#!/usr/bin/python3
# -*- coding: utf-8 -*-
import argparse
import json
from pathlib import Path

import pandas as pd
from tqdm import tqdm

from project_settings import project_path


def get_args():
    parser = argparse.ArgumentParser()

    parser.add_argument(
        "--data_dir",
        default=(project_path / "original_data/dgk_lost_conv-master/results").as_posix(),
        type=str
    )
    parser.add_argument(
        "--output_file",
        default=(project_path / "data/subtitle.jsonl"),
        type=str
    )

    args = parser.parse_args()
    return args


def main():
    args = get_args()

    data_dir = Path(args.data_dir)
    with open(args.output_file, "w", encoding="utf-8") as fout:
        for filename in tqdm(data_dir.glob("*.conv")):
            with open(filename, "r", encoding="utf-8") as fin:
                conversation = list()
                for row in fin:
                    splits = str(row).strip().split(maxsplit=1)

                    if len(splits) == 1:
                        if len(conversation) == 0:
                            continue
                        row_ = {
                            "conversation": conversation,
                            "category": filename.stem,
                            "data_source": "subtitle",
                        }
                        row_ = json.dumps(row_, ensure_ascii=False)
                        fout.write("{}\n".format(row_))
                        conversation = list()
                    else:
                        message = splits[1]

                        conversation.append({
                            "role": "unknown",
                            "message": message,
                        })

    return


if __name__ == '__main__':
    main()