import numpy as np
from state import create_empty_state, to_state_string
from machine_player import MachinePlayer
from human_player import HumanPlayer
from player_provider import PlayerProvider
import json

float_formatter = "{:.3f}".format
np.set_printoptions(formatter={'float_kind': float_formatter})

REPORT_INTERVAL = 1000
report_counter = REPORT_INTERVAL


def report(k):
	global report_counter
	if report_counter > 0:
		report_counter -= 1
	else:
		report_counter = REPORT_INTERVAL-1
		print(f"Iteration: {k:8d}")


def play(player_provider: PlayerProvider, k_max=10000, with_debug=False):
	player_provider.set_debug(with_debug)
	for k in range(0, k_max):
		report(k)
		players = player_provider.choose()
		state = create_empty_state()
		move = 1
		run = True
		other_player = players[-1]
		while run:
			for player in players:
				if with_debug:
					print(f"---------------------------------------------------")
					print(f"- Game {k:06d}, Move {move} -----------------------------")
					print(f"---------------------------------------------------")
				last_state = state
				state, has_moved = player.move(state)
				if with_debug:
					print(to_state_string(last_state, state))
				if not has_moved:
					if with_debug:
						print(f"{player.mark}: No more moves")
						player.reward(0.0)
						other_player.reward(0.0)
					run = False
				if player.has_won(state):
					if with_debug:
						print(f"{player.mark}: Has won the game")
						player.reward(1.0)
						other_player.reward(0.0)
					run = False

				other_player = player
				if not run:
					break

				move += 1


with open("values_x.json", "r") as fp:
	x_values = json.load(fp)

with open("values_o.json", "r") as fp:
	o_values = json.load(fp)

do_training = 1
do_human_player = 1

p_hx = HumanPlayer(mark='X', name="Jens")
p_mx = MachinePlayer(mark='X', params=MachinePlayer.Params(p_explore=0.2, alpha=0.1), values=x_values)
p_mo = MachinePlayer(mark='O', params=MachinePlayer.Params(p_explore=0.2, alpha=0.1), values=o_values)

if do_training:
	play(PlayerProvider(p_mx, p_mo), 10000, False)

	# Values after training
	# Convert and write JSON object to file
	with open("values_x.json", "w") as fp:
		json.dump(p_mx.values, fp, indent=0)

	with open("values_o.json", "w") as fp:
		json.dump(p_mo.values, fp, indent=0)

if do_human_player:
	pp = PlayerProvider(p_hx, p_mo)
else:
	pp = PlayerProvider(p_mx, p_mo)

play(pp, 10, True)