Multithreaded crawler batch download pcgame image url saved as XML implementation code

  • 2020-04-02 09:46:54
  • OfStack


#coding=gbk
from xml.dom import minidom,Node
import urllib2,re,os
def readsrc(src):
    try:
        url = urllib2.urlopen(src)
        content = url.read()#.decode('utf-8')
        return content
    except:
        print 'error'
        return None
def pictype(content):
    '''
     Grab the site navigation bar to get the site's image type 
     Returns the list, each list element is a dictionary, addr Links that represent the image type to, name The name that represents the image type 
     The error will return None
    '''
    p = re.compile(r'<ul>(.*)</ul>',re.S)
    r=p.search(content)
    if r:
        content=r.group()
    else:
        print None
    p = re.compile(r'<lis*.*?>s*<a href *= *"(?P<addr>.*?)">(?P<name>.*?)s*</a>s*</li>')
    l = [i.groupdict() for i in p.finditer(content)]
    l=l[1:]
    if len(l):return l
    else:return None
def pageinfo(src):
    '''
     Gets the details of a page 
     Returns the list of dictionaries for 
    name: Name of picture 
    cutaddr : a zoomed out view 
    picaddr : the address of the actual picture 
    '''
    d=os.path.split(src)[0]
    try:
        url = urllib2.urlopen(src)
        content = url.read()#.decode('utf-8')
    except:
        print 'error'
        return None
    #find all the pictures info in a page
    p = re.compile(r'<ul.*?>(.*?)</ul>',re.S)
    r = p.findall(content)
    if not r: return None
    r = r[1]
    p = re.compile(r'<li><a href="(?P<picaddr>.*?)".*?><img.*?alt="(?P<name>.*?)" *src="(?P<cutaddr>.*?)" */></a>.*?</li>')
    l = [ i.groupdict() for i in p.finditer(r)]
    for i in l:
        i['picaddr']=d+'/'+i['picaddr']
    if len(l): return l
    else: return None
def nextpageaddr(src):
    '''
     From the page html Gets the name of the next page address in the source code, and the last page returns None
    '''
    content=readsrc(src)
    p = re.compile(r'<a class="next" href="(.*?)">.*?</a>')
    r = p.search(content)
    if r:
        return os.path.dirname(src)+"/"+r.group(1)
    else:
        return None
def picinfoaddr(src):
    '''
     Parameter album atlas html code 
     Returns the relative addresses of all images 
    '''
    content=readsrc(src)
    p = re.compile(r'<div class="picinfo">.*?<a href="(?P<addr>.*?)".*?>.*?</div>',re.S)
    r = p.search(content)
    if r:
        return os.path.dirname(src)+"/"+r.group(1)
    else:
        return None
def parseinfo(content):
    '''
     Read all pictures html Code to get the details of an album 
    kw Key words: 
    title : the title 
    type Type: 
    pic : address list of each picture, add at the end _220x165 . _medium,_small  You can get pictures of different sizes 
    '''
    info={}
    temp=str()
    #title
    temp=''
    r=re.search('<h1>(.*?)</h1>',content)#get the pic title
    if r:
        temp = r.group(1)
    info['title']=temp
    #keyword
    temp=''
    r=re.search('<meta name="keywords" content="(.*?)" />',content)
    if r:
        temp = r.group(1)
    info['kw']=temp
    #type
    r=re.findall('<i><a.*?>(.*?)</a></i>.*?>',content)
    if r:
        info['type']=':'.join(r)
    else:
        info['type']=''
    r=re.search('<ul class=".*?">(.*?)</ul>',content,re.S)
    if not r:return None
    content=r.group(1)#filter content
#    print content
    r=re.findall('<a href=".*?<img.*?src="(.*?)".*?</a>',content)
    for index,i in enumerate(r):
        r[index]=i[0:i.rfind('_')]
#        print r[index]
    info['pic']=r
    return info
import threading
class mthread(threading.Thread):
    def __init__(self,tp,addr,lock):
        threading.Thread.__init__(self)
#        self.doc = minidom.Document()
        self.doc=minidom.Document()
        self.tp=tp
        self.lock=lock
        self.addr=addr
        self.thread_stop=False
        self.picdoc=None
    def run(self):
        self.picdoc = self.doc.createElement('urlclass')
#        print self.tp
        self.picdoc.setAttribute('type',self.tp)
#        self.doc.appendChild(self.picdoc)
        m=pageinfo(self.addr)
        while self.addr:
            for i in m:
#                print i['picaddr']
                picaddr=picinfoaddr(i['picaddr'])
#                print picaddr
                info=parseinfo(readsrc(picaddr))
                name=info['title']

                picture=doc.createElement('picture')
                title = doc.createElement('title')
                title.appendChild(doc.createTextNode(info['title']))
                picture.appendChild(title)
                keyword = doc.createElement('keywords')
                keyword.appendChild(doc.createTextNode(info['kw']))
                picture.appendChild(keyword)
                tp = doc.createElement('pictype')
                tp.appendChild(doc.createTextNode(info['type']))
                picture.appendChild(tp)
                cuturl = doc.createElement('piccut')
                cuturl.appendChild(doc.createTextNode(i['cutaddr']))
                picture.appendChild(cuturl)
                urls = doc.createElement('urls')
                self.lock.acquire()
                print 'downloading ',name
                self.lock.release()
                for picurl in info['pic']:
                    singleurl=doc.createElement('url')
                    singleurl.appendChild(doc.createTextNode(picurl+'.jpg'))
                    urls.appendChild(singleurl)
                picture.appendChild(urls)
                self.picdoc.appendChild(picture)
            m=pageinfo(self.addr)
            self.addr=nextpageaddr(self.addr)
#        f = open('c:\'+self.tp+'.xml','w')
#        f.write(doc.toprettyxml(indent = ''))
#        f.close()
    def stop(self):
        self.thread_stop=True

path='C:\pict\'# Download path 
#import sys
sys.exit(12)
content=readsrc('http://photos.pcgames.com.cn/cate/3/1.html')
r=pictype(content)
lt=[]
doc = minidom.Document()
root=doc.createElement('url_resource')
root.setAttribute('type','url')
root.setAttribute('urltype','image')
root.setAttribute('imgfmt','jpg')
doc.appendChild(root)
lock=threading.RLock()
for iaddr in r:
    print 'downloading type: ',iaddr['name']
    addr=iaddr['addr']
    th=mthread(iaddr['name'],addr,lock)
    lt.append(th)
    th.start()
for t in lt:
    t.join()
    root.appendChild(t.picdoc)
print 'write'
f = open('c:\'+'urls'+'.xml','w')
f.write(doc.toprettyxml(indent = ''))
f.close()
print doc.toprettyxml()
print 'end'


Related articles: