import subprocess
import tempfile
import os
from pathlib import Path
from typing import List, Tuple, Optional

import numpy as np
from scipy.io import wavfile

class ConfidenceError(Exception):
    """Raised when an area is found but the energy difference is too low to be confident."""
    pass

class MissingDetectionError(Exception):
    """Raised when the script fails to find both an intro and an outro."""
    pass

def get_video_duration(file: Path) -> float:
    """Uses ffprobe to get the total duration of the video."""
    cmd = [
        "ffprobe", 
        "-v", "error", 
        "-show_entries", "format=duration", 
        "-of", "default=noprint_wrappers=1:nokey=1", 
        str(file)
    ]
    result = subprocess.run(cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE, text=True)
    try:
        return float(result.stdout.strip())
    except ValueError:
        raise RuntimeError(f"Could not determine duration for {file}")

def extract_audio_segment_tmp(
    file: Path,
    start_or_section: float | tuple[float, float],
    end: Optional[float] = None,
    audio_stream: Optional[int] = None
) -> Path:
    """Extracts audio using FFmpeg and saves it to a temporary WAV file."""
    if isinstance(start_or_section, tuple):
        if len(start_or_section) != 2:
            raise ValueError("Tuple must strictly be (start, end)")
        start = float(start_or_section[0])
        duration = float(start_or_section[1])
    elif isinstance(start_or_section, (int, float)):
        if end is None:
            raise ValueError("The 'end' parameter must be provided when using a separate start float.")
        start = float(start_or_section)
        end_time = float(end)
        duration = end_time - start
    else:
        raise TypeError("Invalid type for start_or_section. Expected float or tuple.")
    
    if duration <= 0:
        raise ValueError("Duration must be above 0")

    tmp = tempfile.NamedTemporaryFile(suffix=".wav", delete=False)
    tmp_path = Path(tmp.name)
    tmp.close()

    cmd = [
        "ffmpeg", "-y",
        "-i", str(file),
        "-ss", str(start),
        "-t", str(duration)
    ]

    if audio_stream is not None:
        cmd += ["-map", f"0:a:{audio_stream}"]
    else:
        cmd += ["-map", "0:a:0"]

    cmd += [
        "-vn",
        "-acodec", "pcm_s16le",
        "-ar", "44100",
        "-ac", "2",
        str(tmp_path)
    ]

    subprocess.run(cmd, stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL, check=True)
    return tmp_path

def find_high_power_segment(wav_path: Path, expected_length: float, confidence_threshold: float = 1.0) -> float:
    rate, data = wavfile.read(wav_path)
    
    if len(data.shape) > 1:
        data = data.mean(axis=1)
        
    data = data.astype(np.float32) / (np.max(np.abs(data)) + 1e-9)
    
    chunk_size = rate 
    num_chunks = len(data) // chunk_size
    
    # Calculate RMS (Volume) and Standard Deviation (Variation) for every second
    powers = np.zeros(num_chunks)
    variations = np.zeros(num_chunks)
    
    for i in range(num_chunks):
        chunk = data[i * chunk_size : (i + 1) * chunk_size]
        powers[i] = np.sqrt(np.mean(chunk**2))
        variations[i] = np.std(chunk)

    window_size = int(expected_length)
    if len(powers) < window_size:
        raise ValueError("Audio segment too short.")

    # SCORING ALGORITHM:
    # We want High Power and Low Variation (that "solid block" look)
    # We invert variation so that a small variation results in a higher score.
    scores = []
    for i in range(len(powers) - window_size + 1):
        win_power = np.mean(powers[i : i + window_size])
        win_var = np.mean(variations[i : i + window_size])
        
        # The score increases with power and decreases as the variation increases
        # We add a small constant to win_var to avoid division by zero
        score = win_power / (win_var + 0.05) 
        scores.append(score)

    best_idx = np.argmax(scores)
    
    # Confidence check: is our best score significantly better than the average score?
    avg_score = np.mean(scores)
    if scores[best_idx] < avg_score * confidence_threshold:
        raise ConfidenceError("Detection unreliable: no distinct high-power constant block found.")
        
    return float(best_idx)

def get_intro_outro(file: Path, length: float = 90.0) -> List[Tuple[float, float]]:
    """
    Finds the intro and outro of a video file by analyzing audio power peaks.
    Searches the first 8 minutes and the last 8 minutes.
    """
    results = []
    total_duration = get_video_duration(file)
    search_window = 8 * 60.0  # 8 minutes
    
    # --- 1. FIND INTRO ---
    intro_search_dur = min(search_window, total_duration)
    intro_wav = extract_audio_segment_tmp(file, 0.0, intro_search_dur)
    
    try:
        intro_start_rel = find_high_power_segment(intro_wav, length)
        results.append((intro_start_rel, length))
    except ConfidenceError:
        intro_wav.unlink()
        raise ConfidenceError("Could not confidently detect the Intro.")
    finally:
        if intro_wav.exists():
            intro_wav.unlink() # Cleanup temp file
            
    # --- 2. FIND OUTRO ---
    outro_search_start = max(0.0, total_duration - search_window)
    outro_search_dur = total_duration - outro_search_start
    outro_wav = extract_audio_segment_tmp(file, outro_search_start, total_duration)
    
    try:
        outro_start_rel = find_high_power_segment(outro_wav, length)
        # Add the 8-minute offset back to get the absolute timestamp in the video
        outro_start_abs = outro_search_start + outro_start_rel
        results.append((outro_start_abs, length))
    except ConfidenceError:
        outro_wav.unlink()
        raise ConfidenceError("Could not confidently detect the Outro.")
    finally:
        if outro_wav.exists():
            outro_wav.unlink() # Cleanup temp file

    # Ensure we got both
    if len(results) < 2:
        raise MissingDetectionError("Failed to find both an intro and an outro.")

    return results

# --- Example Usage ---
if __name__ == "__main__":
    try:
        # Requires: pip install numpy scipy
        file_path = Path("/home/honney/mount/Media/tmp/serien/ready/Charlotte/Season 01/[NO CHAPTER]_E04.mkv")
        if file_path.exists():
            timestamps = get_intro_outro(file_path, length=90.0)
            print("Found Intro and Outro [Start, Duration]:", timestamps)
    except Exception as e:
        print(f"Error: {e}")