concat_surahs.py

#!/usr/bin/env python3
"""
Quranic Recitation Surah Concatenator
=====================================
Converts verse-level Opus and JSON files into single Surah-level files
for all 22 reciters.

Output format: {reciter_slug}/{surah_id}/{surah_id}.opus
"""

import os
import glob
import json
import struct
import subprocess
from pathlib import Path
from concurrent.futures import ProcessPoolExecutor, as_completed
import time

def get_ogg_opus_duration(filepath):
    """Fast extraction of Ogg Opus duration in milliseconds."""
    try:
        with open(filepath, 'rb') as f:
            f.seek(0, 2)
            file_size = f.tell()
            
            chunk_size = min(65536, file_size)
            f.seek(file_size - chunk_size)
            data = f.read(chunk_size)
            
            last_oggs_idx = data.rfind(b'OggS')
            if last_oggs_idx == -1:
                return 0
            
            granule_bytes = data[last_oggs_idx+6 : last_oggs_idx+14]
            granule_position = struct.unpack('<q', granule_bytes)[0]
            
            # Granule position in Opus is ALWAYS at 48000 Hz
            return int((granule_position / 48000.0) * 1000)
    except Exception:
        return 0

def process_surah(args):
    reciter_dir, surah_str = args
    surah_path = Path(reciter_dir) / surah_str
    
    # Match verse files: e.g. 001001.opus (length > 3 before extension)
    all_opus = sorted(list(surah_path.glob(f"{surah_str}*.opus")))
    verse_opus_files = [f for f in all_opus if len(f.stem) > 3]
    
    if not verse_opus_files:
        return True
    
    target_opus = surah_path / f"{surah_str}.opus"
    target_json = surah_path / f"{surah_str}.json"
    concat_txt_path = surah_path / f"concat_{surah_str}.txt"
    
    # 1. Compute JSON and write concat file
    combined_json = []
    cumulative_offset = 0
    verse_json_files = []
    
    with open(concat_txt_path, 'w') as f_out:
        for opus_f in verse_opus_files:
            f_out.write(f"file '{opus_f.name}'\n")
            
            json_f = opus_f.with_suffix('.json')
            if json_f.exists():
                verse_json_files.append(json_f)
                with open(json_f, 'r') as jf:
                    data = json.load(jf)
                # Shift timings
                if isinstance(data, list):
                    for seg in data:
                        new_seg = [
                            seg[0],
                            seg[1],
                            int(seg[2]) + cumulative_offset,
                            int(seg[3]) + cumulative_offset
                        ]
                        combined_json.append(new_seg)
            
            duration = get_ogg_opus_duration(opus_f)
            cumulative_offset += duration
            
    # Write combined JSON
    with open(target_json, 'w') as jf:
        # Format like a compact 2D array
        if combined_json:
            inner_strs = [json.dumps(seg) for seg in combined_json]
            json_str = '[\n  ' + ',\n  '.join(inner_strs) + '\n]'
            jf.write(json_str)
        else:
            jf.write('[]')
        
    # 2. Concat Opus Losslessly using ffmpeg
    success = False
    try:
        subprocess.run([
            "ffmpeg", "-y", "-hide_banner", "-loglevel", "error",
            "-f", "concat", "-safe", "0",
            "-i", str(concat_txt_path),
            "-c", "copy",
            str(target_opus)
        ], check=True)
        success = True
    except subprocess.CalledProcessError:
        return False
    finally:
        if concat_txt_path.exists():
            concat_txt_path.unlink()
            
    # 3. Clean up verse-level files if successful
    if success and target_opus.exists():
        for opus_f in verse_opus_files:
            try: opus_f.unlink()
            except: pass
        for json_f in verse_json_files:
            try: json_f.unlink()
            except: pass
            
    return success

def main():
    print("╔══════════════════════════════════════════════════════╗")
    print("║   Surah Concatenator                                 ║")
    print("║   Verse Opus/JSON -> Surah Opus/JSON                 ║")
    print("╚══════════════════════════════════════════════════════╝")
    
    base_dir = Path(".")
    reciters = []
    
    # Identify valid reciter directories
    for entry in base_dir.iterdir():
        if entry.is_dir() and not entry.name.startswith('.') and entry.name not in ('audio', 'wbw_tester', '__pycache__'):
            # Only include if it has surah directories like '001'
            if (entry / '001').exists():
                reciters.append(entry.name)
                
    reciters = sorted(reciters)
    print(f"✓ Found {len(reciters)} reciters to process.")
    
    tasks = []
    for r in reciters:
        for s in range(1, 115):
            tasks.append((r, f"{s:03d}"))
            
    print(f"Total tasks (surahs): {len(tasks)}")
    print(f"Starting parallel processing...")
    
    completed = 0
    errors = 0
    start = time.time()
    
    with ProcessPoolExecutor(max_workers=min(os.cpu_count() or 4, 16)) as executor:
        futures = [executor.submit(process_surah, t) for t in tasks]
        for f in as_completed(futures):
            if not f.result():
                errors += 1
            completed += 1
            if completed % 114 == 0 or completed == len(tasks):
                print(f"  Processed {completed}/{len(tasks)} surahs... ({errors} errors)")
                
    elapsed = time.time() - start
    print(f"\n✓ Done! Converted {len(reciters)} reciters in {elapsed:.1f}s.")

if __name__ == '__main__':
    main()