import json
import os

import pandas as pd

from src.display.formatting import has_no_nan_values, make_clickable_model
from src.display.utils import AutoEvalColumn, EvalQueueColumn
from src.leaderboard.read_evals import get_raw_eval_results
from src.about import Tasks


def get_leaderboard_df(results_path: str, requests_path: str, cols: list, benchmark_cols: list) -> pd.DataFrame:
    """Creates a dataframe from all the individual experiment results"""
    raw_data = get_raw_eval_results(results_path, requests_path)
    all_data_json = [v.to_dict() for v in raw_data]

    df = pd.DataFrame.from_records(all_data_json)
    
    # Handle empty DataFrame case
    if df.empty:
        # Create empty DataFrame with correct columns
        df = pd.DataFrame(columns=cols)
        return df
    
    # Sort by the first task (EMEA NER) since we don't have an average for NER tasks
    # If no results exist yet, just sort by model name
    first_task = list(Tasks)[0]  # emea_ner
    task_col_name = first_task.value.col_name  # Use the col_name directly
    if task_col_name in df.columns:
        df = df.sort_values(by=[task_col_name], ascending=False)
    else:
        # Fallback to sorting by model name if no task results yet
        df = df.sort_values(by=["Model"], ascending=True)
    
    # Only select columns that exist in the DataFrame
    available_cols = [col for col in cols if col in df.columns]
    df = df[available_cols].round(decimals=2)

    # filter out if any of the benchmarks have not been produced
    df = df[has_no_nan_values(df, benchmark_cols)]
    return df


def get_evaluation_queue_df(save_path: str, cols: list) -> list[pd.DataFrame]:
    """Creates the different dataframes for the evaluation queues requestes"""
    entries = [entry for entry in os.listdir(save_path) if not entry.startswith(".")]
    all_evals = []

    for entry in entries:
        if ".json" in entry:
            file_path = os.path.join(save_path, entry)
            try:
                with open(file_path) as fp:
                    data = json.load(fp)

                data[EvalQueueColumn.model.name] = make_clickable_model(data["model"])
                data[EvalQueueColumn.revision.name] = data.get("revision", "main")

                all_evals.append(data)
            except (json.JSONDecodeError, KeyError, IOError) as e:
                print(f"Error processing {file_path}: {e}")
                continue
        elif ".md" not in entry:
            # this is a folder
            sub_entries = [e for e in os.listdir(os.path.join(save_path, entry)) 
                          if os.path.isfile(os.path.join(save_path, entry, e)) and not e.startswith(".")]
            for sub_entry in sub_entries:
                file_path = os.path.join(save_path, entry, sub_entry)
                try:
                    with open(file_path) as fp:
                        data = json.load(fp)

                    data[EvalQueueColumn.model.name] = make_clickable_model(data["model"])
                    data[EvalQueueColumn.revision.name] = data.get("revision", "main")
                    all_evals.append(data)
                except (json.JSONDecodeError, KeyError, IOError) as e:
                    print(f"Error processing {file_path}: {e}")
                    continue

    pending_list = [e for e in all_evals if e["status"] in ["PENDING", "RERUN"]]
    running_list = [e for e in all_evals if e["status"] == "RUNNING"]
    finished_list = [e for e in all_evals if e["status"].startswith("FINISHED") or e["status"] == "PENDING_NEW_EVAL"]
    df_pending = pd.DataFrame.from_records(pending_list, columns=cols)
    df_running = pd.DataFrame.from_records(running_list, columns=cols)
    df_finished = pd.DataFrame.from_records(finished_list, columns=cols)
    return df_finished[cols], df_running[cols], df_pending[cols]