427 lines
15 KiB
Python
427 lines
15 KiB
Python
#-*- coding: utf-8 -*-
|
|
'''
|
|
Created on 2015. 12. 8.
|
|
|
|
@author: cococo
|
|
'''
|
|
import re
|
|
import datetime
|
|
import insta.instaparser as instaparser
|
|
import insta.instaheaders as instaheaders
|
|
import requests
|
|
|
|
from base.baseclasses import SendtoDB
|
|
from base.baseclasses import print_and_flush
|
|
from base.baseclasses import CrawlInit
|
|
from base.baseclasses import wait
|
|
|
|
|
|
def printl(*objects, sep=' ', end='\n', file=None, flush=True):
|
|
print(*objects, sep=sep, end=end, file=file, flush=flush)
|
|
|
|
insta_url = "https://www.instagram.com/"
|
|
insta_tag_url = "https://www.instagram.com/explore/tags/"
|
|
insta_query = "https://www.instagram.com/query/"
|
|
insta_body_url = 'https://www.instagram.com/p/'
|
|
|
|
|
|
class InstaInit(CrawlInit):
|
|
def __init__(self, before_day=0):
|
|
super().__init__(before_day)
|
|
self.urls = dict()
|
|
self.urls[9] = insta_tag_url
|
|
self.urls[10] = insta_url
|
|
|
|
def split_searches(self):
|
|
search = self.searches()
|
|
splited_list = search.split(',')
|
|
trimmed_list = list()
|
|
if self.platform() == 10:
|
|
for x in splited_list:
|
|
trimmed_list.append(x.strip())
|
|
else:
|
|
for x in splited_list:
|
|
trimmed_list.append(self.utf8(x))
|
|
return trimmed_list
|
|
|
|
def make_url(self):
|
|
urls = list()
|
|
for x in self.split_searches():
|
|
url = self.urls[self.platform()] + x
|
|
urls.append(url)
|
|
return urls
|
|
|
|
def get_begin_day(self):
|
|
if self.is_realtime():
|
|
date_now = datetime.datetime.now()
|
|
result = datetime.datetime(year=date_now.year, month=date_now.month, day=date_now.day)
|
|
result += datetime.timedelta(days=self.before_day)
|
|
return result
|
|
else:
|
|
return self.start_day()
|
|
|
|
def get_end_day(self):
|
|
if self.is_realtime():
|
|
date_now = datetime.datetime.now()
|
|
result = datetime.datetime(year=date_now.year, month=date_now.month, day=date_now.day)
|
|
return result
|
|
else:
|
|
return self.end_day()
|
|
|
|
|
|
class ListTag:
|
|
def __init__(self, url):
|
|
self.__r = None
|
|
self.__tag = ''
|
|
self.__url = ''
|
|
self.list_tag = []
|
|
self.end_cursor = None
|
|
self.has_next = False
|
|
self.cookies = {}
|
|
self.load_url(url)
|
|
|
|
def load_url(self, url):
|
|
self.__r = requests.get(url, headers=instaheaders.get_headers_for_list_html())
|
|
printl("<ListTag Start>")
|
|
printl("<ListTag requests>")
|
|
printl('headers = ', end=' ')
|
|
printl(instaheaders.get_headers_for_list_html())
|
|
|
|
self.__r.raise_for_status()
|
|
self.__tag = self.__get_tag(url)
|
|
self.__set_cookies(self.__r.cookies)
|
|
self.__url = url
|
|
self.list_tag, self.end_cursor, self.has_next = instaparser.parse_list_tag_html(self.__r.content)
|
|
printl("<ListTag response>")
|
|
printl('self.__r.cookies=', end='')
|
|
printl(self.__r.cookies)
|
|
printl('end_cursor = ' + str(self.end_cursor))
|
|
printl('has_next = ', end='')
|
|
printl(self.has_next)
|
|
printl("<ListTag End>")
|
|
|
|
return self.list_tag
|
|
|
|
def load_more(self):
|
|
form_data = instaheaders.get_form_data_for_list_tag(self.__tag, self.end_cursor, 12)
|
|
headers = instaheaders.get_headers_for_ajax(self.cookies, self.__url, form_data)
