chg: [update v1.5] add background update: screenshots_crawled

This commit is contained in:
Terrtia 2019-04-24 16:19:35 +02:00
parent 9868833c77
commit e6dca7f8bf
No known key found for this signature in database
GPG key ID: 1E1B1F50D84613D0
5 changed files with 146 additions and 12 deletions

View file

@ -140,7 +140,7 @@ class TorSplashCrawler():
UUID = self.domains[0]+str(uuid.uuid4())
filename_paste_full = os.path.join(self.crawled_paste_filemame, UUID)
relative_filename_paste = os.path.join(self.crawler_path, UUID)
filename_har = os.path.join(self.crawled_har, UUID +'.png')
filename_har = os.path.join(self.crawled_har, UUID)
# # TODO: modify me
# save new paste on disk
@ -180,17 +180,12 @@ class TorSplashCrawler():
self.r_serv_metadata.sadd('paste_children:'+response.meta['father'], relative_filename_paste)
dirname = os.path.dirname(filename_har)
if not os.path.exists(dirname):
os.makedirs(dirname)
if 'png' in response.data:
size_screenshot = (len(response.data['png'])*3) /4
if size_screenshot < 5000000: #bytes
image_content = base64.standard_b64decode(response.data['png'].encode())
hash = sha256(image_content).hexdigest()
print(hash)
img_dir_path = os.path.join(hash[0:2], hash[2:4], hash[4:6], hash[6:8], hash[8:10], hash[10:12])
filename_img = os.path.join(self.crawled_screenshot, 'screenshot', img_dir_path, hash[12:] +'.png')
dirname = os.path.dirname(filename_img)
@ -202,13 +197,13 @@ class TorSplashCrawler():
# add item metadata
self.r_serv_metadata.hset('paste_metadata:{}'.format(relative_filename_paste), 'screenshot', hash)
# add sha256 metadata
self.r_serv_onion.zincrby('screenshot:{}'.format(hash), relative_filename_paste, 1)
self.r_serv_onion.sadd('screenshot:{}'.format(hash), relative_filename_paste)
if 'har' in response.data:
dirname = os.path.dirname(filename_har)
if not os.path.exists(dirname):
os.makedirs(dirname)
with open(filename_har+'har.txt', 'wb') as f:
with open(filename_har+'.json', 'wb') as f:
f.write(json.dumps(response.data['har']).encode())
# save external links in set

View file

@ -31,7 +31,7 @@ if __name__ == "__main__":
db=cfg.getint("ARDB_DB", "db"),
decode_responses=True)
if r_serv.scard('ail:update_v1.5') != 4:
if r_serv.scard('ail:update_v1.5') != 5:
r_serv.delete('ail:update_error')
r_serv.set('ail:update_in_progress', 'v1.5')
r_serv.set('ail:current_background_update', 'v1.5')
@ -50,7 +50,10 @@ if __name__ == "__main__":
if not r_serv.sismember('ail:update_v1.5', 'tags_background'):
update_file = os.path.join(os.environ['AIL_HOME'], 'update', 'v1.4', 'Update-ARDB_Tags_background.py')
process = subprocess.run(['python' ,update_file])
if r_serv.scard('ail:update_v1.5') != 4:
if not r_serv.sismember('ail:update_v1.5', 'crawled_screenshot'):
update_file = os.path.join(os.environ['AIL_HOME'], 'update', 'v1.4', 'Update-ARDB_Onions_screenshots.py')
process = subprocess.run(['python' ,update_file])
if r_serv.scard('ail:update_v1.5') != 5:
r_serv.set('ail:update_error', 'Update v1.5 Failed, please relaunch the bin/update-background.py script')
else:
r_serv.delete('ail:update_in_progress')

