src/utils.py

"""
Utility functions for MUSEval Leaderboard
"""

import pandas as pd
import numpy as np
from typing import Dict, List, Any, Optional

def norm_sNavie(df: pd.DataFrame) -> pd.DataFrame:
    """Normalize dataframe using naive normalization"""
    # Simple normalization - keep as is for now
    return df

def pivot_df(file_path: str, tab_name: str) -> pd.DataFrame:
    """Pivot dataframe from file"""
    try:
        df = pd.read_csv(file_path)
        # Simple pivot - return as is for now
        return df
    except Exception as e:
        print(f"Error reading {file_path}: {e}")
        return pd.DataFrame()

def get_grouped_dfs() -> Dict[str, pd.DataFrame]:
    """Get grouped dataframes for different views"""
    from .load_results import load_results_with_metadata, create_overall_table
    
    # Load results
    results = load_results_with_metadata()
    if not results:
        return {
            'domain': pd.DataFrame(),
            'frequency': pd.DataFrame(), 
            'term_length': pd.DataFrame(),
            'univariate': pd.DataFrame(),
            'overall': pd.DataFrame()
        }
    
    # Create overall dataframe
    overall_df = create_overall_table()
    
    # For now, return the same dataframe for all views
    # In a real implementation, these would be different aggregations
    return {
        'domain': overall_df.copy(),
        'frequency': overall_df.copy(),
        'term_length': overall_df.copy(), 
        'univariate': overall_df.copy(),
        'overall': overall_df.copy()
    }

def pivot_existed_df(df: pd.DataFrame, tab_name: str) -> pd.DataFrame:
    """Pivot existing dataframe"""
    if df.empty:
        return df
    
    # Add tab name as a column for identification
    df_copy = df.copy()
    df_copy['tab'] = tab_name
    return df_copy

def rename_metrics(df: pd.DataFrame) -> pd.DataFrame:
    """Rename metrics columns"""
    if df.empty:
        return df
    
    # Add rank column based on MAE
    if 'MAE' in df.columns:
        df_copy = df.copy()
        df_copy['MASE_Rank'] = df_copy['MAE'].rank(method='min')
        return df_copy
    
    return df

def format_df(df: pd.DataFrame) -> pd.DataFrame:
    """Format dataframe for display"""
    if df.empty:
        return df
    
    df_copy = df.copy()
    
    # Format numeric columns
    numeric_cols = ['MAE', 'Uni-MAE', 'RMSE', 'MAPE', 'R²', 'SMAPE', 'Uni-Multi']
    for col in numeric_cols:
        if col in df_copy.columns:
            if col in ['MAPE', 'SMAPE']:
                df_copy[col] = df_copy[col].apply(lambda x: f"{x:.1f}%" if pd.notna(x) else "")
            else:
                df_copy[col] = df_copy[col].apply(lambda x: f"{x:.3f}" if pd.notna(x) else "")
    
    return df_copy