|
|
printl("<ListTag Start>")
|
|
printl("<ListTag requests>")
|
|
printl('end_cursor = ' + str(self.end_cursor))
|
|
printl('form_data' + form_data)
|
|
printl('headers = ', end=' ')
|
|
printl(headers)
|
|
self.__r = requests.post(insta_query, headers=headers, data=form_data)
|
|
self.__set_cookies(self.__r.cookies)
|
|
self.__r.raise_for_status()
|
|
self.list_tag, self.end_cursor, self.has_next = instaparser.parse_list_ajax(self.__r.content)
|
|
printl("<ListTag response>")
|
|
printl('self.__r.cookies=', end='')
|
|
printl(self.__r.cookies)
|
|
printl('end_cursor = ' + str(self.end_cursor))
|
|
printl('has_next = ', end='')
|
|
printl(self.has_next)
|
|
printl("<ListTag End>")
|
|
return self.list_tag
|
|
|
|
def __get_tag(self, url):
|
|
m = re.search(insta_tag_url + "([^/]*)", url)
|
|
if m:
|
|
return m.group(1)
|
|
else:
|
|
raise RuntimeError('Tag Error')
|
|
|
|
def get_cookies(self):
|
|
return self.cookies
|
|
|
|
def get_url(self):
|
|
return self.__url
|
|
|
|
def set_end_cursor(self, cursor):
|
|
self.end_cursor = cursor
|
|
|
|
def get_end_cursor(self):
|
|
return self.end_cursor
|
|
|
|
def __set_cookies(self, cookies):
|
|
for k, v in cookies.items():
|
|
self.cookies[k] = v
|
|
|
|
def get_list(self):
|
|
return self.list_tag
|
|
|
|
|
|
class ListUser:
|
|
def __init__(self, url):
|
|
self.__r = None
|
|
self.__user = ''
|
|
self.__url = ''
|
|
self.list_user = []
|
|
self.end_cursor = None
|
|
self.has_next = False
|
|
self.cookies = {}
|
|
self.load_url(url)
|
|
|
|
def load_url(self, url):
|
|
self.__r = requests.get(url, headers=instaheaders.get_headers_for_list_html())
|
|
self.__r.raise_for_status()
|
|
self.__url = url
|
|
self.__set_cookies(self.__r.cookies)
|
|
self.list_user, self.end_cursor, self.has_next, self.__user = instaparser.parse_list_user_html(self.__r.content)
|
|
return self.list_user
|
|
|
|
def load_more(self):
|
|
form_data = instaheaders.get_form_data_for_list_user(self.__user, self.end_cursor, 24)
|
|
headers = instaheaders.get_headers_for_ajax(self.cookies, self.__url, form_data)
|
|
printl("<ListUser Start>")
|
|
printl("<ListUser requests>")
|
|
printl('end_cursor = ' + str(self.end_cursor))
|
|
printl('form_data' + form_data)
|
|
printl('headers = ', end=' ')
|
|
printl(headers)
|
|
self.__r = requests.post(insta_query, headers=headers, data=form_data)
|
|
self.__r.raise_for_status()
|
|
self.__set_cookies(self.__r.cookies)
|
|
printl("<ListUser response>")
|
|
printl('self.__r.cookies=', end='')
|
|
printl(self.__r.cookies)
|
|
printl('end_cursor = ' + str(self.end_cursor))
|
|
printl('has_next = ', end='')
|
|
printl(self.has_next)
|
|
printl("<ListUser End>")
|
|
|
|
self.list_user, self.end_cursor, self.has_next = instaparser.parse_list_ajax(self.__r.content)
|
|
return self.list_user
|
|
|
|
def get_cookies(self):
|
|
return self.cookies
|
|
|
|
def get_url(self):
|
|
return self.__url
|
|
|
|
def set_end_cursor(self, cursor):
|
|
self.end_cursor = cursor
|
|
|
|
def get_end_cursor(self):
|
|
return self.end_cursor
|
|
|
|
def __set_cookies(self, cookies):
|
|
for k, v in cookies.items():
|
|
self.cookies[k] = v
|
|
|
|
def get_list(self):
|
|
return self.list_user
|
|
|
|
|
|
class InstaContent:
|
|
