./pachi -r RULESET to set rules from commandline
[pachi/nmclean.git] / playout.h
blobbc13f2dba370bddbd7730d28563b37a5dbe38104
1 #ifndef PACHI_PLAYOUT_H
2 #define PACHI_PLAYOUT_H
4 #define MAX_GAMELEN 600
6 struct board;
7 struct move;
8 enum stone;
9 struct prior_map;
10 struct board_ownermap;
13 /** Playout policy interface: */
15 struct playout_policy;
16 struct playout_setup;
18 /* Initialize policy data structures for new playout; subsequent choose calls
19 * (but not assess/permit calls!) will all be made on the same board; if
20 * setboard is used, it is guaranteed that choose will pick all moves played
21 * on the board subsequently. The routine is expected to initialize b->ps
22 * with internal data. At the playout end, b->ps will be simply free()d,
23 * so make sure all data is within single allocated block. */
24 typedef void (*playoutp_setboard)(struct playout_policy *playout_policy, struct board *b);
26 /* Pick the next playout simulation move. */
27 typedef coord_t (*playoutp_choose)(struct playout_policy *playout_policy, struct playout_setup *playout_setup, struct board *b, enum stone to_play);
29 /* Set number of won (>0) or lost (<0) games for each considerable
30 * move (usually a proportion of @games); can leave some untouched
31 * if policy has no opinion. The number must have proper parity;
32 * just use uct/prior.h:add_prior_value(). */
33 typedef void (*playoutp_assess)(struct playout_policy *playout_policy, struct prior_map *map, int games);
35 /* Allow play of randomly selected move. */
36 typedef bool (*playoutp_permit)(struct playout_policy *playout_policy, struct board *b, struct move *m);
38 /* Tear down the policy state; policy and policy->data will be free()d by caller. */
39 typedef void (*playoutp_done)(struct playout_policy *playout_policy);
41 struct playout_policy {
42 int debug_level;
43 /* We call setboard when we start new playout.
44 * We call choose when we ask policy about next move.
45 * We call assess when we ask policy about how good given move is.
46 * We call permit when we ask policy if we can make a randomly chosen move. */
47 playoutp_setboard setboard;
48 playoutp_choose choose;
49 playoutp_assess assess;
50 playoutp_permit permit;
51 playoutp_done done;
52 /* Particular playout policy's internal data. */
53 void *data;
57 /** Playout engine interface: */
59 /* Engine hook for forcing moves before doing policy decision.
60 * Return pass to forward to policy. */
61 typedef coord_t (*playouth_prepolicy)(struct playout_policy *playout_policy, struct playout_setup *setup, struct board *b, enum stone color);
63 /* Engine hook for choosing moves in case policy did not choose
64 * a move.
65 * Return pass to forward to uniformly random selection. */
66 typedef coord_t (*playouth_postpolicy)(struct playout_policy *playout_policy, struct playout_setup *setup, struct board *b, enum stone color);
68 struct playout_setup {
69 unsigned int gamelen; /* Maximal # of moves in playout. */
70 /* Minimal difference between captures to terminate the playout.
71 * 0 means don't check. */
72 unsigned int mercymin;
74 void *hook_data; // for hook to reference its state
75 playouth_prepolicy prepolicy_hook;
76 playouth_postpolicy postpolicy_hook;
80 struct playout_amafmap {
81 /* We keep record of the game so that we can
82 * examine nakade moves; really going out of our way to
83 * implement nakade AMAF properly turns out to be crucial
84 * when reading some tactical positions in depth (even if
85 * they are just one-stone-snapback). */
86 coord_t game[MAX_GAMELEN];
87 bool is_ko_capture[MAX_GAMELEN];
88 int gamelen;
89 /* Our current position in the game sequence; in AMAF, we search
90 * the range [game_baselen, gamelen[ */
91 int game_baselen;
95 /* >0: starting_color wins, <0: starting_color loses; the actual
96 * number is a DOUBLE of the score difference
97 * 0: superko inside the game tree (XXX: jigo not handled) */
98 int play_random_game(struct playout_setup *setup,
99 struct board *b, enum stone starting_color,
100 struct playout_amafmap *amafmap,
101 struct board_ownermap *ownermap,
102 struct playout_policy *policy);
104 coord_t play_random_move(struct playout_setup *setup,
105 struct board *b, enum stone color,
106 struct playout_policy *policy);
108 #endif