GooglePoiCrawler/swire_docker_itemlist.py

# -*- coding: utf-8 -*-
#from selenium import webdriver
from seleniumwire import webdriver
from selenium.webdriver.common.action_chains import ActionChains
from selenium.webdriver.common.keys import Keys
from selenium.webdriver.support import expected_conditions as EC
from selenium.webdriver.support.wait import WebDriverWait
from selenium.webdriver.common.by import By
import selenium
import traceback
from datetime import datetime
import dataset
import time
import json
import re
import sys, os
import socket
import brotli
from selenium.webdriver.common.desired_capabilities import DesiredCapabilities
import urllib.parse
#chrome_window=False
chrome_window=True

globalkw=None
proxyport=8787


def build_cache(db):
    id_dict={}
    cursor = db.query('SELECT place_id FROM google_poi.swire_store_list;')

    for c in cursor:
        id_dict[c['place_id']]=1
    return id_dict
    #

def brower_start(port):
    global proxyport
    global chrome_window
    print(proxyport)
    options = webdriver.ChromeOptions()
    if chrome_window:
        options.add_argument('--ignore-certificate-errors')
        options.add_argument("--no-sandbox")
        options.add_argument("--headless")
        options.add_argument("--disable-dev-shm-usage")

        browser = webdriver.Chrome(
            desired_capabilities=options.to_capabilities()
        )
    else:
        chrome_options = webdriver.ChromeOptions()
        chrome_options.add_argument('--proxy-server=host.docker.internal:'+str(proxyport))  # Specify your Kubernetes service-name here
        chrome_options.add_argument('--ignore-certificate-errors')
        chrome_options.add_argument("--no-sandbox")
        chrome_options.add_argument("--disable-dev-shm-usage")
        browser = webdriver.Remote(
            command_executor='http://127.0.0.1:'+str(port)+'/wd/hub',
            desired_capabilities=chrome_options.to_capabilities(),
            seleniumwire_options={'addr':'0.0.0.0','port':proxyport,'auto_config': False}

            )
#            seleniumwire_options = {'addr': '172.17.0.2','port':4444})
        browser.set_window_size(1400,1000)
    return browser


def page_down_(driver, xpath_css, time_):
    e = driver.find_element_by_css_selector('span[class="Jl2AFb"]')
    result_count = e.text.split('-')[1].replace(' 項結果','')
    print(result_count)
    if int(result_count) > 5:
        for i in range(time_):
            e = driver.find_elements_by_css_selector('div[class="TFQHme"]')
            action = webdriver.common.action_chains.ActionChains(driver)
            action.move_to_element_with_offset(e[-1], e[-1].size['width'] + 1 , 0)
            action.click()
            action.perform()
            time.sleep(0.5)


def get_url_list(driver):
    page_down_(driver, '//div[@class="TFQHme"]', 8)

    url_soup = BeautifulSoup(driver.page_source, 'html.parser')
    url_list = []
    for i in url_soup.find_all('a'):
        try:
            if i['href'].find('maps/place') != -1:
                url_list += [[i['href'], i['aria-label']]]
        except:
            pass
    # print(len(url_list))
    return url_list


def keyin_keyword(driver, keyword):
    button = driver.find_element_by_id("searchbox")
    driver.implicitly_wait(30)
    ActionChains(driver).move_to_element(button).send_keys(keyword).send_keys(Keys.RETURN).perform()
    time.sleep(3)


def scan_job(db,kw):
    result={'kw':kw}
    cursor = db.query('select t1.num,next-prev as diff from google_poi.conv_log t1, (SELECT num,max(id) mid  FROM google_poi.conv_log group by num  ) t2 where t1.id=t2.mid having diff>0 order by rand()')
    for c in cursor:
        result['num']=c['num']
        break
    cursor = db.query('select lat,lon,loc from lat_lon_loc where num ="'+str(result['num'])+'"')
    for c in cursor:
        result['lat']=c['lat']
        result['lon']=c['lon']
        result['loc']=c['loc']
        return result

def get_next_job(db,repeat=False,repkw=None,repnum=None):
    global globalkw

    result={}
#    if globalkw is not None:
#        cursor = db.query('select distinct(kw),num+1 as num from swire_progress_list where kw="'+globalkw+'"')
#    else:
#        cursor = db.query('select distinct(kw),num+1 as num from swire_progress_list where num < 367 order by rand() limit 1')