def __init__(self, url, cookies, referer):
|
|
self.__r = None
|
|
self.__referer = ''
|
|
self.__code = ''
|
|
self.body = None
|
|
self.reply = []
|
|
self.start_cursor = None
|
|
self.has_previous = False
|
|
self.cookies = {}
|
|
self.load_url(url, cookies, referer)
|
|
|
|
def load_url(self, url, cookies, referer):
|
|
self.__set_cookies(cookies)
|
|
self.__r = requests.get(url, headers=instaheaders.get_headers_for_body_html(self.cookies))
|
|
self.__r.raise_for_status()
|
|
self.__referer = referer
|
|
self.__code = self.__get_code(url)
|
|
self.body, self.reply, self.start_cursor, self.has_previous = instaparser.parse_body_html(self.__r.content)
|
|
return self.body, self.reply
|
|
|
|
def get_body(self):
|
|
return self.body
|
|
|
|
def get_reply(self):
|
|
return self.reply
|
|
|
|
def load_reply_more(self):
|
|
form_data = instaheaders.get_form_data_for_reply(self.__code, self.start_cursor, 20)
|
|
headers = instaheaders.get_headers_for_ajax(self.cookies, self.__referer, form_data)
|
|
printl("<ContentReply Start>")
|
|
printl("<ContentReply requests>")
|
|
printl('start_cursor = ' + self.start_cursor)
|
|
printl('form_data' + form_data)
|
|
printl('headers = ', end=' ')
|
|
printl(headers)
|
|
self.__r = requests.post(insta_query, headers=headers, data=form_data)
|
|
self.__r.raise_for_status()
|
|
self.__set_cookies(self.__r.cookies)
|
|
self.reply, self.start_cursor, self.has_previous = instaparser.parse_reply_ajax(self.__r.content)
|
|
printl("<ContentReply response>")
|
|
printl('self.__r.cookies=', end='')
|
|
printl(self.__r.cookies)
|
|
printl('start_cursor = ' + str(self.start_cursor))
|
|
printl('has_previous = ', end='')
|
|
printl(self.has_previous)
|
|
printl("<ContentReply End>")
|
|
return self.reply
|
|
|
|
def get_cookies(self):
|
|
return self.cookies
|
|
|
|
def __get_code(self, url):
|
|
m = re.search(insta_body_url + "([^/]*)", url)
|
|
if m:
|
|
return m.group(1)
|
|
else:
|
|
raise RuntimeError('Tag Error')
|
|
|
|
def __set_cookies(self, cookies):
|
|
for k, v in cookies.items():
|
|
self.cookies[k] = v
|
|
|
|
|
|
class InstaMainCrawler:
|
|
def __init__(self):
|
|
self.send_to_db = SendtoDB()
|
|
self.crawl_init = InstaInit()
|
|
|
|
def set_keyword_id(self, keyword_id):
|
|
self.keyword_id = keyword_id
|
|
|
|
def crawl_all(self, backup_set=None):
|
|
pass
|
|
|
|
def crawl_content(self, url, cookies, referer):
|
|
content = InstaContent(url, cookies, referer)
|
|
body = content.get_body()
|
|
replies = content.get_reply()
|
|
body['article_url'] = url
|
|
body['keyword_id'] = self.keyword_id
|
|
#printl(body['article_url'])
|
|
while content.has_previous:
|
|
replies = content.load_reply_more() + replies
|
|
wait(2)
|
|
for j in range(0, len(replies)):
|
|
replies[j]['article_url'] = body['article_url']
|
|
replies[j]['platform_id'] = body['platform_id']
|
|
replies[j]['article_order'] = j
|
|
self.send_to_db.delete_url(body['article_url'])
|
|
self.send_to_db.send_body(body)
|
|
if replies:
|
|
self.send_to_db.send_reply(replies)
|
|
printl('ok')
|
|
printl()
|
|
|
|
def start(self):
|
|
self.crawler_start()
|
|
|
|
def set_arguments(self, browser, keyword_id, db_num, before_day, until_page):
|
|
self.init_keyword_id(keyword_id)
|
|
self.init_db(db_num)
