Subversion Repositories SmartDukaan

import urllib2

from BeautifulSoup import BeautifulSoup

import re

from sys import exit

class FlipkartScraper:

    def __init__(self):

        self.count_trials = 0

        self.redirectCount = 0

    def read(self, url):

        print url.replace('http://www.flipkart.com','163.53.77.21')

        print url

        request = urllib2.Request(url)

        request.add_header('User-Agent', 'Mozilla/5.0 (X11; Linux i686; rv:31.0) Gecko/20100101 Firefox/31.0')

        opener = urllib2.build_opener()

        response_data = ""

        try:

            response = urllib2.urlopen(request)

            response_data = response.read()

            print "Fetched response from flipkart for %s" %(url)

            redirect_url = response.url

        except urllib2.HTTPError as e:

            print 'ERROR: ', e

            print 'Retrying'

            self.count_trials += 1

            if self.count_trials < 3:

                return self.read(url)

        self.response_data=response_data

        return self.createData(url,redirect_url)

    def scrapeRedirectedPage(self,soup,redirect_url):

        print soup

        print redirect_url

        t = soup.find("div" , {"class" : "seller-table fk-user-select-none line"})

        print t

        table_rows = t.findAll("tr" , {"class" : re.compile('t-row.*')})

        print table_rows

        for x in table_rows:

            print x

    def createData(self,url, redirect_url):

        print "Creating soup from flipkart data for %s" %(url)

        print redirect_url

        page=self.response_data.decode("utf-8")

        self.soup = BeautifulSoup(page,convertEntities=BeautifulSoup.HTML_ENTITIES)

        page = None

        self.response_data = None

        print "Soup created from flipkart data for %s" %(url)

        if (url==redirect_url):

            return self.scrape(self.soup,url)

        else:

            print self.redirectCount

            self.redirectCount+=1

            if self.redirectCount >4:

                raise

            return self.read(url)

    def scrape(self,soup,url):

        print "Inside json creator for %s" %(url)

        info = []

        oddSeller = soup.findAll("div" , {"class" : "line seller-item odd "})

        for data in oddSeller:

            temp={}

            try:

                businessDays = data.find('span', attrs={'class' : re.compile('fk-deliverable.*')})

                shippingTime = businessDays.find('span', attrs={'class' : re.compile('fk-bold')}).string.replace('to','').replace('business days.','').strip().replace('  ','-')

                temp['shippingTime']=shippingTime

            except:

                pass

            price = data.find('span', attrs={'class' : re.compile('pxs-final-price.*')}).string.strip('Rs.').strip()

            temp['sellingPrice']=float(price)

            for sellerInfo in data.findAll("div",{"class":re.compile(".*seller-info*")}):

                sellerName = sellerInfo.find('a').string

                temp['sellerName'] = sellerName

            for metrics in data.find("div",{"class":"fk-text-right"}):

                try:

                    metric = metrics.findAll('input', {'type': 'submit'})

                except AttributeError:

                    continue

                try:

                    inputTags = metric[0]['data-lst-buytrend']

                except TypeError:

                    continue

                dataMetrics = metric[0]['data-listing-metrics']

                try:

                    buyTrend = inputTags[0:str(inputTags).index('NWSR')].replace('_','')

                except ValueError:

                    buyTrend = inputTags[0:str(inputTags).index('WSR')].replace('_','')

                temp['buyTrend']=buyTrend

                dataMetric = dataMetrics.split(';')

                sellerCode = dataMetric[0]

                temp['sellerCode']=sellerCode

                temp['sellingPriceMetric'] = float(dataMetric[1])

                if not temp.has_key('shippingTime'):

                    print "Populating shipping time from metrics"

                    temp['shippingTime'] = dataMetric[3]

                temp['sellerScore'] = int(dataMetric[4])

                info.append(temp)

        evenSeller = soup.findAll("div" , {"class" : "line seller-item even "})

        for data in evenSeller:

            temp={}

            price = data.find('span', attrs={'class' : re.compile('pxs-final-price.*')}).string.strip('Rs.')

            try:

                businessDays = data.find('span', attrs={'class' : re.compile('fk-deliverable.*')})

                shippingTime = businessDays.find('span', attrs={'class' : re.compile('fk-bold')}).string.replace('to','').replace('business days.','').strip().replace('  ','-')

                temp['shippingTime']=shippingTime

            except:

                pass

            temp['sellingPrice']=float(price)

            for sellerInfo in data.findAll("div",{"class":re.compile(".*seller-info*")}):

                sellerName = sellerInfo.find('a').string

                temp['sellerName'] = sellerName

            for metrics in data.find("div",{"class":"fk-text-right"}):

                try:

                    metric = metrics.findAll('input', {'type': 'submit'})

                except AttributeError:

                    continue

                try:

                    inputTags = metric[0]['data-lst-buytrend']

                except TypeError:

                    continue

                dataMetrics = metric[0]['data-listing-metrics']

                try:

                    buyTrend = inputTags[0:str(inputTags).index('NWSR')].replace('_','')

                except ValueError:

                    buyTrend = inputTags[0:str(inputTags).index('WSR')].replace('_','')

                temp['buyTrend']=buyTrend

                dataMetric = dataMetrics.split(';')

                temp['sellerCode'] = dataMetric[0] 

                temp['sellingPriceMetric'] = float(dataMetric[1])

                if not temp.has_key('shippingTime'):

                    print "Populating shipping time from metrics"

                    temp['shippingTime'] = dataMetric[3]

                temp['sellerScore'] = int(dataMetric[4])

                info.append(temp)

        print "Returning Json response from flipkart for %s" %(url)

        return info

if __name__ == '__main__':

    scraper = FlipkartScraper()

    print scraper.read('http://www.flipkart.com/ps/MOBDYFURT9PKAPSX')