From: Christian Heller Date: Sun, 8 Mar 2015 03:09:37 +0000 (+0100) Subject: Server/py: Add movable AI. X-Git-Tag: tce~377 X-Git-Url: https://plomlompom.com/repos/?a=commitdiff_plain;h=94504421e041769058facdea5dbf310d3561c3aa;p=plomrogue Server/py: Add movable AI. --- diff --git a/libplomrogue.c b/libplomrogue.c index fd66755..384c81d 100644 --- a/libplomrogue.c +++ b/libplomrogue.c @@ -419,3 +419,157 @@ extern uint8_t build_fov_map(uint8_t y, uint8_t x, free_angles(shadows); return 0; } + +static uint16_t * score_map = NULL; +static uint16_t neighbor_scores[6]; + +/* Init AI score map. Return 1 on failure, else 0. */ +extern uint8_t init_score_map() +{ + uint32_t map_size = maplength * maplength; + score_map = malloc(map_size * sizeof(uint16_t)); + if (!score_map) + { + return 1; + } + uint32_t i = 0; + for (; i < map_size; i++) + { + score_map[i] = UINT16_MAX; + } + return 0; +} + +/* Set score_map[pos] to score. Return 1 on failure, else 0. */ +extern uint8_t set_map_score(uint16_t pos, uint16_t score) +{ + if (!score_map) + { + return 1; + } + score_map[pos] = score; +/* + uint32_t mup_size = maplength * maplength; + uint32_t pus; + FILE * file = fopen("test_set", "w"); + for (pus = 0; pus < mup_size; pus++) + { + fprintf(file, "%d ", score_map[pus]); + if (0 == pus % maplength) + { + fprintf(file, "\n"); + } + } + fclose(file); +*/ + return 0; +} + +/* Get score_map[pos]. Return uint16_t value on success, -1 on failure. */ +extern int32_t get_map_score(uint16_t pos) +{ + if (!score_map) + { + return -1; + } + return score_map[pos]; +} + +/* Free score_map. */ +extern void free_score_map() +{ + free(score_map); + score_map = NULL; +} + +/* Write into "neighbors" scores of the immediate neighbors of the score_map + * cell at pos_i (array index), as found in the directions north-east, east, + * south-east etc. (clockwise order). Use kill_score for illegal neighborhoods + * (i.e. if direction would lead beyond the map's border). + */ +static void get_neighbor_scores(uint16_t pos_i, uint16_t kill_score, + uint16_t * neighbors) +{ + uint32_t map_size = maplength * maplength; + uint8_t open_north = pos_i >= maplength; + uint8_t open_east = pos_i + 1 % maplength; + uint8_t open_south = pos_i + maplength < map_size; + uint8_t open_west = pos_i % maplength; + uint8_t is_indented = (pos_i / maplength) % 2; + uint8_t open_diag_west = is_indented || open_west; + uint8_t open_diag_east = !is_indented || open_east; + neighbors[0] = !(open_north && open_diag_east) ? kill_score : + score_map[pos_i - maplength + is_indented]; + neighbors[1] = !(open_east) ? kill_score : score_map[pos_i + 1]; + neighbors[2] = !(open_south && open_diag_east) ? kill_score : + score_map[pos_i + maplength + is_indented]; + neighbors[3] = !(open_south && open_diag_west) ? kill_score : + score_map[pos_i + maplength - !is_indented]; + neighbors[4] = !(open_west) ? kill_score : score_map[pos_i - 1]; + neighbors[5] = !(open_north && open_diag_west) ? kill_score : + score_map[pos_i - maplength - !is_indented]; +} + +/* Call get_neighbor_scores() on neighbor_scores buffer. Return 1 on error. */ +extern uint8_t ready_neighbor_scores(uint16_t pos) +{ + if (!score_map) + { + return 1; + } + get_neighbor_scores(pos, UINT16_MAX, neighbor_scores); + return 0; +} + +/* Return i-th position from neighbor_scores buffer.