kopia lustrzana https://github.com/bellingcat/auto-archiver
refactoring to use vk_url_scraper
rodzic
177e3a623e
commit
88ede91304
2
Pipfile
2
Pipfile
|
@ -22,8 +22,8 @@ google-auth-oauthlib = "*"
|
|||
oauth2client = "*"
|
||||
python-slugify = "*"
|
||||
pyyaml = "*"
|
||||
vk-api = "*"
|
||||
dateparser = "*"
|
||||
vk-url-scraper = "*"
|
||||
|
||||
[requires]
|
||||
python_version = "3.9"
|
||||
|
|
|
@ -1,7 +1,7 @@
|
|||
{
|
||||
"_meta": {
|
||||
"hash": {
|
||||
"sha256": "d06498403429a8fffcd6d049b314872c0095abee7fb9c6ffd3ba3d7b0c31c8cd"
|
||||
"sha256": "eacd9633c33d4d526d7737fc6bf83ab713205f28f819530f549378fbd14da3d8"
|
||||
},
|
||||
"pipfile-spec": 6,
|
||||
"requires": {
|
||||
|
@ -50,19 +50,19 @@
|
|||
},
|
||||
"boto3": {
|
||||
"hashes": [
|
||||
"sha256:0821212ff521cb934801b1f655cef3c0e976775324b1018f1751700d0f42dbb4",
|
||||
"sha256:87d34861727699c795bf8d65703f2435e75f12879bdd483e08b35b7c5510e8c8"
|
||||
"sha256:0b9757575b8003928defc5fb6e816936fa1bdb1384d0edec6622bb9fb104e96c",
|
||||
"sha256:f39b91a4c3614db8e44912ee82426fb4b16d5df2cd66883f3aff6f76d7f5d310"
|
||||
],
|
||||
"index": "pypi",
|
||||
"version": "==1.24.9"
|
||||
"version": "==1.24.12"
|
||||
},
|
||||
"botocore": {
|
||||
"hashes": [
|
||||
"sha256:5669b982b0583e73daef1fe0a4df311055e6287326f857dbb1dcc2de1d8412ad",
|
||||
"sha256:7a7588b0170e571317496ac4104803329d5bc792bc008e8a757ffd440f1b6fa6"
|
||||
"sha256:17d3ec9f684d21e06b64d9cb224934557bcd95031e2ecb551bf16271e8722fec",
|
||||
"sha256:b8ac156e55267da6e728ea0b806bfcd97adf882801cffe7849c4b88ce4780326"
|
||||
],
|
||||
"markers": "python_version >= '3.7'",
|
||||
"version": "==1.27.9"
|
||||
"version": "==1.27.12"
|
||||
},
|
||||
"brotli": {
|
||||
"hashes": [
|
||||
|
@ -149,11 +149,11 @@
|
|||
},
|
||||
"certifi": {
|
||||
"hashes": [
|
||||
"sha256:9c5705e395cd70084351dd8ad5c41e65655e08ce46f2ec9cf6c2c08390f71eb7",
|
||||
"sha256:f1d53542ee8cbedbe2118b5686372fb33c297fcd6379b050cca0ef13a597382a"
|
||||
"sha256:84c85a9078b11105f04f3036a9482ae10e4621616db313fe045dd24743a0820d",
|
||||
"sha256:fe86415d55e84719d75f8b69414f6438ac3547d2078ab91b67e779ef69378412"
|
||||
],
|
||||
"markers": "python_full_version >= '3.6.0'",
|
||||
"version": "==2022.5.18.1"
|
||||
"markers": "python_version >= '3.6'",
|
||||
"version": "==2022.6.15"
|
||||
},
|
||||
"cffi": {
|
||||
"hashes": [
|
||||
|
@ -308,11 +308,11 @@
|
|||
},
|
||||
"google-api-core": {
|
||||
"hashes": [
|
||||
"sha256:958024c6aa3460b08f35741231076a4dd9a4c819a6a39d44da9627febe8b28f0",
|
||||
"sha256:ce1daa49644b50398093d2a9ad886501aa845e2602af70c3001b9f402a9d7359"
|
||||
"sha256:06f7244c640322b508b125903bb5701bebabce8832f85aba9335ec00b3d02edc",
|
||||
"sha256:93c6a91ccac79079ac6bbf8b74ee75db970cc899278b97d53bc012f35908cf50"
|
||||
],
|
||||
"markers": "python_full_version >= '3.6.0'",
|
||||
"version": "==2.8.1"
|
||||
"markers": "python_version >= '3.6'",
|
||||
"version": "==2.8.2"
|
||||
},
|
||||
"google-api-python-client": {
|
||||
"hashes": [
|
||||
|
