import cv2
import mediapipe as mp
import numpy as np
import math, time
from pythonosc import udp_client

# -------------------------------
# SETTINGS
# -------------------------------
TOUCH_CAM_INDEX = 1        # deine Touch-Kamera
GESTURE_CAM_INDEX = 0      # deine Clap / Gesture Kamera

GAME_SCREEN_WIDTH  = 900   # muss zu deinem Pygame-Fenster passen!
GAME_SCREEN_HEIGHT = 600

client = udp_client.SimpleUDPClient("127.0.0.1", 5005)

# -------------------------------
# MAIN FUNCTION
# -------------------------------
def run_gesture_input():

    mp_hands = mp.solutions.hands
    mp_draw = mp.solutions.drawing_utils

    hands_touch = mp_hands.Hands(max_num_hands=1, min_detection_confidence=0.6)
    hands_gesture = mp_hands.Hands(max_num_hands=2, min_detection_confidence=0.6)

    # Kameras öffnen
    cam_touch   = cv2.VideoCapture(TOUCH_CAM_INDEX)
    cam_gesture = cv2.VideoCapture(GESTURE_CAM_INDEX)

    if not cam_touch.isOpened():
        print("❌ Touch-Kamera konnte NICHT geöffnet werden!")
    else:
        print(f"✅ Touch-Kamera geöffnet (Index {TOUCH_CAM_INDEX})")

    if not cam_gesture.isOpened():
        print("❌ Gesture-Kamera konnte NICHT geöffnet werden!")
    else:
        print(f"✅ Gesture-Kamera geöffnet (Index {GESTURE_CAM_INDEX})")

    prev_clap_time = 0
    clap_cooldown = 1.5

    while True:

        ok1, frame_touch = cam_touch.read()
        ok2, frame_gest  = cam_gesture.read()

        if not ok1 or not ok2:
            print("⚠️ Eine Kamera liefert kein Bild.")
            break

        frame_touch = cv2.flip(frame_touch, 1)
        frame_gest = cv2.flip(frame_gest, 1)

        # ---------------------------------------
        # TOUCH (Zeigefinger) ohne Kalibrierung
        # ---------------------------------------
        rgb_t = cv2.cvtColor(frame_touch, cv2.COLOR_BGR2RGB)
        res_t = hands_touch.process(rgb_t)

        th, tw, _ = frame_touch.shape

        if res_t.multi_hand_landmarks:
            lm = res_t.multi_hand_landmarks[0]
            mp_draw.draw_landmarks(frame_touch, lm, mp_hands.HAND_CONNECTIONS)

            fx = int(lm.landmark[8].x * tw)
            fy = int(lm.landmark[8].y * th)

            # einfache Skalierung auf dein Spiel-Fenster
            sx = int(fx * (GAME_SCREEN_WIDTH / tw))
            sy = int(fy * (GAME_SCREEN_HEIGHT / th))

            # Finger unten? (Touch)
            if lm.landmark[8].y > 0.8:
                client.send_message("/touch", [sx, sy])
                cv2.putText(frame_touch, f"Touch {sx},{sy}", (40, 60),
                            cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0, 255, 0), 2)

        # ---------------------------------------
        # CLAP (zwei Hände)
        # ---------------------------------------
        rgb_g = cv2.cvtColor(frame_gest, cv2.COLOR_BGR2RGB)
        res_g = hands_gesture.process(rgb_g)

        gh, gw, _ = frame_gest.shape

        if res_g.multi_hand_landmarks and len(res_g.multi_hand_landmarks) == 2:
            h1, h2 = res_g.multi_hand_landmarks

            x1 = np.mean([p.x for p in h1.landmark]) * gw
            y1 = np.mean([p.y for p in h1.landmark]) * gh
            x2 = np.mean([p.x for p in h2.landmark]) * gw
            y2 = np.mean([p.y for p in h2.landmark]) * gh

            dist = math.hypot(x2 - x1, y2 - y1)

            if dist < 100 and (time.time() - prev_clap_time) > clap_cooldown:
                prev_clap_time = time.time()
                client.send_message("/clap", 1)
                cv2.putText(frame_gest, "👏", (int(gw/2)-20, 80),
                            cv2.FONT_HERSHEY_SIMPLEX, 2, (0, 255, 255), 3)

        cv2.imshow("Touch-Cam", frame_touch)
        cv2.imshow("Gesture-Cam", frame_gest)

        if cv2.waitKey(5) & 0xFF == 27:
            break

    cam_touch.release()
    cam_gesture.release()
    cv2.destroyAllWindows()


if __name__ == "__main__":
    run_gesture_input()