import json
import os

import pandas as pd
import numpy as np

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


# 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, cols: list, benchmark_cols: list) -> pd.DataFrame:
    """Creates a dataframe from all the individual experiment results"""

    # iterate thorugh all files in the results path and read them into json
    all_data_json = []
    res_path = os.path.join(results_path, "demo-leaderboard", "syntherela-demo")
    for entry in os.listdir(res_path):
        if entry.endswith(".json"):
            file_path = os.path.join(res_path, entry)
            with open(file_path) as fp:
                data = json.load(fp)
                all_data_json.append(data)

    multi_table_metrics = [
        "AggregationDetection-LogisticRegression",
        "AggregationDetection-XGBClassifier",
        "CardinalityShapeSimilarity",
    ]

    # create empty dataframe with the columns multi_table_metrics
    multitable_df = pd.DataFrame(columns=["Dataset", "Model"] + multi_table_metrics)

    # iterate through all json files and add the data to the dataframe
    for data in all_data_json:
        model = data["model"]
        dataset = data["dataset"]
        row = {"Dataset": dataset, "Model": model}
        for metric in multi_table_metrics:
            if metric in data["multi_table_metrics"]:
                metric_values = []
                for table in data["multi_table_metrics"][metric].keys():
                    if "accuracy" in data["multi_table_metrics"][metric][table]:
                        metric_values.append(data["multi_table_metrics"][metric][table]["accuracy"])
                    if "statistic" in data["multi_table_metrics"][metric][table]:
                        metric_values.append(data["multi_table_metrics"][metric][table]["statistic"])

                row[metric] = np.mean(metric_values)
        multitable_df = pd.concat([multitable_df, pd.DataFrame([row])], ignore_index=True)
    return multitable_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]