backend/utils.py

import os
import shutil
from pydub import AudioSegment

RAVDESS_MAP = {
    '01': 'neutral', '02': 'calm', '03': 'happiness', '04': 'sadness',
    '05': 'anger', '06': 'fear', '07': 'disgust', '08': 'surprise'
}

KNOWN_EMOTIONS = ['anger', 'neutral', 'sadness', 'happiness', 'fear', 'disgust', 'surprise', 'calm']
# Handling some common synonyms/variations
EMOTION_ALIASES = {
    'happy': 'happiness',
    'sad': 'sadness',
    'angry': 'anger',
    'surprised': 'surprise'
}

def convert_to_wav(source_path, target_path):
    """Converts audio file to WAV format."""
    try:
        audio = AudioSegment.from_file(source_path)
        audio.export(target_path, format="wav")
        return True
    except Exception as e:
        print(f"Error converting {source_path}: {e}")
        return False

def get_label_from_filename(filename):
    """Extracts emotion label from filename based on patterns."""
    filename = os.path.basename(filename).lower()

    # Rule 1: RAVDESS dataset (e.g., 03-01-01-01-01-01-01.wav)
    if filename.count('-') == 6 and filename.startswith('03'):
        parts = filename.replace('.wav', '').split('-')
        if len(parts) > 2:
            return RAVDESS_MAP.get(parts[2])

    # Rule 2: General emotion words
    # Check for exact matches or surrounded by underscores/dots
    for emotion in KNOWN_EMOTIONS + list(EMOTION_ALIASES.keys()):
        # Simple containment check might be too loose (e.g., "fearless"), 
        # but matches the original script's logic roughly.
        # Improving it with delimiters for safety.
        patterns = [
            f"_{emotion}_", 
            f"{emotion}_", 
            f"_{emotion}.", 
            f"upload_{emotion}_"
        ]
        
        # Also check if it STARTS with the emotion (common in some datasets)
        if filename.startswith(f"{emotion}_") or any(p in filename for p in patterns):
            normalized = EMOTION_ALIASES.get(emotion, emotion)
            return normalized
            
    return None