@ -351,7 +351,7 @@
|
|||
"sha256:023eaea9d8c1cceccd9587c6af6c20f33eeeb05d4148670f2b0322dc1511700c",
|
||||
"sha256:b09b56f5463070c2153753ef123f07d2e49235e89148e9b2459ec8ed2f68d7d3"
|
||||
],
|
||||
"markers": "python_full_version >= '3.6.0'",
|
||||
"markers": "python_version >= '3.6'",
|
||||
"version": "==1.56.2"
|
||||
},
|
||||
"gspread": {
|
||||
|
@ -367,7 +367,7 @@
|
|||
"sha256:70813c1135087a248a4d38cc0e1a0181ffab2188141a93eaf567940c3957ff06",
|
||||
"sha256:8ddd78563b633ca55346c8cd41ec0af27d3c79931828beffb46ce70a379e7442"
|
||||
],
|
||||
"markers": "python_full_version >= '3.6.0'",
|
||||
"markers": "python_version >= '3.6'",
|
||||
"version": "==0.13.0"
|
||||
},
|
||||
"httplib2": {
|
||||
|
@ -412,11 +412,11 @@
|
|||
},
|
||||
"jmespath": {
|
||||
"hashes": [
|
||||
"sha256:a490e280edd1f57d6de88636992d05b71e97d69a26a19f058ecf7d304474bf5e",
|
||||
"sha256:e8dcd576ed616f14ec02eed0005c85973b5890083313860136657e24784e4c04"
|
||||
"sha256:02e2e4cc71b5bcab88332eebf907519190dd9e6e82107fa7f83b1003a6252980",
|
||||
"sha256:90261b206d6defd58fdd5e85f478bf633a2901798906be2ad389150c5c60edbe"
|
||||
],
|
||||
"markers": "python_version >= '3.7'",
|
||||
"version": "==1.0.0"
|
||||
"version": "==1.0.1"
|
||||
},
|
||||
"loguru": {
|
||||
"hashes": [
|
||||
|
@ -562,7 +562,7 @@
|
|||
"sha256:23a8208d75b902797ea29fd31fa80a15ed9dc2c6c16fe73f5d346f83f6fa27a2",
|
||||
"sha256:6db33440354787f9b7f3a6dbd4febf5d0f93758354060e802f6c06cb493022fe"
|
||||
],
|
||||
"markers": "python_full_version >= '3.6.0'",
|
||||
"markers": "python_version >= '3.6'",
|
||||
"version": "==3.2.0"
|
||||
},
|
||||
"outcome": {
|
||||
|
@ -690,7 +690,7 @@
|
|||
"sha256:5eb116118f9612ff1ee89ac96437bb6b49e8f04d8a13b514ba26f620208e26eb",
|
||||
"sha256:dc9c10fb40944260f6ed4c688ece0cd2048414940f1cea51b8b226318411c519"
|
||||
],
|
||||
"markers": "python_full_version >= '3.6.0'",
|
||||
"markers": "python_version >= '3.6'",
|
||||
"version": "==2.12.0"
|
||||
},
|
||||
"pyopenssl": {
|
||||
|
@ -724,6 +724,14 @@
|
|||
"markers": "python_version >= '2.7' and python_version not in '3.0, 3.1, 3.2, 3.3'",
|
||||
"version": "==2.8.2"
|
||||
},
|
||||
"python-dotenv": {
|
||||
"hashes": [
|
||||
"sha256:b7e3b04a59693c42c36f9ab1cc2acc46fa5df8c78e178fc33a8d4cd05c8d498f",
|
||||
"sha256:d92a187be61fe482e4fd675b6d52200e7be63a12b724abbf931a40ce4fa92938"
|
||||
],
|
||||
"markers": "python_version >= '3.5'",
|
||||
"version": "==0.20.0"
|
||||
},
|
||||
"python-slugify": {
|
||||
"hashes": [
|
||||
"sha256:272d106cb31ab99b3496ba085e3fea0e9e76dcde967b5e9992500d1f785ce4e1",
|
||||
|
@ -863,11 +871,13 @@
|
|||
"sha256:fbc88d3ba402b5d041d204ec2449c4078898f89c4a6e6f0ed1c1a510ef1e221d",
|
||||
"sha256:fbd3fe37353c62fd0eb19fb76f78aa693716262bcd5f9c14bb9e5aca4b3f0dc4"
|
||||
],
|
||||
"markers": "python_full_version >= '3.6.0'",
|
||||
"markers": "python_version >= '3.6'",
|
||||
"version": "==2022.3.2"
|
||||
},
|
||||
"requests": {
|
||||
"extras": [],
|
||||
"extras": [
|
||||
"socks"
|
||||
],
|
||||