*/ +extern uint16_t get_neighbor_score(uint8_t i) +{ + return neighbor_scores[i]; +} + +/* Iterate over scored cells in score_map geometry. Compare each cell's score + * against the score of its immediate neighbors in 6 directions. If any + * neighbor's score is at least two points lower than the current cell's score, + * re-set it to 1 point higher than its lowest-scored neighbor. Repeat this + * whole process until all cells have settled on their final score. Ignore cells + * whose score is greater than UINT16_MAX - 1 (treat those as unreachable). + * Return 1 on error, else 0. + */ +extern uint8_t dijkstra_map() +{ + if (!score_map) + { + return 1; + } + uint16_t max_score = UINT16_MAX - 1; + uint32_t map_size = maplength * maplength; + uint32_t pos; + uint16_t i_scans, neighbors[6], min_neighbor; + uint8_t scores_still_changing = 1; + uint8_t i_dirs; + for (i_scans = 0; scores_still_changing; i_scans++) + { + scores_still_changing = 0; + for (pos = 0; pos < map_size; pos++) + { + if (score_map[pos] <= max_score) + { + get_neighbor_scores(pos, max_score, neighbors); + min_neighbor = max_score; + for (i_dirs = 0; i_dirs < 6; i_dirs++) + { + if (min_neighbor > neighbors[i_dirs]) + { + min_neighbor = neighbors[i_dirs]; + } + } + if (score_map[pos] > min_neighbor + 1) + { + score_map[pos] = min_neighbor + 1; + scores_still_changing = 1; + } + } + } + } + return 0; +} diff --git a/plomrogue-server.py b/plomrogue-server.py index 5223e4b..a63eedd 100755 --- a/plomrogue-server.py +++ b/plomrogue-server.py @@ -42,6 +42,16 @@ def prep_library(): libpr.build_fov_map.argtypes = [ctypes.c_uint8, ctypes.c_uint8, ctypes.c_char_p, ctypes.c_char_p] libpr.build_fov_map.restype = ctypes.c_uint8 + libpr.init_score_map.restype = ctypes.c_uint8 + libpr.set_map_score.argtypes = [ctypes.c_uint16, ctypes.c_uint16] + libpr.set_map_score.restype = ctypes.c_uint8 + libpr.get_map_score.argtypes = [ctypes.c_uint16] + libpr.get_map_score.restype = ctypes.c_int32 + libpr.get_neighbor_score.argtypes = [ctypes.c_uint8] + libpr.get_neighbor_score.restype = ctypes.c_uint16 + libpr.ready_neighbor_scores.argtpes = [ctypes.c_uint16] + libpr.ready_neighbor_scores.restype = ctypes.c_uint8 + libpr.dijkstra_map.restype = ctypes.c_uint8 return libpr @@ -756,9 +766,178 @@ def hunger(t): decrement_lifepoints(t) -def get_dir_to_nearest_target(t, c): - # Dummy - return False +def get_dir_to_target(t, filter): + """Try to set T_COMMAND/T_ARGUMENT for move to "filter"-determined target. + + The path-wise nearest target is chosen, via the shortest available path. + Target must not be t. On succcess, return positive value, else False. + Filters: + "a": Thing in FOV is below a certain distance, animate, but of ThingType + that is not t's, and starts out weaker than t is; build path as + avoiding things of t's ThingType + "f": neighbor cell (not inhabited by any animate Thing) further away from + animate Thing not further than x steps away and in FOV and of a + ThingType that is not t's, and starts out stronger or as strong as t + is currently; or (cornered), if no such flight cell, but Thing of + above criteria is too near,1 a cell closer