scripts/calculate_annotator_audio_minutes.py

#!/usr/bin/env python3
"""
Script to calculate total minutes of audio data assigned to each annotator.

This script queries the database to find all audio files assigned to each annotator
through AnnotationInterval ranges, loads the actual audio files to calculate their
durations, and reports the total minutes per annotator.
"""

import argparse
import sys
import os
import time
from typing import Dict, List, Tuple
from sqlalchemy import and_
from sqlalchemy.exc import OperationalError

# Add project root to Python path
project_root = os.path.abspath(os.path.join(os.path.dirname(__file__), '..'))
if project_root not in sys.path:
    sys.path.insert(0, project_root)

from utils.database import get_db, get_db_readonly
from utils.cloud_server_audio_loader import CloudServerAudioLoader
from data.models import Annotator, AnnotationInterval, TTSData
from utils.logger import Logger
from utils.sentry_integration import capture_custom_event
import sentry_sdk
from config import conf

log = Logger()

def get_assigned_tts_data_for_annotator(db, annotator_id: int) -> List[TTSData]:
    """
    Get all TTSData items assigned to a specific annotator through AnnotationInterval ranges.
    
    Args:
        db: Database session
        annotator_id: ID of the annotator
        
    Returns:
        List of TTSData objects assigned to the annotator
    """
    max_retries = 3
    retry_delay = 5  # seconds
    
    for attempt in range(max_retries):
        try:
            # Get all annotation intervals for this annotator
            intervals = db.query(AnnotationInterval).filter(
                AnnotationInterval.annotator_id == annotator_id
            ).all()
            
            if not intervals:
                return []
            
            # Collect all TTSData IDs within the assigned ranges
            assigned_tts_data = []
            for interval in intervals:
                if interval.start_index is not None and interval.end_index is not None:
                    tts_data_in_range = db.query(TTSData).filter(
                        and_(
                            TTSData.id >= interval.start_index,
                            TTSData.id <= interval.end_index
                        )
                    ).all()
                    assigned_tts_data.extend(tts_data_in_range)
            
            return assigned_tts_data
            
        except OperationalError as e:
            if "Lost connection to MySQL server" in str(e) and attempt < max_retries - 1:
                log.warning(f"Database connection lost, retrying in {retry_delay} seconds... (attempt {attempt + 1}/{max_retries})")
                time.sleep(retry_delay)
                # Refresh the database session
                db.rollback()
                continue
            else:
                raise

def calculate_audio_duration_seconds(filename: str, loader: CloudServerAudioLoader) -> float:
    """
    Calculate the duration of an audio file in seconds.
    
    Args:
        filename: Name of the audio file
        loader: CloudServerAudioLoader instance
        
    Returns:
        Duration in seconds, or 0.0 if file cannot be loaded
    """
    try:
        sample_rate, samples = loader.load_audio(filename)
        # Calculate duration in seconds
        if samples.ndim == 1:
            # Mono audio
            duration_seconds = len(samples) / sample_rate
        else:
            # Multi-channel audio - use length of first channel
            duration_seconds = samples.shape[0] / sample_rate
        
        return duration_seconds
    except Exception as e:
        log.warning(f"Failed to load audio file '{filename}': {e}")
        sentry_sdk.capture_exception(e, extra={
            'operation': 'calculate_audio_duration',
            'filename': filename
        })
        return 0.0

def calculate_annotator_audio_minutes(annotator_name: str = None):
    """
    Calculate and report the total minutes of audio assigned to each annotator.
    
