码迷,mamicode.com
首页 > 数据库 > 详细

python+Sqlite+Dataframe打造金融股票数据结构

时间:2020-01-27 19:26:57      阅读:112      评论:0      收藏:0      [点我收藏+]

标签:font   commit   5*   com   import   open   print   rom   tor   

5. 本地数据库

很简单的用本地Sqlite查找股票数据。

DataSource类,返回的是Dataframe物件。这个Dataframe物件,在之后的业务,如计算股票指标,还需要特别处理。

import os
import sqlite3 as sqlite3
import numpy as np
import pandas as pd


# 数据源
class DataSource:
    def __init__(self):
        self.db = None              # 数据库
        self.cursor = None          # 指针
        self.stocks = {}            # 股票池
        self.indexs = {}            # 指数池
        self.name = unit_test.db  # 数据源名称

    def connect(self):
        self.db = sqlite3.connect(os.path.abspath(self.name))
        self.cursor = self.db.cursor()

    def get_stocks(self, ucodes):
        # 股票池
        try:
            self.stocks = {}
            self.connect()
            self.db.row_factory = lambda cursor, row: row[0]
            for ucode in ucodes:
                sql = """SELECT t.code, t.lot, t.nmll, t.stime, t.high, t.low, t.open, t.close, t.volume
                            FROM (SELECT n.code, n.lot, n.nmll, c.stime, c.high, c.low, c.open, c.close, c.volume 
                                FROM s_{} AS c INNER JOIN name AS n 
                                    ON c.code=n.code ORDER BY c.stime DESC LIMIT 365*20) AS t 
                                /*INNER JOIN financial AS f 
                            ON t.code=f.code AND substr(t.stime,1,4)=f.year*/
                        ORDER BY t.stime""".format(ucode)
                self.cursor.execute(sql)
                columns = [code, lot, nmll, sdate, high, low, open, last, vol]
                self.stocks[ucode] = pd.DataFrame(self.cursor.fetchall(), columns=columns)
            self.db.commit()
            self.cursor.close()
            self.db.close()
            return self.stocks
        except sqlite3.Error as e:
            print(e)

    def get_indexs(self, indexs):
        try:
            # 指数池
            self.indexs = {}
            self.connect()
            self.db.row_factory = lambda cursor, row: row[0]
            for index in indexs:
                sql = """SELECT t.code, t.lot, t.nmll, t.stime, t.high, t.low, t.open, t.close, t.volume
                            FROM (SELECT n.code, n.lot, n.nmll, c.stime, c.high, c.low, c.open, c.close, c.volume 
                                FROM s_{} AS c INNER JOIN name AS n 
                                    ON c.code=n.code ORDER BY c.stime DESC LIMIT 365*20) AS t 
                                /*INNER JOIN financial AS f 
                            ON t.code=f.code AND substr(t.stime,1,4)=f.year*/
                        ORDER BY t.stime""".format(index.upper())
                self.cursor.execute(sql)
                columns = [code, lot, nmll, sdate, high, low, open, last, vol]
                self.indexs[index] = pd.DataFrame(self.cursor.fetchall(), columns=columns)
            self.db.commit()
            self.cursor.close()
            self.db.close()
            return self.indexs
        except sqlite3.Error as e:
            print(e)


data_source = DataSource()
df1 = data_source.get_stocks([00700])
df2 = data_source.get_indexs([hsi])

 

python+Sqlite+Dataframe打造金融股票数据结构

标签:font   commit   5*   com   import   open   print   rom   tor   

原文地址:https://www.cnblogs.com/chenkuang/p/12236519.html

(0)
(0)
   
举报
评论 一句话评论(0
登录后才能评论!
© 2014 mamicode.com 版权所有  联系我们:gaon5@hotmail.com
迷上了代码!