View file

@ -0,0 +1,133 @@
#!/usr/bin/env python3
# -*-coding:UTF-8 -*
import os
import sys
import time
import redis
import datetime
import configparser
from hashlib import sha256
def rreplace(s, old, new, occurrence):
li = s.rsplit(old, occurrence)
return new.join(li)
def substract_date(date_from, date_to):
date_from = datetime.date(int(date_from[0:4]), int(date_from[4:6]), int(date_from[6:8]))
date_to = datetime.date(int(date_to[0:4]), int(date_to[4:6]), int(date_to[6:8]))
delta = date_to - date_from # timedelta
l_date = []
for i in range(delta.days + 1):
date = date_from + datetime.timedelta(i)
l_date.append( date.strftime('%Y%m%d') )
return l_date
if __name__ == '__main__':
start_deb = time.time()
configfile = os.path.join(os.environ['AIL_BIN'], 'packages/config.cfg')
if not os.path.exists(configfile):
raise Exception('Unable to find the configuration file. \
Did you set environment variables? \
Or activate the virtualenv.')
cfg = configparser.ConfigParser()
cfg.read(configfile)
SCREENSHOT_FOLDER = os.path.join(os.environ['AIL_HOME'], cfg.get("Directories", "crawled_screenshot"))
NEW_SCREENSHOT_FOLDER = os.path.join(os.environ['AIL_HOME'], cfg.get("Directories", "crawled_screenshot"), 'screenshot')
PASTES_FOLDER = os.path.join(os.environ['AIL_HOME'], cfg.get("Directories", "pastes")) + '/'
r_serv = redis.StrictRedis(
host=cfg.get("ARDB_DB", "host"),
port=cfg.getint("ARDB_DB", "port"),
db=cfg.getint("ARDB_DB", "db"),
decode_responses=True)
r_serv_metadata = redis.StrictRedis(
host=cfg.get("ARDB_Metadata", "host"),
port=cfg.getint("ARDB_Metadata", "port"),
db=cfg.getint("ARDB_Metadata", "db"),
decode_responses=True)
r_serv_tag = redis.StrictRedis(
host=cfg.get("ARDB_Tags", "host"),
port=cfg.getint("ARDB_Tags", "port"),
db=cfg.getint("ARDB_Tags", "db"),
decode_responses=True)
r_serv_onion = redis.StrictRedis(
host=cfg.get("ARDB_Onion", "host"),
port=cfg.getint("ARDB_Onion", "port"),
db=cfg.getint("ARDB_Onion", "db"),
decode_responses=True)
r_serv.set('ail:current_background_script', 'crawled_screenshot')
r_serv.set('ail:current_background_script_stat', 0)
## Update Onion ##
print('Updating ARDB_Onion ...')
index = 0
start = time.time()
# clean down domain from db
date_from = '20180801'
date_today = datetime.date.today().strftime("%Y%m%d")
list_date = substract_date(date_from, date_today)
nb_done = 0
last_progress = 0
total_to_update = len(list_date)
for date in list_date:
screenshot_dir = os.path.join(SCREENSHOT_FOLDER, date[0:4], date[4:6], date[6:8])
if os.path.isdir(screenshot_dir):
print(screenshot_dir)
for file in os.listdir(screenshot_dir):
if file.endswith(".png"):
index += 1
#print(file)
img_path = os.path.join(screenshot_dir, file)
with open(img_path, 'br') as f:
image_content = f.read()
hash = sha256(image_content).hexdigest()
img_dir_path = os.path.join(hash[0:2], hash[2:4], hash[4:6], hash[6:8], hash[8:10], hash[10:12])
filename_img = os.path.join(NEW_SCREENSHOT_FOLDER, img_dir_path, hash[12:] +'.png')
dirname = os.path.dirname(filename_img)
if not os.path.exists(dirname):
os.makedirs(dirname)
if not os.path.exists(filename_img):
os.rename(img_path, filename_img)
item = os.path.join('crawled', date[0:4], date[4:6], date[6:8], file[:-4])
# add item metadata
r_serv_metadata.hset('paste_metadata:{}'.format(item), 'screenshot', hash)
# add sha256 metadata
r_serv_onion.sadd('screenshot:{}'.format(hash), item)
if file.endswith('.pnghar.txt'):
har_path = os.path.join(screenshot_dir, file)
new_file = rreplace(file, '.pnghar.txt', '.json', 1)
new_har_path = os.path.join(screenshot_dir, new_file)
os.rename(har_path, new_har_path)
progress = int((nb_done * 100) /total_to_update)
# update progress stats
if progress != last_progress:
r_serv.set('ail:current_background_script_stat', progress)
print('{}/{} screenshot updated {}%'.format(nb_done, total_to_update, progress))
last_progress = progress
nb_done += 1
r_serv.set('ail:current_background_script_stat', 100)
end = time.time()
print('Updating ARDB_Onion Done => {} paths: {} s'.format(index, end - start))
print()
print('Done in {} s'.format(end - start_deb))
r_serv.sadd('ail:update_v1.5', 'crawled_screenshot')

View file

@ -160,7 +160,7 @@ DiffMaxLineLength = int(cfg.get("Flask", "DiffMaxLineLength"))#Use to display t
bootstrap_label = ['primary', 'success', 'danger', 'warning', 'info']
dict_update_description = {'v1.5':{'nb_background_update': 4, 'update_warning_message': 'An Update is running on the background. Some informations like Tags, screenshot can be',
dict_update_description = {'v1.5':{'nb_background_update': 5, 'update_warning_message': 'An Update is running on the background. Some informations like Tags, screenshot can be',
'update_warning_message_notice_me': 'missing from the UI.'}
}

View file

@ -658,7 +658,10 @@ def show_domain():
unpack_url = faup.get()
domain = unpack_url['domain'].decode()
if not port:
port = unpack_url['port'].decode()
if unpack_url['port']:
port = unpack_url['port'].decode()
else:
port = 80
try:
port = int(port)
except: