import json
import os

import pandas as pd
from dataclasses import fields
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_model_info
from src.display.utils import ModelType


def get_model_info_df(results_path: str) -> pd.DataFrame:
    """Creates a dataframe from all the individual experiment results"""
    raw_data = get_model_info(results_path)
    all_data_json = [v.to_dict() for v in raw_data]
    df = pd.DataFrame.from_records(all_data_json)
    return df


def get_merged_df(result_df: pd.DataFrame, model_info_df: pd.DataFrame) -> pd.DataFrame:
    """Merges the model info dataframe with the results dataframe"""
    result_df = result_df.rename(columns={"Model": "tmp_name"})
    merged_df = pd.merge(model_info_df, result_df, on="tmp_name", how="inner")
    assert len(merged_df) == len(
        result_df
    ), f"missing model info for: {set(result_df['tmp_name'].unique()) - set(model_info_df['tmp_name'].unique())}"
    merged_df = merged_df.drop(columns=["Model", "tmp_name"])
    merged_df = merged_df.rename(columns={"model_w_link": "Model"})
    return merged_df


# 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)
#     df = df.sort_values(by=[AutoEvalColumn.average.name], ascending=False)
#     df = df[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_leaderboard_df(results_path: str, requests_path: str, cols: list, benchmark_cols: list) -> pd.DataFrame:
    """
    Processes a STATIC results CSV file to generate a leaderboard DataFrame with formatted columns and sorted values.
    Args:
        results_path (str): The file path to the results CSV file.
    Returns:
        pd.DataFrame: A processed DataFrame with renamed columns, additional formatting, and sorted values.
    Notes:
        - The function reads a CSV file from the given `results_path`.
        - Internal column names are mapped to display names using `AutoEvalColumn`.
        - A new column for model type symbols is created by parsing the `model_type` column.
        - The `model_type` column is updated to prepend the model type symbol.
        - The DataFrame is sorted by the `Rank_scaled` column in ascending order.
    """

    df = pd.read_csv(results_path)
    # Create the mapping from internal column name to display name

    column_mapping = {field.name: getattr(AutoEvalColumn, field.name).name for field in fields(AutoEvalColumn)}
    # Assuming `df` is your DataFrame:
    df.rename(columns=column_mapping, inplace=True)
    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)
            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)
        elif ".md" not in entry:
            # this is a folder
            sub_entries = [
                e for e in os.listdir(f"{save_path}/{entry}") if os.path.isfile(e) and not e.startswith(".")
            ]
            for sub_entry in sub_entries:
                file_path = os.path.join(save_path, entry, sub_entry)
                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)

    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]