    Args:
        annotator_name: Optional name of specific annotator to calculate for
    """
    try:
        # Initialize audio loader
        loader = CloudServerAudioLoader(conf.FTP_URL)
        
        # First, get the annotators list with a fresh connection
        annotator_data = []
        with get_db_readonly() as db:
            # Get annotators based on filter
            if annotator_name:
                annotators = db.query(Annotator).filter(
                    Annotator.is_active == True,
                    Annotator.name == annotator_name
                ).all()
                if not annotators:
                    log.error(f"No active annotator found with name: {annotator_name}")
                    return
            else:
                annotators = db.query(Annotator).filter(Annotator.is_active == True).all()
            
            # Extract the data we need before the session closes
            annotator_data = [(ann.id, ann.name) for ann in annotators]
        
        if not annotator_data:
            log.info("No active annotators found.")
            return
        
        log.info("--- Annotator Audio Duration Report ---")
        log.info("Calculating total minutes of assigned audio per annotator...")
        log.info("")
        
        total_annotators = len(annotator_data)
        annotator_results = []
        
        for idx, (annotator_id, annotator_name) in enumerate(annotator_data, 1):
            log.info(f"Processing annotator {idx}/{total_annotators}: {annotator_name} (ID: {annotator_id})")
            
            # Get assigned TTSData for this annotator with a fresh connection
            assigned_tts_data = []
            with get_db_readonly() as db:
                assigned_tts_data = get_assigned_tts_data_for_annotator(db, annotator_id)
            
            if not assigned_tts_data:
                log.info(f"  No audio files assigned to {annotator_name}")
                annotator_results.append((annotator_name, 0, 0.0))
                continue
            
            total_duration_seconds = 0.0
            successful_files = 0
            failed_files = 0
            
            log.info(f"  Calculating duration for {len(assigned_tts_data)} assigned audio files...")
            
            # Calculate duration for each assigned audio file
            for tts_data in assigned_tts_data:
                duration = calculate_audio_duration_seconds(tts_data.filename, loader)
                if duration > 0:
                    total_duration_seconds += duration
                    successful_files += 1
                else:
                    failed_files += 1
            
            total_minutes = total_duration_seconds / 60.0
            
            log.info(f"  Successfully processed: {successful_files} files")
            if failed_files > 0:
                log.warning(f"  Failed to process: {failed_files} files")
            log.info(f"  Total duration: {total_duration_seconds:.2f} seconds ({total_minutes:.2f} minutes)")
            
            annotator_results.append((annotator_name, len(assigned_tts_data), total_minutes))
            log.info("")
            
            # Print summary report
            log.info("=" * 60)
            log.info("SUMMARY REPORT")
            log.info("=" * 60)
            log.info(f"{'Annotator':<20} {'Files':<8} {'Minutes':<12} {'Hours':<8}")
            log.info("-" * 60)
            
            total_files = 0
            total_minutes = 0.0
            
            for annotator_name, file_count, minutes in annotator_results:
                hours = minutes / 60.0
                log.info(f"{annotator_name:<20} {file_count:<8} {minutes:<12.2f} {hours:<8.2f}")
                total_files += file_count
                total_minutes += minutes
            
            log.info("-" * 60)
            total_hours = total_minutes / 60.0
            log.info(f"{'TOTAL':<20} {total_files:<8} {total_minutes:<12.2f} {total_hours:<8.2f}")
            log.info("=" * 60)
            
            # Capture analytics event
            capture_custom_event(
                'annotator_audio_calculation_completed',
                {
                    'total_annotators': total_annotators,
                    'total_files_processed': total_files,
                    'total_minutes': total_minutes,
                    'total_hours': total_hours
                }
            )

    except Exception as e:
        log.error(f"Failed to calculate annotator audio minutes: {e}")
        sentry_sdk.capture_exception(e, extra={
            'operation': 'calculate_annotator_audio_minutes'
        })
        raise

def main():
    """Main entry point for the script."""
    parser = argparse.ArgumentParser(
        description="Calculate total minutes of audio data assigned to each annotator"
    )
    parser.add_argument(
        '--annotator',
        type=str,
        help="Calculate for a specific annotator by name (optional, calculates for all if not specified)"
    )
    
    args = parser.parse_args()
    
    if args.annotator:
        log.info(f"Calculating audio minutes for annotator: {args.annotator}")
        calculate_annotator_audio_minutes(args.annotator)
    else:
        log.info("Calculating audio minutes for all annotators")
        calculate_annotator_audio_minutes()

if __name__ == "__main__":
    main()