"hashes": [
|
||||
"sha256:bc7861137fbce630f17b03d3ad02ad0bf978c844f3536d0edda6499dafce2b6f",
|
||||
"sha256:d568723a7ebd25875d8d1eaf5dfa068cd2fc8194b2e483d7b1f7c81918dbec6b"
|
||||
|
@ -903,7 +913,7 @@
|
|||
"sha256:5c6bd9dc7a543b7fe4304a631f8a8a3b674e2bbfc49c2ae96200cdbe55df6b17",
|
||||
"sha256:95c5d300c4e879ee69708c428ba566c59478fd653cc3a22243eeb8ed846950bb"
|
||||
],
|
||||
"markers": "python_version < '4' and python_full_version >= '3.6.0'",
|
||||
"markers": "python_version >= '3.6' and python_version < '4'",
|
||||
"version": "==4.8"
|
||||
},
|
||||
"s3transfer": {
|
||||
|
@ -957,7 +967,7 @@
|
|||
"sha256:3b2503d3c7084a42b1ebd08116e5f81aadfaea95863628c80a3b774a11b7c759",
|
||||
"sha256:fc53893b3da2c33de295667a0e19f078c14bf86544af307354de5fcf12a3f30d"
|
||||
],
|
||||
"markers": "python_full_version >= '3.6.0'",
|
||||
"markers": "python_version >= '3.6'",
|
||||
"version": "==2.3.2.post1"
|
||||
},
|
||||
"telethon": {
|
||||
|
@ -1011,7 +1021,7 @@
|
|||
"sha256:238e70234214138ed7b4e8a0fab0e5e13872edab3be586ab8198c407620e2ab9",
|
||||
"sha256:8b536a8ec63dc0751342b3984193a3118f8fca2afe25752bb9b7fffd398552d3"
|
||||
],
|
||||
"markers": "python_full_version >= '3.6.0'",
|
||||
"markers": "python_version >= '3.6'",
|
||||
"version": "==2022.1"
|
||||
},
|
||||
"tzlocal": {
|
||||
|
@ -1019,7 +1029,7 @@
|
|||
"sha256:89885494684c929d9191c57aa27502afc87a579be5cdd3225c77c463ea043745",
|
||||
"sha256:ee5842fa3a795f023514ac2d801c4a81d1743bbe642e3940143326b3a00addd7"
|
||||
],
|
||||
"markers": "python_full_version >= '3.6.0'",
|
||||
"markers": "python_version >= '3.6'",
|
||||
"version": "==4.2"
|
||||
},
|
||||
"uritemplate": {
|
||||
|
@ -1027,7 +1037,7 @@
|
|||
"sha256:4346edfc5c3b79f694bccd6d6099a322bbeb628dbf2cd86eea55a456ce5124f0",
|
||||
"sha256:830c08b8d99bdd312ea4ead05994a38e8936266f84b9a7878232db50b044e02e"
|
||||
],
|
||||
"markers": "python_full_version >= '3.6.0'",
|
||||
"markers": "python_version >= '3.6'",
|
||||
"version": "==4.1.1"
|
||||
},
|
||||
"urllib3": {
|
||||
|
@ -1047,9 +1057,16 @@
|
|||
"sha256:11c731e214ebc7fa911db81efb021f97587493a5402b992f24748fe1cd9d7afc",
|
||||
"sha256:d0ae766fa93a40d47c5da045d94201721bf766dbde122a1d2253516b35c5edf3"
|
||||
],
|
||||
"index": "pypi",
|
||||
"version": "==11.9.8"
|
||||
},
|
||||
"vk-url-scraper": {
|
||||
"hashes": [
|
||||
"sha256:de74b161e8bae153160e1a6f0521457cb38a02a91e1dc598a41aef236d966b70",
|
||||
"sha256:fcd2ec01ac217f5257252cd83b1fccb2af59121c757b60488ba0e125d8bac7f1"
|
||||
],
|
||||
"index": "pypi",
|
||||
"version": "==0.1.5"
|
||||
},
|
||||
"websockets": {
|
||||
"hashes": [
|
||||
"sha256:07cdc0a5b2549bcfbadb585ad8471ebdc7bdf91e32e34ae3889001c1c106a6af",
|
||||
|
|
|
@ -1,8 +1,8 @@
|
|||
import re, json, requests
|
||||
import re, json
|
||||
|
||||
import vk_api, dateparser
|
||||
from bs4 import BeautifulSoup
|
||||
from loguru import logger
|
||||
from utils.misc import DateTimeEncoder
|
||||
from vk_url_scraper import VkScraper
|
||||
|
||||
