一、Python爬虫
---------------------------------------------------
1.测试
# -*- encoding=utf-8 -*-
import urllib.request
# 打开url上的资源
resp = urllib.request.urlopen("http://focus.tianya.cn/")
# 读取内容，返回byte数组
mybytes = resp.read()
# 解码bytes成为string
mystr = mybytes.decode("utf-8")
# 关闭资源
resp.close()
# 输出
print(mystr)
# 导入正则表达式模块
import re
ptn = u'<a\s*href="([\u0000-\uffff&&^"]*?)"' #非贪婪模式
res = re.finditer(ptn,mystr)
for r in res:
addr = r.group(1);
print(addr)
2.递归爬并保存
# -*- encoding=utf-8 -*-
import urllib.request
import os;
import re;
def fileExists(url,localpath):
path = url ;
path = path.replace(":", "_");
path = path.replace("/", "$");
path = path.replace("?", "$");
path = localpath + "/" + path;
return os.path.exists(path) ;
#下载网页方法
def download(url):
#处理处理问题
path = url ;
path = path.replace(":","_");
path = path.replace("/","$");
path = path.replace("?","$");
path = "d:/py/data/" + path;
#判断当前的网页是否已经下载
resp = urllib.request.urlopen(url)
pageBytes = resp.read()
resp.close
if not os.path.exists(path):
#保存文件到磁盘
f = open(path,"wb");
f.write(pageBytes) ;
f.close();
try:
#解析网页的内容
pageStr = pageBytes.decode("utf-8");
#解析href地址
pattern = u'<a[\u0000-\uffff&&^[href]]*href="([\u0000-\uffff&&^"]*?)"'
res = re.finditer(pattern, pageStr)
for r in res:
addr = r.group(1);
print(addr)
if addr.startswith("//"):
addr = addr.replace("//","http://");
#判断网页中是否包含自己的地址
if (addr.startswith("http://") and not fileExists(addr,"d:/py/data")):
download(addr) ;
except Exception as e:
#print(url + " : 不是文本") ;
#print(Exception)
print(e)
# print(pageBytes.decode("gbk", errors='ignore'));
return ;
download("http://www.jd.com");
二、Python协同hbase实现数据的写入
--------------------------------------------------------------
1.启动hbase集群
a.启动zk
b.启动hadoop集群
c.启动hbase集群
如果时钟不同步。
$>su root
$>xcall.sh "ntpdate asia.pool.ntp.org"
2.s100上启动hbase的thriftserver服务器，满足和第三方应用通信
$> hbase-daemon.sh start thrift2
3.查看webui
http://s100:9095/ //webui端口
$> netstat -anop | grep 9090 //9090 rpc端口
4.下载windows下thrift的编译器,不需要安装，仅仅是个工具。
thrift-0.10.0.exe
5.下载并安装thrift的python模块.
5.1)下载文件
thrift-0.10.0.tar.gz
5.2)tar开文件
5.3)进入目录
cmd>cd thrift-0.10.0\lib\py
cmd>setup.py install
...
Using c:\users\administrator\appdata\local\programs\python\python37\lib\site-packages\six-1.11.0-py3.7.egg
Finished processing dependencies for thrift==0.10.0
6.测试在py文件中是否能够导入
from thrift import Thrift
from thrift.transport import TSocket
from thrift.transport import TTransport
from thrift.protocol import TBinaryProtocol
7.找到hbase.thrift文件进行编译，产生python文件，使用以下命令进行编译
[hbase.thrift文件位于hbase安装包下,找不到就去网上下载]
cmd> thrift-0.10.0.exe -o ./out -gen py hbase.thrift
8.将生成的文件夹拷贝到idea/python模块下
a.在模块下新建一个pythonpackage, 叫mythrift
b.将生成的py文件夹下的hbase文件夹拷贝到mythrift下
9.使用python操作hbase的表
# -*- encoding=utf-8 -*-
# 导入原生模块
import os
# 导入thrift的python模块
from thrift import Thrift
from thrift.transport import TSocket
from thrift.transport import TTransport
from thrift.protocol import TBinaryProtocol
# 导入自已编译生成的hbase python模块
from mythrift.hbase import THBaseService
from mythrift.hbase.ttypes import *
from mythrift.hbase.ttypes import TResult
# 创建Socket连接，到s100:9090
transport = TSocket.TSocket('s100', 9090)
transport = TTransport.TBufferedTransport(transport)
protocol = TBinaryProtocol.TBinaryProtocol(transport)
client = THBaseService.Client(protocol)
# 打开传输端口
transport.open()