#    cursor = db.query('select kw,num  from areacodes where expand=0 order by rand()')
    cursor = db.query('select kw,num  from areacodes order by rand()')

    for c in cursor:
#        repkw=c['kw']
        if repkw is None:
            repkw=c['kw']
        result['kw']=c['kw']
        result['num']=c['num']
        break
    if repkw is not None:
        result['kw']=repkw
    if result.get('num') is not None:
        cursor = db.query('select lat,lon,loc from lat_lon_loc where num ="'+str(result['num'])+'"')
        for c in cursor:
            result['lat']=c['lat']
            result['lon']=c['lon']
            result['loc']=c['loc']
            break

    if repeat and repkw!= 'REP':
        result['kw']=repkw
        result['num']=repnum

    if 'REP' in repkw:
        if repnum=='REP':
            repnum=None
#            cursor = db.query('select  num from swire_store_list where num not in (select num from conv_log) order by rand() limit 1')

            cursor = db.query('select  num from swire_store_list  order by rand() limit 1')
            for c in cursor:
                repnum=c['num']
                break
        if repnum is None:
            cursor = db.query('select  num from swire_store_list  order by rand() limit 1')
            for c in cursor:
                repnum=c['num']
                break


#        cursor = db.query('select  lat_txt,lon_txt,keyword,num from swire_store_list order by rand() limit 1')
        cursor = db.query('select  lat_txt,lon_txt,keyword,num from swire_store_list where num="'+str(repnum)+'" limit 1')

        for c in cursor:
            result['kw']=c['keyword']
            result['num']=c['num']
            result['lat']=c['lat_txt']
            result['lon']=c['lon_txt']
            result['loc']=''
            return result
    


    if repeat:
#        cursor = db.query('select  lat_txt,lon_txt from swire_store_list where num ="'+str(result['num'])+'" and keyword="'+result['kw']+'" order by rand() limit 1')
        cursor = db.query('select  lat_txt,lon_txt,keyword from swire_store_list order by rand() limit 1')

        for c in cursor:
            result['kw']=c['keyword']
            result['lat']=c['lat_txt']
            result['lon']=c['lon_txt']

    return result


def write_to_file(jsobj,fname):
    import codecs
    fw=codecs.open(fname,'w','utf-8')
    fw.write(str(jsobj))
    fw.close()

def parsing_js(orig):
    resultobj=[]
    content=""
    lines=orig.split('\n')
    for l in lines:
        newl=l.replace('\\"','"')
#        if '\\\\"' in newl:
#            print(newl)
#        newl=newl.repace('\\\\"','')
        newl=newl.replace('\\"','"')

        content+=newl
    result=re.search(r'\[\["',content)
    print(result)
    content_begin=result.start()

    result=re.search(r'\]\]"',content)
    print(result)

    content_end=result.end()

    jscontent=content[content_begin:content_end-1]
#    write_to_file(jscontent,'c:/tmp/debug.txt')
    jsobj=json.loads(jscontent)
    for x in jsobj[0][1][1:]:
        print(x[14][11])
        print(x[14][9])
        reviews_cnt=None
        photo=None
        rating=None
        biz_id=None
        loc_x=None
        loc_y=None
        addr_elmts=None
        tel=None
        try:
            rating=x[14][4][7]
            reviews_cnt=x[14][4][8]
        except:
            traceback.print_exc()

        try:
            photo=x[14][37][0][0][0]
            num_photos=x[14][37][0][0][6][1]
        except:
            traceback.print_exc()

        try:
            loc_x=x[14][37][0][0][29][0]
            loc_y=x[14][37][0][0][29][1]
        except:
            traceback.print_exc()

        try:
            biz_id=x[14][57][2]
            tel=x[14][178][0][3]
        except:
            traceback.print_exc()

        try:
            addr_elmts=str(x[14][82])
        except:
            traceback.print_exc()




        category=str(x[14][13])
        topic=str(x[14][89])
        print(x[14][13])

        print(x[14][10])
        print(x[14][2])
        print(x[14][78])
        try:
            resultobj.append({'name':x[14][11],'fid':x[14][10],'addr':x[14][2][0],'addr_elmts':addr_elmts,'place_id':x[14][78],'category':category,'rating':rating,'reviews_cnt':reviews_cnt,'lat':x[14][9][2],'lat_txt':str(x[14][9][2]),'lon':x[14][9][3],'lon_txt':str(x[14][9][3]),'topic':topic,'photo':photo,'num_photos':num_photos,'loc_x':loc_x,'loc_y':loc_y,'biz_id':biz_id,'tel':tel,'crawler_date':datetime.today().strftime("%Y/%m/%d %H:%M")})
        except:
            traceback.print_exc()
    return resultobj

def save_js_to_db(jsobj,num,keyword):
    global store_list_table
    global iddict
    for r in jsobj:
        if iddict.get(r['place_id']) is not None:
            continue
        r['num']=num
        r['keyword']=keyword

        try:
            store_list_table.insert(r)

