and
> estimate_strategical_value more readable. I think what my patch does is
> most quickly summarized by (discard_rules) reproduced below.

This looks okay to me if it passes the regressions. A few comments for
later improvements:

> +struct discard_rule {
> +  int reason_type[40];
> +  discard_condition_fn_ptr condition;
> +  int flags;
> +  char trace_message[160];

Both array sizes should be symbolic constants. For the first we
already have the symbol MAX_REASONS.

> +/* Interprets the object of a reason and returns its position.
> + * If the object is a pair (of worms or dragons), the position of the first
> + * object is returned. (This is only used for trace outputs.) Returns
> + * -1 if move does not point to a location.

The usual convention is to return NO_MOVE in the last case.

> +/* Bitmap values for move_reason.status */
> +#define ACTIVE 0
> +#define TERRITORY_REDUNDANT 1
> +#define STRATEGICALLY_REDUNDANT 2
> +#define REDUNDANT 3
> +#define SECONDARY 4

It seems safer to have

#define REDUNDANT (TERRITORY_REDUNDANT | STRATEGICALLY_REDUNDANT)

/Gunnar