physics_repeat_check/ir_db_establish.py

import os
import time
import shutil
import sqlite3

from config import sqlite_path, sqlite_copy_path, mongo_coll
from word_segment import Word_Segment

# 建立倒排索引
class Inverted_Index_Establish():
    def __init__(self, n_grams_flag=False):
        # 获取mongodb数据
        self.mongo_coll = mongo_coll
        # 获取数据库地址
        self.sqlite_path = sqlite_path
        self.sqlite_copy_path = sqlite_copy_path
        # 分词算法
        self.word_seg = Word_Segment(n_grams_flag=n_grams_flag)

    def __call__(self):
        # 从mongodb读取数据集
        origin_dataset = self.mongo_coll.find(no_cursor_timeout=True, batch_size=5)
        # 将数据构建倒排索引
        inverted_index = self.inverted_index_compute(origin_dataset)
        # 将倒排索引数据写入sqlite
        self.sqlite_write(inverted_index)
        # 备份sqlite数据库,防止出现损坏
        shutil.copy(self.sqlite_path, self.sqlite_copy_path)
    
    # 读取数据构建倒排索引
    def inverted_index_compute(self, origin_dataset):
        start = time.time()
        # 倒排索引词典
        inverted_index = dict()
        # 计算文档总长度,用于计算平均长度
        all_doc_length = 0
        for i,data in enumerate(origin_dataset):
            if "content_clear" not in data:
                continue
            seg_list, _ = self.word_seg(data["content_clear"])
            # 计算每篇文档长度和总文档长度
            doc_length = len(seg_list)
            all_doc_length += doc_length
            
            # 存储每个文档中每个词的出现次数
            tf_dict = dict()
            for word in seg_list:
                tf_dict[word] = tf_dict.get(word, 0) + 1
            # 将每个term对应的文档按固定格式存入倒排索引词典
            for term,tf in tf_dict.items():
                # 文档间通过"\n"分隔,文档内通过"\t"分隔(term_freq, doc_length)
                doc_info = "{}\t{}".format(tf, doc_length)
                if term not in inverted_index:
                    inverted_index[term] = [str(data["id"]), doc_info]
                else:
                    inverted_index[term][0] += "\n" + str(data["id"])
                    inverted_index[term][1] += "\n" + doc_info
        # 获取文档总数与文档总长度存入倒排索引词典
        inverted_index["doc_data_statistics"] = [i + 1, str(all_doc_length)]
        print(time.time()-start)
        print("倒排索引计算完毕")

        return inverted_index
    
    # 将倒排索引数据写入sqlite
    def sqlite_write(self, inverted_index):
        # 若已存在sqlite数据库,则删除数据库重建
        if os.path.exists(self.sqlite_path):
            os.remove(self.sqlite_path)
        # 建立sqlite数据库链接
        sqlite_conn = sqlite3.connect(self.sqlite_path)
        # 创建游标对象cursor
        cursor = sqlite_conn.cursor()
        # 创建数据表
        cursor.execute("DROP TABLE IF EXISTS physics")
        cursor.execute("CREATE TABLE physics (term TEXT PRIMARY KEY, doc_freq TEXT, docs TEXT)")
        
        for key, value in inverted_index.items():
            t = (key, value[0], value[1])
            cursor.execute("INSERT INTO physics VALUES (?, ?, ?)", t)
        # 提交事务
        sqlite_conn.commit()
        # 关闭数据库连接
        cursor.close()
        sqlite_conn.close()


if __name__ == "__main__":
    # 初始化倒排索引建立(n-grams为True)
    sql_etl = Inverted_Index_Establish(True)
    sql_etl()