利用Python进行数据分析_Pandas_数据加载、存储与文件格式

申明：本系列文章是自己在学习《利用Python进行数据分析》这本书的过程中，为了方便后期自己巩固知识而整理。

1 pandas读取文件的解析函数

read_csv 读取带分隔符的数据，默认分隔符逗号

read_table 读取带分隔符的数据，默认分隔符 “ ”

read_fwf 读取定宽、列格式数据（无分隔符）

read_clipboard 读取剪贴板中的数据（将网页转换为表格）

1.1 读取excel数据

import pandas as pd
import numpy as np
file = 'D:example.xls'
pd = pd.read_excel(file)
pd

运行结果：

1.1.1 不显示表头

pd = pd.read_excel(file,header=None)

运行结果：

1.1.2 设置表头

pd = pd.read_excel(file,names=['Year','Name','Math','Chinese','EngLish','Avg'])

运行结果：

1.1.3 指定索引

pd = pd.read_excel(file,index_col= '姓名')

运行结果：

2 读取CSV数据

import pandas as pd
import numpy as np
pd = pd.read_csv("d:\test.csv",engine='python')
pd

运行结果：

import pandas as pd
import numpy as np
pd = pd.read_table("d:\test.csv",engine='python')
pd

运行结果：

import pandas as pd
import numpy as np
pd = pd.read_fwf("d:\test.csv",engine='python')
pd

运行结果：

3 将数据写出到文本格式

将数据写出到csv格式，默认分隔符逗号

import pandas as pd
import numpy as np
pd = pd.read_fwf("d:\test.csv",engine='python')
pd.to_csv("d:\test1.csv",encoding='gbk')

运行结果：

4 手工处理分隔符格式

单字符分隔符文件，直接用csv模块

import pandas as pd
import numpy as np
import csv
file = 'D:\test.csv'
pd = pd.read_csv(file,engine='python')
pd.to_csv("d:\test1.csv",encoding='gbk',sep='/')
f = open("d:\test1.csv")
reader = csv.reader(f)
for line in reader:
print(line)

运行结果：

4.1 缺失值填充

import pandas as pd
import numpy as np
import csv
file = 'D:\test.csv'
pd = pd.read_csv(file,engine='python')
pd.to_csv("d:\test1.csv",encoding='gbk',sep='/',na_rep='NULL')
f = open("d:\test1.csv")
reader = csv.reader(f)
for line in reader:
    print(line)

运行结果：

4.2 JSON

4.2.1 json.loads 可将JSON字符串转换成Python形式

import pandas as pd
import numpy as np
import json
obj = """{
  "sucess" : "1",
  "header" : {
    "version" : 0,
    "compress" : false,
    "times" : 0
  },
  "data" : {
    "name" : "BankForQuotaTerrace",
    "attributes" : {
      "queryfound" : "1",
      "numfound" : "1",
      "reffound" : "1"
    },
    "columnmeta" : {
      "a0" : "DATE",
      "a1" : "DOUBLE",
      "a2" : "DOUBLE",
      "a3" : "DOUBLE",
      "a4" : "DOUBLE",
      "a5" : "DOUBLE",
      "a6" : "DATE",
      "a7" : "DOUBLE",
      "a8" : "DOUBLE",
      "a9" : "DOUBLE",
      "b0" : "DOUBLE",
      "b1" : "DOUBLE",
      "b2" : "DOUBLE",
      "b3" : "DOUBLE",
      "b4" : "DOUBLE",
      "b5" : "DOUBLE"
    },
    "rows" : [ [ "2017-10-28", 109.8408691012081, 109.85566362201733, 0.014794520809225841, 1.0, null, "", 5.636678251676443, 5.580869556115291, 37.846934105222246, null, null, null, null, null, 0.061309012867495856 ] ]
  }
}
"""
result = json.loads(obj)
result

