ioistired-pleroma-ebooks/fetch_posts.py

260 lines
8.6 KiB
Python
Raw Normal View History

2018-10-09 03:11:51 +02:00
#!/usr/bin/env python3
# SPDX-License-Identifier: AGPL-3.0-only
2018-10-09 03:11:51 +02:00
2021-06-16 03:59:57 +02:00
import sys
import anyio
import aiohttp
import platform
import pendulum
import operator
import aiosqlite
2021-06-16 03:59:57 +02:00
import contextlib
2021-09-21 15:58:20 +02:00
from yarl import URL
2023-01-11 07:23:42 +01:00
from pleroma import Pleroma, HandleRateLimits
from bs4 import BeautifulSoup
from functools import partial
from typing import Iterable, NewType
2023-01-11 07:23:42 +01:00
from utils import shield, suppress, http_session_factory
from third_party.utils import extract_post_content
UTC = pendulum.timezone('UTC')
JSON_CONTENT_TYPE = 'application/json'
ACTIVITYPUB_CONTENT_TYPE = 'application/activity+json'
2021-09-17 08:34:44 +02:00
MIGRATION_VERSION = 1
class PostFetcher:
def __init__(self, *, config):
self.config = config
self.erroneous_accounts = []
async def __aenter__(self):
stack = contextlib.AsyncExitStack()
self._fedi = await stack.enter_async_context(
Pleroma(api_base_url=self.config['site'], access_token=self.config['access_token']),
)
self._http = await stack.enter_async_context(
2023-01-11 07:23:42 +01:00
http_session_factory(
headers={'Accept': ', '.join([JSON_CONTENT_TYPE, ACTIVITYPUB_CONTENT_TYPE])},
trust_env=True,
raise_for_status=True,
),
)
2021-09-17 08:35:54 +02:00
self._rl_handler = HandleRateLimits(self._http)
2021-07-26 08:47:43 +02:00
self._db = await stack.enter_async_context(aiosqlite.connect(self.config['db_path']))
2021-09-17 08:34:44 +02:00
await self._maybe_run_migrations()
await self._db.commit()
self._db.row_factory = aiosqlite.Row
self._ctx_stack = stack
return self
2021-09-17 08:34:44 +02:00
async def _maybe_run_migrations(self):
async with self._db.cursor() as cur, suppress(aiosqlite.OperationalError):
if await (await cur.execute('SELECT migration_version FROM migrations')).fetchone(): return
await self._run_migrations()
async def _run_migrations(self):
# TODO proper migrations, not just "has the schema ever been run" migrations
async with await (anyio.Path(__file__).parent/'schema.sql').open() as f:
schema = await f.read()
async with self._db.cursor() as cur:
await cur.executescript(schema)
await cur.execute('INSERT INTO migrations (migration_version) VALUES (?)', (MIGRATION_VERSION,))
async def __aexit__(self, *excinfo):
return await self._ctx_stack.__aexit__(*excinfo)
# username@instance
AccountHandle = NewType('AccountHandle', str)
async def fetch_all(self):
"""fetch all following accounts, or an iterable of accounts if provided"""
await self._fedi.verify_credentials()
self._completed_accounts = {}
async with anyio.create_task_group() as tg:
2021-09-21 15:58:20 +02:00
for fqn in map(self.fqn, await self._fedi.following()):
tg.start_soon(self._do_account, fqn)
def fqn(self, acc: dict):
try:
return acc['fqn']
except KeyError:
fqn = acc['acct']
if '@' in fqn: return fqn
return fqn + '@' + URL(self.config['site']).host
2021-06-16 03:59:57 +02:00
async def _do_account(self, acc: AccountHandle):
2021-06-16 03:59:57 +02:00
async with anyio.create_task_group() as tg:
self._completed_accounts[acc] = done_ev = anyio.Event()
tx, rx = anyio.create_memory_object_stream()
async with rx, tx:
tg.start_soon(self._process_pages, rx, acc)
tg.start_soon(self._fetch_account, tx, acc)
await done_ev.wait()
# processing is complete, so halt fetching.
# processing may complete before fetching if we get caught up on new posts.
tg.cancel_scope.cancel()
async def _process_pages(self, stream, account):
done_ev = self._completed_accounts[account]
2021-06-16 03:59:57 +02:00
try:
async for activity in stream:
try:
await self._insert_activity(activity)
except aiosqlite.IntegrityError as exc:
# LOL sqlite error handling is so bad
if exc.args[0].startswith('UNIQUE constraint failed: '):
# this means we've encountered an item we already have saved
2021-09-17 08:37:39 +02:00
# TODO we need to ignore this if we don't actually have all the posts.
# For example, if a prior fetch was interrupted, we'll have k pages of the most recent posts,
# but no more. But since we still have the most recent page saved, it'll *look* like
# we've saved everything, since we stop as soon as we encounter a post we already have.
# To fix this we can check against totalItems in the user's outbox.
break
self.erroneous_accounts.append(account)
raise
finally:
print('Saving posts from', account, 'to the DB')
await self._db.commit()
done_ev.set()
async def _insert_activity(self, activity):
if activity['type'] != 'Create':
# this isn't a post but something else (like, boost, reaction, etc)
return
obj = activity['object']
await self._db.execute(
"""
INSERT INTO posts (post_id, summary, content, published_at)
VALUES (?, ?, ?, ?)