#            store_list_table.upsert(r,keys=['place_id'])
        except:
            traceback.print_exc()
#        store_list_table.upsert(r,keys=['place_id'])

def process_web_request(db,driver,area_num,keyword):
    global prev_cnt
#    query = WebDriverWait(driver, 30).until(EC.presence_of_element_located((By.XPATH, '//button[@vet="19128"]')))
    time.sleep(0.8)
    time.sleep(3)
    print("ppppppppp&**********************")
    for request in driver.requests:
        if 'search?' in request.url :
            print('searching.....')
#        else:
#            print(request.url[20:60])
        if request.response:
#            if 'https://www.google.com.tw/search?tbm=map' in request.url :
            if 'search?' in request.url :
                print('parsing js:')
                resp = brotli.decompress(request.response.body)
                jstext=resp.decode('utf-8')
                resultobj=parsing_js(jstext)
                print("before",datetime.now())
                print("num: "+str(area_num))
                save_js_to_db(resultobj,area_num,keyword)
                print("after",datetime.now())

                aft_cnt=0
                cursor = db.query('select count(*) as cnt from swire_store_list where num="'+str(area_num)+'" ')
                for c in cursor:
                    aft_cnt=c['cnt']
                    break
                db['conv_log'].insert({'num':area_num,'prev':prev_cnt,'next':aft_cnt,'dt':datetime.now()})



#    time.sleep(9999)


def main():
    global chrome_window
    global store_list_table
    global globalkw
    global proxyport
    global iddict
    global prev_cnt

    port=4444
    # if len(sys.argv) == 3 :
    #     port=int(sys.argv[1])
    #     proxyport=int(sys.argv[2])

    if len(sys.argv)>1:
        globalkw=sys.argv[1]
        port=int(sys.argv[2])
        proxyport=int(sys.argv[3])
    print(globalkw, port, proxyport)
    failcnt=0
    localip=socket.gethostbyname(socket.gethostname())
#    if localip=='192.168.1.108':
#        chrome_window=True
#        chrome_window=False

    db = dataset.connect('mysql://choozmo:pAssw0rd@db.ptt.cx:3306/google_poi?charset=utf8mb4')
    iddict=build_cache(db)
    store_list_table = db['swire_store_list']

#    table2 = db['swire_progress_list']
    table2 = db['swire_area_progress']

    if not chrome_window:
        print('restart docker p{}'.format(port))
#        os.system('sudo docker container restart p'+str(port))
        os.system('docker container restart p'+str(port))

        time.sleep(10)

    print('drvier start...')
    driver = brower_start(port)
    
    area_num=None
    while True:
        try:
            if len(sys.argv) > 4 :
                repkw=sys.argv[1]
                repnum=sys.argv[2]
                if 'SCAN' in repkw:
                    job=scan_job(db,repnum)
                else:
                    job=get_next_job(db,repeat=True,repkw=repkw,repnum=repnum)
            else:
                job=get_next_job(db, repkw=globalkw)
            print(job)
            keyword  = job['kw']
            latitude = job['lat'] #緯度
            longitude = job['lon'] #精度
            area_num=job['num']
            safe_string = urllib.parse.quote_plus(keyword)
            url = 'https://www.google.com.tw/maps/@{},{},18z?hl=zh-TW'.format(latitude, longitude)

            prev_cnt=0
            cursor = db.query('select count(*) as cnt from swire_store_list where num="'+str(area_num)+'" ')
            for c in cursor:
                prev_cnt=c['cnt']
                break


#            url = 'https://www.google.com/maps/search/'+safe_string+'/@{},{},16z/data=!3m1!4b1'.format(latitude, longitude)
#            url='https://www.google.com/maps/search/'+safe_string+'/@24.7962279,121.0449762,15z/data=!3m1!4b1?hl=zh-TW'
#            print(url)
#            url='https://www.google.com/maps/search/%E7%81%AB%E9%8D%8B%E9%A4%90%E5%BB%B3/@24.772608,121.0515456,13z'
            driver.get(url)
#            time.sleep(3)
            keyin_keyword(driver, keyword)

            process_web_request(db,driver,area_num,keyword)

            pagecnt=0
            while True:
                element = driver.find_element_by_id('ppdPk-Ej1Yeb-LgbsSe-tJiF1e')
                if element.get_attribute('disabled'):
                    break
    #               driver.implicitly_wait(30)
                ActionChains(driver).move_to_element(element).click(element).perform() 
                process_web_request(db,driver,area_num,keyword)
                pagecnt+=1
                if pagecnt>=5:
                    break


#            table2.upsert({'kw':keyword,'num':job['num']},['kw'])
            table2.insert({'kw':keyword,'num':job['num']},['kw'])
            db.query('update areacodes set expand = 1 where num="'+str(job['num'])+'" and kw="'+keyword+'" ')

        except:
            traceback.print_exc()
            failcnt+=1
            if failcnt>=15:
                sys.exit()
            pass



if __name__ == '__main__':
    main()