|
|
self.init_before_day(before_day)
|
|
self.init_until_page(until_page)
|
|
|
|
def init_browser(self, browser):
|
|
pass
|
|
|
|
def init_keyword_id(self, keyword_id):
|
|
if type(keyword_id) != int:
|
|
self.keyword_id = int(keyword_id)
|
|
else:
|
|
self.keyword_id = keyword_id
|
|
self.crawl_init.get_keyword_parameters(keyword_id)
|
|
self.crawl_init.disconnect()
|
|
|
|
def init_db(self, db_num):
|
|
self.send_to_db.set_db(db_num)
|
|
|
|
def init_before_day(self, before_day):
|
|
self.crawl_init.set_before_day(before_day)
|
|
|
|
def init_until_page(self, until_page):
|
|
self.crawl_init.set_until_page(until_page)
|
|
|
|
def crawler_start(self):
|
|
real_time = True
|
|
|
|
while real_time:
|
|
print_and_flush("Crawling Start")
|
|
url_list = self.crawl_init.make_url()
|
|
i = 0
|
|
end_cursor = None
|
|
backup_set = set()
|
|
while i < len(url_list):
|
|
# first connect
|
|
try:
|
|
printl(url_list[i] + "\n")
|
|
if insta_tag_url in url_list[i]:
|
|
list_crawler = ListTag(url_list[i])
|
|
else:
|
|
list_crawler = ListUser(url_list[i])
|
|
wait(1)
|
|
insta_list = list_crawler.get_list()
|
|
is_load_more = list_crawler.has_next
|
|
for element in insta_list:
|
|
old_elements = 0
|
|
if element['date'].date() > self.crawl_init.get_end_day():
|
|
printl(element['url'])
|
|
printl(element['date'].strftime("%Y-%m-%d %H:%M:%S"))
|
|
continue
|
|
elif element['date'].date() < self.crawl_init.get_begin_day():
|
|
printl(element['url'])
|
|
printl(element['date'].strftime("%Y-%m-%d %H:%M:%S"))
|
|
old_elements += 1
|
|
if old_elements > 8:
|
|
is_load_more = False
|
|
break
|
|
else:
|
|
if not element['url'] in backup_set:
|
|
printl(element['url'])
|
|
printl(element['date'].strftime("%Y-%m-%d %H:%M:%S"))
|
|
wait(1.5)
|
|
self.crawl_content(element['url'], list_crawler.get_cookies(), list_crawler.get_url())
|
|
backup_set.add(element['url'])
|
|
|
|
# ajax load
|
|
while is_load_more:
|
|
if end_cursor:
|
|
list_crawler.end_cursor = end_cursor
|
|
end_cursor = None
|
|
wait(1)
|
|
insta_list = list_crawler.load_more()
|
|
is_load_more = list_crawler.has_next
|
|
old_elements = 0
|
|
printl("list length = " + str(len(insta_list)))
|
|
for element in insta_list:
|
|
if element['date'].date() > self.crawl_init.get_end_day():
|
|
printl(element['url'])
|
|
printl(element['date'].strftime("%Y-%m-%d %H:%M:%S"))
|
|
continue
|
|
elif element['date'].date() < self.crawl_init.get_begin_day():
|
|
printl(element['url'])
|
|
printl(element['date'].strftime("%Y-%m-%d %H:%M:%S"))
|
|
|
|
old_elements += 1
|
|
if old_elements > 8:
|
|
is_load_more = False
|
|
break
|
|
else:
|
|
if not element['url'] in backup_set:
|
|
printl(element['url'])
|
|
printl(element['date'].strftime("%Y-%m-%d %H:%M:%S"))
|
|
wait(1.5)
|
|
try:
|
|
self.crawl_content(element['url'], list_crawler.get_cookies(),
|
|
list_crawler.get_url())
|
|
except Exception as e:
|
|
printl(e)
|
|
backup_set.add(element['url'])
|
|
i += 1
|
|
except Exception as e:
|
|
printl(e)
|
|
end_cursor = list_crawler.end_cursor
|
|
printl('end_cursor=' + end_cursor)
|
|
if e.args:
|
|
wait(300)
|
|
real_time = self.crawl_init.is_realtime()
|
|
printl("Finished Crawling :)")
|
|
self.send_to_db.close()
|
|
#self.driver.quit()
|