spider2/lite/process.py

import json
from pathlib import Path


def judge_db_type(instance_id: str):
    if instance_id.startswith("bq") or instance_id.startswith("ga"):
        return "big_query"
    elif instance_id.startswith("sf"):
        return "snowflake"
    elif instance_id.startswith("local"):
        return "sqlite"

    return None


def process_schema(table: dict):
    format_table = dict()

    format_table["db_id"] = table["db_id"]
    format_table["db_size"] = table["db_stats"]["No. of columns"]
    # format_table["db_type"] = judge_db_type(table["db_type"])  # 可能需要根据数据来判断
    format_table["db_type"] = None,
    format_table["table_names_original"] = table["table_names_original"]
    format_table["column_names_original"] = table["column_names_original"]
    format_table["column_types"] = table["column_types"]
    format_table["column_descriptions"] = table["column_descriptions"]
    # format_table["sample_rows"] = table["sample_rows"]
    format_table["table_to_projDataset"] = table["table_to_projDataset"]

    return format_table


def process_dataset(dataset: dict):
    format_dataset = dict()


    format_dataset["instance_id"] = dataset["instance_id"]
    format_dataset["db_id"] = dataset["db_id"]
    format_dataset["question"] = dataset["question"]
    format_dataset["db_type"] = judge_db_type(dataset["instance_id"])
    format_dataset["db_size"] = dataset['No. of candidate columns']
    format_dataset["query"] = dataset["query"]

    format_dataset["external_path"] = dataset['external_knowledge']

    return format_dataset


if __name__ == '__main__':

    with open("origin_schema.json", 'r', encoding='utf-8') as f:
        schema = json.load(f)

    with open("origin_dataset.json", 'r', encoding='utf-8') as f:
        data = json.load(f)

    # 处理 schema
    tabel_data = []
    for row in schema:
        res = process_schema(row)
        tabel_data.append(res)

    # 处理 dataset
    dataset_data = []
    db_type_map = dict()
    for row in data:
        res = process_dataset(row)
        dataset_data.append(res)
        if res["db_type"] not in db_type_map:
            db_type_map[res["db_id"]] = res["db_type"]

    for row in tabel_data:
        db_type = db_type_map[row["db_id"]]
        row["db_type"] = db_type

    with open("./snow/dataset.json", 'w', encoding='utf-8') as f:
        json.dump(dataset_data, f, ensure_ascii=False, indent=4)

    with open("./snow/schema.json", 'w', encoding='utf-8') as f:
        json.dump(tabel_data, f, ensure_ascii=False, indent=4)