from storages import Storage
|
||||
from .base_archiver import Archiver, ArchiveResult
|
||||
|
@ -17,73 +17,40 @@ class VkArchiver(Archiver):
|
|||
name = "vk"
|
||||
wall_pattern = re.compile(r"(wall.{0,1}\d+_\d+)")
|
||||
photo_pattern = re.compile(r"(photo.{0,1}\d+_\d+)")
|
||||
onclick_pattern = re.compile(r"({.*})")
|
||||
|
||||
def __init__(self, storage: Storage, driver, config: VkConfig):
|
||||
super().__init__(storage, driver)
|
||||
if config != None:
|
||||
self.vk_session = vk_api.VkApi(config.username, config.password)
|
||||
self.vk_session.auth(token_only=True)
|
||||
self.vks = VkScraper(config.username, config.password)
|
||||
|
||||
def download(self, url, check_if_exists=False):
|
||||
# detect URLs that this archiver can handle
|
||||
_id, method = None, None
|
||||
if has_wall := self.wall_pattern.search(url):
|
||||
_id = has_wall[0]
|
||||
method = self.archive_wall
|
||||
elif has_photo := self.photo_pattern.search(url):
|
||||
_id = has_photo[0]
|
||||
method = self.archive_photo
|
||||
else: return False
|
||||
if not hasattr(self, "vks") or self.vks is None:
|
||||
logger.debug("VK archiver was not supplied with credentials.")
|
||||
return False
|
||||
|
||||
logger.info(f"found valid {_id=} from {url=}")
|
||||
proper_url = f'https://vk.com/{_id}'
|
||||
|
||||
# if check if exists will not download again
|
||||
key = self.get_html_key(proper_url)
|
||||
key = self.get_html_key(url)
|
||||
if check_if_exists and self.storage.exists(key):
|
||||
screenshot = self.get_screenshot(proper_url)
|
||||
screenshot = self.get_screenshot(url)
|
||||
cdn_url = self.storage.get_cdn_url(key)
|
||||
return ArchiveResult(status="already archived", cdn_url=cdn_url, screenshot=screenshot)
|
||||
|
||||
try:
|
||||
return method(proper_url, _id)
|
||||
except Exception as e:
|
||||
logger.error(f"something went wrong with vk archive, possibly 404 causing index out of range, or missing key: {e}")
|
||||
return False
|
||||
results = self.vks.scrape(url) # some urls can contain multiple wall/photo/... parts and all will be fetched
|
||||
if len(results) == 0:
|
||||
return False
|
||||
|
||||
def archive_photo(self, photo_url, photo_id):
|
||||
headers = {"access_token": self.vk_session.token["access_token"], "photos": photo_id.replace("photo", ""), "extended": "1", "v": self.vk_session.api_version}
|
||||
req = requests.get("https://api.vk.com/method/photos.getById", headers)
|
||||
res = req.json()["response"][0]
|
||||
title = res["text"][:200] # more on the page
|
||||
img_url = res["orig_photo"]["url"]
|
||||
time = dateparser.parse(str(res["date"]), settings={"RETURN_AS_TIMEZONE_AWARE": True, "TO_TIMEZONE": "UTC"})
|
||||
|
||||
page_cdn, page_hash, thumbnail = self.generate_media_page([img_url], photo_url, res)
|
||||
screenshot = self.get_screenshot(photo_url)
|
||||
return ArchiveResult(status="success", cdn_url=page_cdn, screenshot=screenshot, hash=page_hash, thumbnail=thumbnail, timestamp=time, title=title)
|
||||
|
||||