to it, or, if less near, + just wait + "c": Thing in memorized map is consumable + "s": memory map cell with greatest-reachable degree of unexploredness + """ + + def set_map_score(pos, score): + test = libpr.set_map_score(pos, score) + if test: + raise RuntimeError("No score map allocated for set_map_score().") + + def get_map_score(pos): + result = libpr.get_map_score(pos) + if result < 0: + raise RuntimeError("No score map allocated for get_map_score().") + return result + + def seeing_thing(): + if t["fovmap"] and ("a" == filter or "f" == filter): + for id in world_db["Things"]: + Thing = world_db["Things"][id] + if Thing != t and Thing["T_LIFEPOINTS"] and \ + t["T_TYPE"] != Thing["T_TYPE"] and \ + 'v' == chr(t["fovmap"][(Thing["T_POSY"] + * world_db["MAP_LENGTH"]) + + Thing["T_POSX"]]): + ThingType = world_db["ThingTypes"][Thing["T_TYPE"]] + if ("f" == filter and ThingType["TT_LIFEPOINTS"] >= + t["T_LIFEPOINTS"]) \ + or ("a" == filter and ThingType["TT_LIFEPOINTS"] < + t["T_LIFEPOINTS"]): + return True + elif t["T_MEMMAP"] and "c" == filter: + for mt in t["T_MEMTHING"]: + if ' ' != chr(t["T_MEMMAP"][(mt[1] * world_db["MAP_LENGTH"]) + + mt[2]]) \ + and world_db["ThingTypes"][mt[0]]["TT_CONSUMABLE"]: + return True + return False + + def init_score_map(): + test = libpr.init_score_map() + if test: + raise RuntimeError("Malloc error in init_score_map().") + for i in [i for i in range(world_db["MAP_LENGTH"] ** 2) + if '.' == chr(t["T_MEMMAP"][i])]: + set_map_score(i, 65535 - 1) + if "a" == filter: + for id in world_db["Things"]: + Thing = world_db["Things"][id] + pos = Thing["T_POSY"] * world_db["MAP_LENGTH"] + Thing["T_POSX"] + if t != Thing and Thing["T_LIFEPOINTS"] and \ + t["T_TYPE"] != Thing["T_TYPE"] and \ + 'v' == chr(t["fovmap"][pos]) and \ + t["T_LIFEPOINTS"] > \ + world_db["ThingTypes"][Thing["T_TYPE"]]["TT_LIFEPOINTS"]: + set_map_score(pos, 0) + elif t["T_TYPE"] == Thing["T_TYPE"]: + set_map_score(pos, 65535) + elif "f" == filter: + for id in [id for id in world_db["Things"] + if world_db["Things"][id]["T_LIFEPOINTS"]]: + Thing = world_db["Things"][id] + pos = Thing["T_POSY"] * world_db["MAP_LENGTH"] + Thing["T_POSX"] + if t["T_TYPE"] != Thing["T_TYPE"] and \ + 'v' == chr(t["fovmap"][pos]) and \ + t["T_LIFEPOINTS"] <= \ + world_db["ThingTypes"][Thing["T_TYPE"]]["TT_LIFEPOINTS"]: + set_map_score(pos, 0) + elif "c" == filter: + for mt in [mt for mt in t["T_MEMTHING"] + if ' ' != chr(t["T_MEMMAP"][mt[1] + * world_db["MAP_LENGTH"] + + mt[2]]) + if world_db["ThingTypes"][mt[0]]["TT_CONSUMABLE"]]: + set_map_score(mt[1] * world_db["MAP_LENGTH"] + mt[2], 0) + elif "s" == filter: + for i in [i for i in range(world_db["MAP_LENGTH"] ** 2) + if t["T_MEMDEPTHMAP"][i] == mem_depth_c[0]]: + set_map_score(i, 0) + + def rand_target_dir(neighbors, cmp, dirs): + candidates = [] + n_candidates = 0 + for i in range(len(dirs)): + if cmp == neighbors[i]: + candidates.append(dirs[i]) + n_candidates += 1 + return candidates[rand.next() % n_candidates] if n_candidates else 0 + + def get_neighbor_scores(dirs, eye_pos): + scores = [] + if libpr.ready_neighbor_scores(eye_pos): + raise RuntimeError("No score map allocated for " + + "ready_neighbor_scores.