Implement running Discord and 4chan gathering simultaneously
This commit is contained in:
216
sources/ch4.py
Normal file
216
sources/ch4.py
Normal file
@@ -0,0 +1,216 @@
|
||||
# Python modules can't start with a number...
|
||||
import json
|
||||
import random
|
||||
import string
|
||||
from datetime import datetime
|
||||
from typing import Any, Dict
|
||||
|
||||
import treq
|
||||
from bs4 import BeautifulSoup
|
||||
from siphashc import siphash
|
||||
from twisted.internet.defer import inlineCallbacks
|
||||
|
||||
import db
|
||||
import util
|
||||
|
||||
|
||||
class Chan4(object):
|
||||
"""
|
||||
4chan indexer, crawler and ingester.
|
||||
"""
|
||||
|
||||
def __init__(self):
|
||||
name = self.__class__.__name__
|
||||
self.log = util.get_logger(name)
|
||||
|
||||
self.api_endpoint = "https://a.4cdn.org"
|
||||
self.boards = []
|
||||
self.thread_list = {}
|
||||
|
||||
self.thread_deferreds = []
|
||||
|
||||
self.log.info(f"Starting crawler bot to {self.api_endpoint}")
|
||||
|
||||
self.hash_key = db.r.get("hashing_key")
|
||||
if not self.hash_key:
|
||||
letters = string.ascii_lowercase
|
||||
self.hash_key = "".join(random.choice(letters) for i in range(16))
|
||||
self.log.debug(f"Created new hash key: {self.hash_key}")
|
||||
db.r.set("hashing_key", self.hash_key)
|
||||
else:
|
||||
|
||||
self.hash_key = self.hash_key.decode("ascii")
|
||||
self.log.debug(f"Decoded hash key: {self.hash_key}")
|
||||
|
||||
@inlineCallbacks
|
||||
def run(self):
|
||||
yield self.get_board_list()
|
||||
yield self.get_thread_lists()
|
||||
yield self.get_thread_contents()
|
||||
|
||||
def get_board_list(self):
|
||||
self.log.info("Getting board list")
|
||||
response = self.api_call("boards.json")
|
||||
response.addCallback(self.got_board_list)
|
||||
return response
|
||||
|
||||
def got_board_list(self, board_list):
|
||||
if board_list["success"]:
|
||||
for board in board_list["response"]["boards"]:
|
||||
self.boards.append(board["board"])
|
||||
|
||||
@inlineCallbacks
|
||||
def get_thread_lists(self):
|
||||
for board in self.boards:
|
||||
yield self.get_thread_list(board)
|
||||
# self.thread_deferreds.append(d)
|
||||
# yield defer.gatherResults(self.thread_deferreds)
|
||||
# self.thread_deferreds = []
|
||||
# self.log.info("Finished getting thread lists")
|
||||
|
||||
@inlineCallbacks
|
||||
def get_thread_contents(self):
|
||||
for board in self.thread_list.keys():
|
||||
for page in self.thread_list[board]:
|
||||
for threads in page["threads"]:
|
||||
no = threads["no"]
|
||||
yield self.get_thread_content(board, no)
|
||||
# self.content_deferreds.append(d)
|
||||
# al = yield defer.gatherResults(self.content_deferreds)
|
||||
# self.content_deferreds = []
|
||||
# self.log.info("Finished getting content")
|
||||
|
||||
def get_thread_list(self, board):
|
||||
self.log.info(f"Getting thread list for {board}")
|
||||
response = self.api_call(f"{board}/catalog.json")
|
||||
response.addCallback(self.got_thread_list, board)
|
||||
return response
|
||||
|
||||
def got_thread_list(self, thread_list, board):
|
||||
if thread_list["success"]:
|
||||
self.thread_list[board] = thread_list["response"]
|
||||
self.log.info(f"Got thread list for {board}: {len(thread_list)}")
|
||||
|
||||
def get_thread_content(self, board, thread):
|
||||
self.log.info(f"Getting information for thread {thread} on board {board}")
|
||||
response = self.api_call(f"{board}/thread/{thread}.json")
|
||||
response.addCallback(self.got_thread_content, board, thread)
|
||||
return response
|
||||
|
||||
def got_thread_content(self, thread_content, board, thread):
|
||||
if thread_content["success"]:
|
||||
self.log.info(f"Got thread content for thread {thread} on board {board}")
|
||||
for post in thread_content["response"]["posts"]:
|
||||
# print(post)
|
||||
self.handle_post(board, thread, post)
|
||||
else:
|
||||
self.log.error(
|
||||
(
|
||||
f"Error fetching thread {thread} on board {board}: "
|
||||
f"{thread_content['message']}"
|
||||
)
|
||||
)
|
||||
|
||||
def handle_post(self, board, thread, post):
|
||||
name_map = {
|
||||
"no": "msg_id",
|
||||
"now": "ts",
|
||||
"name": "user",
|
||||
"trip": "nick",
|
||||
"id": "nick_id",
|
||||
"resto": "id_reply",
|
||||
"com": "msg",
|
||||
"ext": "file_ext",
|
||||
"w": "file_w",
|
||||
"h": "file_h",
|
||||
"tn_w": "file_tn_w",
|
||||
"tn_h": "file_tn_h",
|
||||
"tim": "file_tim",
|
||||
"fsize": "file_size",
|
||||
"md5": "file_md5",
|
||||
"filedeleted": "file_deleted",
|