def archive_wall(self, wall_url, wall_id):
|
||||
headers = {"access_token": self.vk_session.token["access_token"], "posts": wall_id.replace("wall", ""), "extended": "1", "copy_history_depth": "2", "v": self.vk_session.api_version}
|
||||
req = requests.get("https://api.vk.com/method/wall.getById", headers)
|
||||
res = req.json()["response"]
|
||||
wall = res["items"][0]
|
||||
img_urls = []
|
||||
if "attachments" in wall:
|
||||
for a in wall["attachments"]:
|
||||
attachment = a[a["type"]]
|
||||
if "thumb" in attachment:
|
||||
attachment = attachment["thumb"]
|
||||
if "sizes" in attachment:
|
||||
try: img_urls.append(attachment["sizes"][-1]["url"])
|
||||
except Exception as e:
|
||||
logger.warning(f"could not get image from attachment: {e}")
|
||||
|
||||
|
||||
title = wall["text"][:200] # more on the page
|
||||
time = dateparser.parse(str(wall["date"]), settings={"RETURN_AS_TIMEZONE_AWARE": True, "TO_TIMEZONE": "UTC"})
|
||||
|
||||
page_cdn, page_hash, thumbnail = self.generate_media_page(img_urls, wall_url, res)
|
||||
screenshot = self.get_screenshot(wall_url)
|
||||
dump_payload = lambda p : json.dumps(p, ensure_ascii=False, indent=4, cls=DateTimeEncoder)
|
||||
textual_output = ""
|
||||
title, time = results[0]["text"], results[0]["datetime"]
|
||||
urls_found = []
|
||||
for res in results:
|
||||
textual_output+= f"id: {res['id']}<br>time utc: {res['datetime']}<br>text: {res['text']}<br>payload: {dump_payload(res['payload'])}<br><hr/><br>"
|
||||
title = res["text"] if len(title) == 0 else title
|
||||
time = res["datetime"] if not time else time
|
||||
for attachments in res["attachments"].values():
|
||||
urls_found.extend(attachments)
|
||||
|
||||
page_cdn, page_hash, thumbnail = self.generate_media_page(urls_found, url, textual_output)
|
||||
# if multiple wall/photos/videos are present the screenshot will only grab the 1st
|
||||
screenshot = self.get_screenshot(url)
|
||||
return ArchiveResult(status="success", cdn_url=page_cdn, screenshot=screenshot, hash=page_hash, thumbnail=thumbnail, timestamp=time, title=title)
|
||||
|
|
|
@ -1,5 +1,6 @@
|
|||
|
||||
import os, sys, requests
|
||||
import os, json, requests
|
||||
from datetime import datetime
|
||||
from loguru import logger
|
||||
|
||||
|
||||
|
@ -19,10 +20,19 @@ def expand_url(url):
|
|||
logger.error(f'Failed to expand url {url}')
|
||||
return url
|
||||
|
||||
def getattr_or(o: object, prop: str, default = None):
|
||||
try:
|
||||
|
||||
def getattr_or(o: object, prop: str, default=None):
|
||||
try:
|
||||
res = getattr(o, prop)
|
||||
if res is None: raise
|
||||
return res
|
||||
except:
|
||||
return default
|
||||
return default
|
||||
|
||||
|
||||
class DateTimeEncoder(json.JSONEncoder):
|
||||
# to allow json.dump with datetimes do json.dumps(obj, cls=DateTimeEncoder)
|
||||
def default(self, o):
|
||||
if isinstance(o, datetime):
|
||||
return str(o) # with timezone
|
||||
return json.JSONEncoder.default(self, o)
|
||||
|
|
Ładowanie…
Reference in New Issue