()") + for i in range(len(dirs)): + scores.append(libpr.get_neighbor_score(i)) + return scores + + def get_dir_from_neighbors(): + dir_to_target = False + dirs = "edcxsw" + eye_pos = t["T_POSY"] * world_db["MAP_LENGTH"] + t["T_POSX"] + neighbors = get_neighbor_scores(dirs, eye_pos) + if "f" == filter: + inhabited = [world_db["Things"][id]["T_POSY"] + * world_db["MAP_LENGTH"] + + world_db["Things"][id]["T_POSX"] + for id in world_db["Things"] + if world_db["Things"][id]["T_LIFEPOINTS"]] + for i in range(len(dirs)): + mv_yx_in_dir_legal(dirs[i], t["T_POSY"], t["T_POSX"]) + pos_cmp = libpr.result_y() * world_db["MAP_LENGTH"] \ + + libpr.result_x() + for pos in [pos for pos in inhabited if pos == pos_cmp]: + neighbors[i] = 65535 + break + minmax_start = 0 if "f" == filter else 65535 - 1 + minmax_neighbor = minmax_start + for i in range(len(dirs)): + if ("f" == filter and get_map_score(eye_pos) < neighbors[i] and + minmax_neighbor < neighbors[i] and 65535 != neighbors[i]) \ + or ("f" != filter and minmax_neighbor > neighbors[i]): + minmax_neighbor = neighbors[i] + if minmax_neighbor != minmax_start: + dir_to_target = rand_target_dir(neighbors, minmax_neighbor, dirs) + if "f" == filter: + if not dir_to_target: + if 1 == get_map_score(eye_pos): + dir_to_target = rand_target_dir(neighbors, 0, dirs) + elif 3 >= get_map_score(eye_pos): + t["T_COMMAND"] = [id for id in world_db["ThingActions"] + if world_db["ThingActions"][id]["TA_NAME"] + == "wait"][0] + return 1; + elif dir_to_target and 3 < get_map_score(eye_pos): + dir_to_target = 0 + elif "a" == filter and 10 <= get_map_score(eye_pos): + dir_to_target = 0 + return dir_to_target + + dir_to_target = False + mem_depth_c = b' ' + run_i = 9 + 1 if "s" == filter else 1 + while run_i and not dir_to_target and ("s" == filter or seeing_thing()): + run_i -= 1 + init_score_map() + mem_depth_c = b'9' if b' ' == mem_depth_c \ + else bytes([mem_depth_c[0] - 1]) + if libpr.dijkstra_map(): + raise RuntimeError("No score map allocated for dijkstra_map().") + dir_to_target = get_dir_from_neighbors() + libpr.free_score_map() + if dir_to_target and str == type(dir_to_target): + t["T_COMMAND"] = [id for id in world_db["ThingActions"] + if world_db["ThingActions"][id]["TA_NAME"] + == "move"][0] + t["T_ARGUMENT"] = ord(dir_to_target) + return dir_to_target def standing_on_consumable(t): @@ -799,7 +978,7 @@ def ai(t): """ t["T_COMMAND"] = [id for id in world_db["ThingActions"] if world_db["ThingActions"][id]["TA_NAME"] == "wait"][0] - if not get_dir_to_nearest_target(t, "f"): + if not get_dir_to_target(t, "f"): sel = get_inventory_slot_to_consume(t) if -1 != sel: t["T_COMMAND"] = [id for id in world_db["ThingActions"] @@ -810,9 +989,9 @@ def ai(t): t["T_COMMAND"] = [id for id in world_db["ThingActions"] if world_db["ThingActions"][id]["TA_NAME"] == "pick_up"][0] - elif (not get_dir_to_nearest_target(t, "c")) and \ - (not get_dir_to_nearest_target(t, "a")): - get_dir_to_nearest_target(t, "s") + elif (not get_dir_to_target(t, "c")) and \ + (not get_dir_to_target(t, "a")): + get_dir_to_target(t, "s") def turn_over(): @@ -832,7 +1011,6 @@ def turn_over(): whilebreaker = True break ai(Thing) - Thing["T_COMMAND"] = 1 try_healing(Thing) Thing["T_PROGRESS"] += 1 taid = [a for a in world_db["ThingActions"]