运行结果：

4.2.2 json.dumps可将Python字符串转换成JSON形式

result = json.loads(obj)
asjson=json.dumps(result)
asjson

运行结果：

4.2.3 JSON数据转换成DataFrame

import pandas as pd
import numpy as np
from pandas import DataFrame
import json
obj = """{
  "sucess" : "1",
  "header" : {
    "version" : 0,
    "compress" : false,
    "times" : 0
  },
  "data" : {
    "name" : "BankForQuotaTerrace",
    "attributes" : {
      "queryfound" : "1",
      "numfound" : "1",
      "reffound" : "1"
    },
    "columnmeta" : {
      "a0" : "DATE",
      "a1" : "DOUBLE",
      "a2" : "DOUBLE",
      "a3" : "DOUBLE",
      "a4" : "DOUBLE",
      "a5" : "DOUBLE",
      "a6" : "DATE",
      "a7" : "DOUBLE",
      "a8" : "DOUBLE",
      "a9" : "DOUBLE",
      "b0" : "DOUBLE",
      "b1" : "DOUBLE",
      "b2" : "DOUBLE",
      "b3" : "DOUBLE",
      "b4" : "DOUBLE",
      "b5" : "DOUBLE"
    },
    "rows" : [ [ "2017-10-28", 109.8408691012081, 109.85566362201733, 0.014794520809225841, 1.0, null, "", 5.636678251676443, 5.580869556115291, 37.846934105222246, null, null, null, null, null, 0.061309012867495856 ] ]
  }
}
"""
result = json.loads(obj)
result
jsondf = DataFrame(result['data'],columns = ['name','attributes','columnmeta'],index={1,2,3})
jsondf

运行结果：

备注：其中attributes和columnmeta，存在嵌套，这个问题后面再补充。

4.3 XML和HTML

爬取同花顺网页中的列表数据，并转换成DataFrame

在爬取的时候，我这里没有考虑爬分页的数据，有兴趣的可以自己尝试，我这里主要是想尝试爬取数据后转成DataFrame

代码如下：

import pandas as pd
import numpy as np
from pandas.core.frame import DataFrame
from lxml.html import parse
import requests
from bs4 import BeautifulSoup
import time

url = 'http://data.10jqka.com.cn/market/longhu/'
headers = {"User-Agent":"Mozilla/5.0 (Windows NT 6.1; WOW64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/63.0.3239.132 Safari/537.36"}
response = requests.get(url = url,headers = headers)
html = response.content
soup = BeautifulSoup(html, 'lxml')
s = soup.find_all('div','yyb')

# 获取dataframe所需的columns
def getcol():
    col = []
    for i in s: 
        lzs = i.find_all('thead')
        for k in lzs: 
            lbs = k.find_all('th') 
            for j in lbs:
                col.append(j.text.strip('
'))
            return col
        
# 获取dataframe所需的values
def getvalues():
    val = []
    for j in s:
        v = j.find_all('tbody')
        for k in v:
            vv = k.find_all('tr')
            list = []
            for l in vv:
                tdlist = []
                vvv = l.find_all('td')
                for m in vvv:
                    tdlist.append(m.text)
                list.append(tdlist)
            return(list)

if __name__ == "__main__":
    cols = getcol()
    values = getvalues()
    data=DataFrame(values,columns=cols)
    print(data)

运行结果：

4.4 二进制数据格式

pandas对象的save方法保存，load方法读回到Python

4.5 HDF5格式

HDF是层次型数据格式，HDF5文件含一个文件系统式的节点结构，支持多个数据集、元数据，可以高效的分块读写。Python中的HDF5库有2个接口：PyTables和h5py。

海量数据应该考虑用这个，现在我没用着，先不研究了。

4.6 使用HTML和Web API

import requests
import pandas as pd
from pandas import DataFrame
import json
url = 'http://t.weather.sojson.com/api/weather/city/101030100'
resp = requests.get(url)
data = json.loads(resp.text)#这里的data是一个dict
jsondf = DataFrame(data['cityInfo'],columns =['city','cityId','parent','updateTime'],index=[1])#实例化
jsondf

运行结果：

4.7 使用数据库

4.7.1 sqlite3

import sqlite3
import pandas.io.sql as sql

con = sqlite3.connect()

sql.read_frame('select * from test',con)#con 是一个连接对象

4.7.1 MongoDB

没装。先搁置。