# # put操作
# table = b'ns1:t1'
# row = b'row2'
# v1 = TColumnValue(b'f1', b'id', b'101')
# v2 = TColumnValue(b'f1', b'name', b'tomas')
# v3 = TColumnValue(b'f1', b'age', b'12')
# vals = [v1, v2, v3]
# put = TPut(row, vals)
# client.put(table, put)
# print("okkkk!!")
# transport.close()
# # get操作
# table = b'ns1:t1'
# rowkey=b"row2"
# col_id = TColumn(b"f1",b"id")
# col_name = TColumn(b"f1",b"name")
# col_age = TColumn(b"f1",b"age")
#
# cols = [col_id,col_name,col_age]
# get = TGet(rowkey,cols)
# res = client.get(table,get)
# print(res.columnValues)
# print(bytes.decode(res.columnValues[0].qualifier))
# print(bytes.decode(res.columnValues[0].family))
# print(res.columnValues[0].timestamp)
# print(bytes.decode(res.columnValues[0].value))
# # delete操作
# table = b'ns1:t1'
# rowkey = b"row2"
# col_id = TColumn(b"f1", b"id")
# col_name = TColumn(b"f1", b"name")
# col_age = TColumn(b"f1", b"age")
# cols = [col_id, col_name, col_age]
#
# #构造删除对象
# delete = TDelete(rowkey,cols)
# res = client.deleteSingle(table, delete)
# transport.close()
# print("ok")
# scan 扫描操作
table = b'call:calllogs'
startRow = b'34,13520401111,20180114152647,0,13269364444,406'
stopRow = b'90,15032295555,20180922165903,0,15778421111,298'
dur = TColumn(b"f1", b"callDuration")
time = TColumn(b"f1", b"callTime")
caller = TColumn(b"f1", b"caller")
callee = TColumn(b"f1", b"callee")
cols = [dur, time,caller,callee]
scan = TScan(startRow=startRow,stopRow=stopRow,columns=cols)
r = client.getScannerResults(table,scan,100);
for x in r:
print("============")
print(bytes.decode(x.columnValues[0].qualifier))
print(bytes.decode(x.columnValues[0].family))
print(x.columnValues[0].timestamp)
print(bytes.decode(x.columnValues[0].value))
# scan 全表扫描操作
table = b'call:calllogs'
# startRow = b'34,13520401111,20180114152647,0,13269364444,406'
# stopRow = b'90,15032295555,20180922165903,0,15778421111,298'
dur = TColumn(b"f1", b"callDuration")
time = TColumn(b"f1", b"callTime")
caller = TColumn(b"f1", b"caller")
callee = TColumn(b"f1", b"callee")
cols = [dur, time,caller,callee]
scan = TScan(columns=cols)
r = client.getScannerResults(table,scan,100);
print(len(r))
for x in r:
print("============")
print(bytes.decode(x.columnValues[0].qualifier))
print(bytes.decode(x.columnValues[0].family))
print(x.columnValues[0].timestamp)
print(bytes.decode(x.columnValues[0].value))
三、SparkShell使用Python进行WorldCount
-----------------------------------------------------------------
1.本地模式[可以使用scala和python编写]
a.移除spark/conf/core-site.xml | hdfs-site.xml | hive-site.xml文件
[这样spark就不会去集成hive了]
b.进入pyspark shell
$> cd /soft/spark/bin
$> ./pyspark --master local[*]
>>> arr = [1,2,3,4]
>>> rdd = sc.parellize(arr);
>>> rdd.map(lambda e : (e,1)) #python的lamba表达式
2.WorldCountDemo
>>> arr = ["tom","tom1","tom1","tom3"]
>>> rdd1 = sc.parallelize(arr)
>>> rdd1.collect()
['tom', 'tom1', 'tom1', 'tom3']
>>> rdd1.map(lambda e : (e,1))
PythonRDD[4] at RDD at PythonRDD.scala:48
>>> rdd2 = rdd1.map(lambda e : (e,1))
>>> rdd2.collect()
[('tom', 1), ('tom1', 1), ('tom1', 1), ('tom3', 1)]
>>> rdd3 = rdd2.reduceByKey()
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
TypeError: reduceByKey() takes at least 2 arguments (1 given)
>>> rdd3 = rdd2.reduceByKey(lambda a,b : a + b)
>>> rdd3.collect()