""",
(
obj['id'],
# Pleroma returns an empty string here for posts without a CW,
# which is semantically incorrect IMO
obj['summary'] or None,
extract_post_content(obj['content']),
pendulum.parse(obj['published']).astimezone(pendulum.timezone('UTC')).timestamp(),
),
)
# TODO figure out why i put shield here lol
@shield
async def _fetch_account(self, tx, account: AccountHandle):
done_ev = self._completed_accounts[account]
2021-06-16 03:59:57 +02:00
try:
outbox = await self.fetch_outbox(account)
except Exception as exc:
2021-06-16 03:59:57 +02:00
import traceback
traceback.print_exception(type(exc), exc, exc.__traceback__)
done_ev.set()
self.erroneous_accounts.append(account)
return
print(f'Fetching posts for {account}...')
next_page_url = outbox['first']
while True:
print(f'Fetching {next_page_url}... ')
2021-09-17 08:35:54 +02:00
async with self._rl_handler.request('GET', next_page_url) as resp: page = await resp.json()
for activity in page['orderedItems']:
try:
await tx.send(activity)
except anyio.BrokenResourceError:
# already closed means we're already done
return
# show progress
#print('.', end='', flush=True)
if not (next_page_url := page.get('next')):
#done_ev.set()
break
done_ev.set()
async def fetch_outbox(self, handle):
"""finger handle, a fully-qualified ActivityPub actor name, returning their outbox URL"""
# it's fucking incredible how overengineered ActivityPub is btw
print('Fingering ', handle, '...', sep='')
username, at, instance = handle.lstrip('@').partition('@')
assert at == '@'
# i was planning on doing /.well-known/host-meta to find the webfinger URL, but
# 1) honk does not support host-meta
# 2) WebFinger is always located at the same location anyway
profile_url = await self._finger_actor(username, instance)
try:
async with self._http.get(profile_url) as resp: profile = await resp.json()
except aiohttp.ContentTypeError:
# we didn't get JSON, so just guess the outbox URL
outbox_url = profile_url + '/outbox'
else:
outbox_url = profile['outbox']
async with self._http.get(outbox_url) as resp: outbox = await resp.json()
assert outbox['type'] == 'OrderedCollection'
return outbox
async def _finger_actor(self, username, instance):
# despite HTTP being a direct violation of the WebFinger spec, assume e.g. Tor instances do not support
# HTTPS-over-onion
finger_url = f'http://{instance}/.well-known/webfinger?resource=acct:{username}@{instance}'
async with self._http.get(finger_url) as resp: finger_result = await resp.json()
return (profile_url := self._parse_webfinger_result(username, instance, finger_result))
def _parse_webfinger_result(self, username, instance, finger_result):
"""given webfinger data, return profile URL for handle"""
def check_content_type(type, ct): return ct == type or ct.startswith(type+';')
check_ap = partial(check_content_type, ACTIVITYPUB_CONTENT_TYPE)
try:
# note: the server might decide to return multiple links
# so we need to decide how to prefer one.
2021-07-26 08:04:14 +02:00
# i'd put "and yarl.URL(template).host == instance" here,
# but some instances have no subdomain for the handle yet use a subdomain for the canonical URL.
# Additionally, an instance could theoretically serve profile pages over I2P and the clearnet,
# for example.
return (profile_url := next(
link['href']
for link in finger_result['links']
if link['rel'] == 'self' and check_ap(link['type'])
))
except StopIteration:
# this should never happen either
raise RuntimeError(f'fatal: while fingering {username}@{instance}, failed to find a profile URL')
async def amain():
import json5 as json
2021-07-26 08:04:32 +02:00
import third_party.utils as utils
args = utils.arg_parser_factory(description='Fetch posts from all followed accounts').parse_args()
config = utils.load_config(args.cfg)
async with PostFetcher(config=config) as fetcher: await fetcher.fetch_all()
if (accs := fetcher.erroneous_accounts):
print(
'Exiting unsuccessfully due to previous errors in these accounts:',
', '.join(accs),
file=sys.stderr,
)
sys.exit(1)
def main():
try:
anyio.run(amain)
except KeyboardInterrupt:
# suppress the lengthy traceback
sys.exit(1)
2018-10-09 03:11:51 +02:00
2021-06-16 03:59:57 +02:00
if __name__ == '__main__':
main()