import json
import os

import pandas as pd
from typing import List

from src.display.formatting import has_no_nan_values, make_clickable_model
from src.display.utils import AutoEvalColumn

# Import SAGE-specific modules - avoid transformers dependency
process_sage_results_for_leaderboard = None
try:
    # Import SAGE modules without triggering transformers dependency
    import sys
    import os
    import json
    from dataclasses import dataclass
    from typing import Dict, List, Any
    import numpy as np
    
    # Copy SAGEResult class locally to avoid import issues
    @dataclass
    class SAGEResult:
        submission_id: str
        organization: str
        email: str
        tokens: str
        accuracy: float
        mg_pass_2: float
        mg_pass_4: float
        submitted_time: str
        status: str = "EVALUATED"
        
        def to_dict(self):
            """Converts the SAGE Result to a dict compatible with our dataframe display"""
            # Extract model name from submission_id or use model_name directly
            if hasattr(self, 'model_name'):
                model_name = self.model_name
            elif self.submission_id.startswith("oss_"):
                # Extract model name from submission_id
                model_name = self.submission_id.split("_", 2)[-1].replace("_", " ")
            else:
                model_name = self.submission_id
            
            # Create display name
            display_name = f"**{model_name}**"
            model_symbol = "🤖"
            
            # Format date to YYYY-MM-DD only
            formatted_date = self.submitted_time
            if isinstance(self.submitted_time, str):
                # Try to parse and reformat date
                try:
                    if 'T' in self.submitted_time:
                        # ISO format like "2025-09-09T14:37:23.616340"
                        formatted_date = self.submitted_time.split('T')[0]
                    else:
                        # Already in simple format
                        formatted_date = self.submitted_time.split(' ')[0]
                except:
                    formatted_date = self.submitted_time
            
            data_dict = {
                "Model": display_name,
                "Organization": self.organization,
                "Accuracy (%)": round(self.accuracy, 2),
                "mG-Pass@2 (%)": round(self.mg_pass_2, 2),
                "mG-Pass@4 (%)": round(self.mg_pass_4, 2),
                "Submission Date": formatted_date,
            }
            
            return data_dict
    
    def load_initial_sage_results_from_oss() -> List[SAGEResult]:
        """Load initial SAGE results from OSS"""
        sage_results = []
        
        try:
            # 导入OSS排行榜管理器
            from src.oss.oss_leaderboard_manager import OSSLeaderboardManager
            
            # 从OSS加载排行榜数据
            leaderboard_manager = OSSLeaderboardManager()
            initial_data = leaderboard_manager.load_leaderboard_from_oss()
            
            if initial_data:
                print(f"✅ 从OSS加载了 {len(initial_data)} 条排行榜记录")
                
                for i, entry in enumerate(initial_data):
                    sage_result = SAGEResult(
                        submission_id=f"oss_{i:02d}_{entry['model_name'].replace(' ', '_').replace('-', '_')}",
                        organization=entry['organization'],
                        email=entry.get('contact_email', f"contact@{entry['organization'].lower().replace(' ', '')}.com"),
                        tokens=entry.get('tokens', 'N/A'),
                        accuracy=entry.get('accuracy', 0.0),
                        mg_pass_2=entry.get('mg_pass_2', 0.0),
                        mg_pass_4=entry.get('mg_pass_4', 0.0),
                        submitted_time=entry["submitted_time"],
                        status="EVALUATED"
                    )
                    # Add model_name as additional attribute for display
                    sage_result.model_name = entry['model_name']
                    sage_results.append(sage_result)
            else:
                print("⚠️ OSS中未找到排行榜数据")
                
        except Exception as e:
            print(f"❌ 从OSS加载排行榜失败: {e}")
        
        return sage_results
    
    def process_sage_results_for_leaderboard_oss() -> List[SAGEResult]:
        """Process all SAGE results from OSS"""
        return load_initial_sage_results_from_oss()
    
    # Set the function
    process_sage_results_for_leaderboard = process_sage_results_for_leaderboard_oss
    
except ImportError as e:
    print(f"Could not set up SAGE results processing: {e}")
    process_sage_results_for_leaderboard = None


def get_sage_leaderboard_df() -> pd.DataFrame:
    """Creates a dataframe from SAGE evaluation results"""
    if process_sage_results_for_leaderboard is None:
        return pd.DataFrame()
    
    # Get SAGE results
    sage_results = process_sage_results_for_leaderboard()
    all_data_json = [result.to_dict() for result in sage_results]
    
    if not all_data_json:
        return pd.DataFrame()
    
    df = pd.DataFrame.from_records(all_data_json)
    
    # Remove duplicates: for same Model+Organization, keep only the latest submission date
    if "Model" in df.columns and "Organization" in df.columns and "Submission Date" in df.columns:
        # Sort by Submission Date (descending) to get the latest first
        df = df.sort_values(by=["Submission Date"], ascending=False)
        # Drop duplicates, keeping first (latest) occurrence of each Model+Organization combination
        df = df.drop_duplicates(subset=["Model", "Organization"], keep="first")
    
    # Sort by accuracy (descending)
    if "Accuracy (%)" in df.columns:
        df = df.sort_values(by=["Accuracy (%)"], ascending=False)
    
    # Round numeric columns
    numeric_cols = ["Accuracy (%)", "mG-Pass@2 (%)", "mG-Pass@4 (%)"]
    for col in numeric_cols:
        if col in df.columns:
            df[col] = df[col].round(2)
    
    return df