[('tom1', 2), ('tom3', 1), ('tom', 1)]
四、爬虫程序更改 -- 将爬到的网页储存到hbase中
---------------------------------------------------------
1.使用base64进行编解码
import base64;
url = b"http://tianya.cn";
b = base64.encodebytes(url);
print(b)
bb = base64.decodebytes(b)
print(bb)
2.创建hbase表：pages
hbase> create 'ns1:pages','f1'
3.编写pageDao.py,专门处理hbase表的crud
# -*- encoding=utf-8 -*-
# 导入原生模块
import os
import base64
# 导入thrift的python模块
from thrift import Thrift
from thrift.transport import TSocket
from thrift.transport import TTransport
from thrift.protocol import TBinaryProtocol
# 导入自已编译生成的hbase python模块
from mythrift.hbase import THBaseService
from mythrift.hbase.ttypes import *
from mythrift.hbase.ttypes import TResult
# 创建Socket连接，到s100:9090
transport = TSocket.TSocket('s100', 9090)
transport = TTransport.TBufferedTransport(transport)
protocol = TBinaryProtocol.TBinaryProtocol(transport)
client = THBaseService.Client(protocol)
#定义函数，保存网页
def savePage(url,page):
#
transport.open()
#对url进行base64编码，形成bytes,作为rowkey
urlBase64Bytes = base64.encodebytes(url.encode("utf-8"))
# put操作
table = b'ns1:pages'
rowkey = urlBase64Bytes
v1 = TColumnValue(b'f1', b'page', page)
vals = [v1]
put = TPut(rowkey, vals)
client.put(table, put)
transport.close()
#判断网页是否存在
def exists(url):
transport.open()
# 对url进行base64编码，形成bytes,作为rowkey
urlBase64Bytes = base64.encodebytes(url.encode("utf-8"))
print(urlBase64Bytes)
table = b'ns1:pages'
rowkey = urlBase64Bytes
col_page = TColumn(b"f1",b"page")
cols = [col_page]
get = TGet(rowkey,cols)
res = client.get(table, get)
transport.close()
return res.row is not None
4.编写爬虫程序
# -*- encoding=utf-8 -*-
import urllib.request
import os
import re
import pageDao
#下载网页方法
def download(url):
#判断当前的网页是否已经下载
resp = urllib.request.urlopen(url)
pageBytes = resp.read()
resp.close
if not pageDao.exists(url):
pageDao.savePage(url, pageBytes);
try:
#解析网页的内容
pageStr = pageBytes.decode("utf-8");
#解析href地址
pattern = u'<a[\u0000-\uffff&&^[href]]*href="([\u0000-\uffff&&^"]*?)"'
res = re.finditer(pattern, pageStr)
for r in res:
addr = r.group(1);
print(addr)
if addr.startswith("//"):
addr = addr.replace("//","http://");
#判断网页中是否包含自己的地址
if addr.startswith("http://") and url != addr and (not pageDao.exists(addr)):
download(addr) ;
except Exception as e:
print(e)
print(pageBytes.decode("gbk", errors='ignore'));
return ;
download("http://jd.com");
五、使用python实现spark的数据分析,生成分析图表
-----------------------------------------------------------------
1.Win上安装pip安装python的模块
1.numpy
cmd> pip install -i https://pypi.tuna.tsinghua.edu.cn/simple numpy
2.scipy
cmd> pip install -i https://pypi.tuna.tsinghua.edu.cn/simple scipy
3.matplotpy
cmd> pip install -i https://pypi.tuna.tsinghua.edu.cn/simple matplotlib
cmd> python -m pip install -U pip setuptools
cmd> python -m pip install matplotlib
2.ubuntu下安装numpy，scipy，pandas，matplotlib模块
a.numpy
安装：
如果你的ubuntu中没有安装python，请首先安装python
在终端输入以下命令：
$> sudo apt-get update
$> sudo apt-get install python-numpy
$> sudo apt-get install python3-numpy
如果是python3，则将上面的python-numpy换成python3-numpy即可
b.scipy
$> sudo apt-get update
$> sudo apt-get install python-scipy
c.pandas
$> sudo apt-get update
$> sudo apt-get install python-pandas
d.matplotlib
$> sudo apt-get update
$> sudo apt-get install python-matplotlib
e.scikit-learn