||||
"spoiler": "file_spoiler",
|
||||
"custom_spoiler": "file_custom_spoiler",
|
||||
"m_img": "file_m_img",
|
||||
"time": "unix_time",
|
||||
}
|
||||
post["type"] = "msg"
|
||||
|
||||
# Calculate hash for post
|
||||
post_normalised = json.dumps(post, sort_keys=True)
|
||||
hash = siphash(self.hash_key, post_normalised)
|
||||
hash = str(hash)
|
||||
redis_key = f"cache.{board}.{thread}.{post['no']}"
|
||||
key_content = db.r.get(redis_key)
|
||||
if key_content:
|
||||
key_content = key_content.decode("ascii")
|
||||
if key_content == hash:
|
||||
return
|
||||
else:
|
||||
post["type"] = "update"
|
||||
db.r.set(redis_key, hash)
|
||||
# Check if hash exists
|
||||
# Store the hash
|
||||
for key, value in list(post.items()):
|
||||
if key in name_map:
|
||||
post[name_map[key]] = post[key]
|
||||
del post[key]
|
||||
if "ts" in post:
|
||||
old_time = post["ts"]
|
||||
# '08/30/22(Tue)02:25:37'
|
||||
time_spl = old_time.split(":")
|
||||
if len(time_spl) == 3:
|
||||
old_ts = datetime.strptime(old_time, "%m/%d/%y(%a)%H:%M:%S")
|
||||
else:
|
||||
old_ts = datetime.strptime(old_time, "%m/%d/%y(%a)%H:%M")
|
||||
new_ts = old_ts.isoformat()
|
||||
post["ts"] = new_ts
|
||||
if "msg" in post:
|
||||
soup = BeautifulSoup(post["msg"], "html.parser")
|
||||
msg = soup.get_text(separator="\n")
|
||||
post["msg"] = msg
|
||||
|
||||
post["src"] = "4ch"
|
||||
|
||||
# print({name_map[name]: val for name, val in post.items()})
|
||||
db.store_message(post)
|
||||
|
||||
@inlineCallbacks
|
||||
def callback_api_call(self, response, result):
|
||||
try:
|
||||
text = yield response.content()
|
||||
except: # noqa
|
||||
self.log.error("Error with API call")
|
||||
return
|
||||
try:
|
||||
result["response"] = json.loads(text)
|
||||
except json.decoder.JSONDecodeError:
|
||||
result["success"] = "ERROR"
|
||||
result["message"] = "Error parsing JSON."
|
||||
return result
|
||||
result["status"] = response.code
|
||||
if response.code == 200:
|
||||
result["success"] = True
|
||||
result["message"] = "OK"
|
||||
else:
|
||||
result["message"] = "API ERROR"
|
||||
|
||||
return result
|
||||
|
||||
def api_call(self, method: str):
|
||||
headers = {
|
||||
"User-Agent": (
|
||||
"Mozilla/5.0 (Windows NT 10.0; rv:68.0) Gecko/20100101 Firefox/68.0"
|
||||
)
|
||||
}
|
||||
url = f"{self.api_endpoint}/{method}"
|
||||
self.log.debug(f"GET {url}")
|
||||
response = treq.get(url, headers=headers)
|
||||
result: Dict[str, Any] = {
|
||||
"success": False,
|
||||
"message": "Invalid Method",
|
||||
"response": None,
|
||||
"status": None,
|
||||
}
|
||||
response.addCallback(self.callback_api_call, result)
|
||||
return response
|
||||
62
sources/dis.py
Normal file
62
sources/dis.py
Normal file
@@ -0,0 +1,62 @@
|
||||
#!/usr/bin/env python
|
||||
from operator import attrgetter
|
||||
|
||||
import discord
|
||||
|
||||
import db
|
||||
import util
|
||||
|
||||
ATTRMAP = {
|
||||
"msg": "content",
|
||||
"msg_id": "id",
|
||||
"nick": "author.name",
|
||||
"host": "author.discriminator",
|
||||
"ident": "author.nick",
|
||||
"time": "created_at",
|
||||
"channel": "channel.name",
|
||||
"channel_nsfw": "channel.nsfw",
|
||||
"bot": "author.bot",
|
||||
"user_id": "author.id",
|
||||
"channel_id": "channel.id",
|
||||
"net": "author.guild.name",
|
||||
"net_id": "author.guild.id",
|
||||
"guild_member_count": "author.guild.member_count",
|
||||
"channel_category": "channel.category.name",
|
||||
"channel_category_id": "channel.category.id",
|
||||
"channel_category_nsfw": "channel.category.nsfw",
|
||||
}
|
||||
|
||||
|
||||
class DiscordClient(discord.Client):
|
||||
def __init__(self, *args, **kwargs):
|
||||
self.logger = None
|
||||
self.did_something = False
|
||||
name = self.__class__.__name__
|
||||
self.log = util.get_logger(name)
|
||||
|
||||
super().__init__(*args, **kwargs)
|
||||
|
||||
async def on_ready(self):
|
||||
self.log.info("Discord connection established.")
|
||||
|
||||
def recurse_dict(self, obj):
|
||||
to_return = {}
|
||||
for key, mapped in ATTRMAP.items():
|
||||
try:
|
||||
to_return[key] = attrgetter(mapped)(obj)
|
||||
except AttributeError:
|
||||
continue
|
||||
|
||||
return to_return
|
||||
|
||||
async def on_message(self, message):
|
||||
if not message.content:
|
||||
return
|
||||
|
||||
a = self.recurse_dict(message)
|
||||
a["ts"] = a["time"].isoformat()
|
||||
del a["time"]
|
||||
a["type"] = "msg"
|
||||
a["src"] = "dis"
|
||||
|
||||
db.store_message(a)
|
||||
Reference in New Issue
Block a user