$> sudo apt-get update
$> sudo apt-get install python-sklearn
3.进入PythonSparkShell
cmd> ./pyspark --master local[*]
$> ./pyspark --master local[*]
4.粘贴下面代码，创建数据框
from pyspark.sql import Row
import matplotlib.pyplot as plt
import numpy as np
import pylab as P
plt.rcdefaults()
dataDir ="file:///D://share//python//ml-data//ml-1m//users.dat"
dataDir ="file:///mnt/hgfs/share/python/ml-data/ml-1m/users.dat"
lines = sc.textFile(dataDir)
splitLines = lines.map(lambda l: l.split("::"))
usersRDD = splitLines.map(lambda p: Row(id=p[0],gender=p[1],age=int(p[2]), occupation=p[3], zipcode=p[4]))
usersDF = spark.createDataFrame(usersRDD)
usersDF.createOrReplaceTempView("users")
usersDF.show()
5.生成图表
#生成直方图
ageDF = spark.sql("SELECT age FROM users")
ageList = ageDF.rdd.map(lambda p: p.age).collect()
ageDF.describe().show()
plt.hist(ageList)
plt.title("Age distribution of the users\n")
plt.xlabel("Age")
plt.ylabel("Number of users")
plt.show(block=False)
#密度图
from scipy.stats import gaussian_kde
density = gaussian_kde(ageList)
xAxisValues = np.linspace(0,100,1000)
density.covariance_factor = lambda : .5
density._compute_covariance()
plt.title("Age density plot of the users\n")
plt.xlabel("Age")
plt.ylabel("Density")
plt.plot(xAxisValues, density(xAxisValues))
plt.show(block=False)
#生成嵌套子图
plt.subplot(121)
plt.hist(ageList)
plt.title("Age distribution of the users\n")
plt.xlabel("Age")
plt.ylabel("Number of users")
plt.subplot(122)
plt.title("Summary of distribution\n")
plt.xlabel("Age")
plt.boxplot(ageList, vert=False)
plt.show(block=False)
#柱状图
occ10 = spark.sql("SELECT occupation, count(occupation) as usercount FROM users GROUP BY occupation ORDER BY usercount DESC LIMIT 10")
occ10.show()
occTuple = occ10.rdd.map(lambda p:(p.occupation,p.usercount)).collect()
occList, countList = zip(*occTuple)
occList
y_pos = np.arange(len(occList))
plt.barh(y_pos, countList, align='center', alpha=0.4)
plt.yticks(y_pos, occList)
plt.xlabel('Number of users')
plt.title('Top 10 user types\n')
plt.gcf().subplots_adjust(left=0.15)
plt.show(block=False)
#堆栈条形图
occGender = spark.sql("SELECT occupation, gender FROM users")
occGender.show()
occCrossTab = occGender.stat.crosstab("occupation","gender")
occupationsCrossTuple = occCrossTab.rdd.map(lambda p:(p.occupation_gender,p.M, p.F)).collect()
occList, mList, fList = zip(*occupationsCrossTuple)
N = len(occList)
ind = np.arange(N)
width = 0.75
p1 = plt.bar(ind, mList, width, color='r')
p2 = plt.bar(ind, fList, width, color='y', bottom=mList)
plt.ylabel('Count')
plt.title('Gender distribution by occupation\n')
plt.xticks(ind + width/2., occList, rotation=90)
plt.legend((p1[0], p2[0]), ('Male', 'Female'))
plt.gcf().subplots_adjust(bottom=0.25)
plt.show(block=False)
#饼图
occupationsBottom10 = spark.sql("SELECT occupation,count(occupation) as usercount FROM users GROUP BY occupation ORDER BY usercount LIMIT 10")
occupationsBottom10Tuple = occupationsBottom10.rdd.map(lambda p:(p.occupation,p.usercount)).collect()
occupationsBottom10List, countBottom10List =zip(*occupationsBottom10Tuple)
explode = (0, 0.3, 0.2, 0.15,0.1,0,0,0,0,0.1)
plt.pie(countBottom10List, explode=explode,labels=occupationsBottom10List, autopct='%1.1f%%', shadow=True,startangle=90)
plt.title('Bottom 10 user types\n')
plt.show(block=False)