"""Utilities for ActivityStreams 1 objects.
* https://activitystrea.ms/specs/json/schema/activity-schema.html
* http://activitystrea.ms/specs/json/1.0/
"""
import collections
import copy
import logging
from operator import itemgetter
import re
from oauth_dropins.webutil import util
logger = logging.getLogger(__name__)
CONTENT_TYPE = 'application/stream+json'
# maps each AS1 RSVP verb to the collection inside an event object that the
# RSVPing actor would go into.
RSVP_VERB_TO_COLLECTION = collections.OrderedDict(( # in priority order
('rsvp-yes', 'attending'),
('rsvp-no', 'notAttending'),
('rsvp-maybe', 'maybeAttending'),
('rsvp-interested', 'interested'),
('invite', 'invited'),
))
VERBS_WITH_OBJECT = frozenset((
'accept',
'block',
'flag',
'follow',
'like',
'react',
'reject',
'share',
'stop-following',
'undo',
)) | set(RSVP_VERB_TO_COLLECTION.keys())
CRUD_VERBS = frozenset((
'delete',
'post',
'update',
# not in AS1 spec, undo isn't a real AS1 verb, but we map it to AS2 Undo
# https://activitystrea.ms/specs/json/schema/activity-schema.html#verbs
'undo',
))
# objectTypes that can be actors. technically this is based on AS2 semantics,
# since it's unspecified in AS1.
# https://www.w3.org/TR/activitystreams-core/#actors
# https://activitystrea.ms/specs/json/1.0/#activity
ACTOR_TYPES = frozenset((
'application',
'group',
'organization',
'person',
'service',
))
# this isn't well defined 🤷
POST_TYPES = frozenset((
'article',
'comment',
'link',
'mention',
'note',
))
# used in original_post_discovery
_PERMASHORTCITATION_RE = re.compile(r'\(([^:\s)]+\.[^\s)]{2,})[ /]([^\s)]+)\)$')
HASHTAG_RE = re.compile(r'(^|\s)[#＃](\w+)\b', re.UNICODE)
[docs]
def object_type(obj):
"""Returns the object type, or the verb if it's an activity object.
Details: http://activitystrea.ms/specs/json/1.0/#activity-object
Args:
obj (dict): decoded JSON ActivityStreams object
Returns:
str: ActivityStreams object type
"""
type = obj.get('objectType')
return type if type and type != 'activity' else obj.get('verb')
[docs]
def get_object(obj, field='object'):
"""Extracts and returns a field value as an object.
If the field value is a string, returns an object with it as the id, eg
``{'id': val}``. If the field value is a list, returns the first element.
Args:
obj (dict): decoded JSON ActivityStreams object
field (str)
Returns:
dict:
"""
if not obj:
return {}
val = util.get_first(obj, field, {}) or {}
return {'id': val} if isinstance(val, str) else val
[docs]
def get_objects(obj, field='object'):
"""Extracts and returns a field's values as objects.
If a field value is a string, generates an object with it as the id, eg
``{'id': val}``.
Args:
obj (dict): decoded JSON ActivityStreams object
field (str)
Returns:
sequence of dict:
"""
if not obj:
return []
return [{'id': val} if isinstance(val, str) else val
for val in util.get_list(obj, field)]
[docs]
def get_owner(obj):
"""Returns an object's author or actor.
Prefers author, then actor. If that's an object, returns its id, if available.
If neither field exists, but ``obj.objectType`` is in ``ACTOR_TYPES``,
returns its own id, if available.
For posts, updates, and deletes, falls back to the inner object's owner if the
outer activity has no actor.
Args:
obj (dict): decoded JSON ActivityStreams object
Returns:
str:
"""
if not obj:
return None
elif not isinstance(obj, dict):
raise ValueError(f'{obj} is not a dict')
ids = get_ids(obj, 'author') or get_ids(obj, 'actor')
if ids:
return ids[0]
if obj.get('objectType') in ACTOR_TYPES:
ids = util.get_list(obj, 'id')
if ids:
return ids[0]
if obj.get('verb') in ('post', 'update', 'delete'):
return get_owner(get_object(obj))
[docs]
def get_url(obj):
"""Returns the url field's first text value, or ``''``.
Somewhat duplicates :func:`microformats2.get_text`.
"""
if not obj:
return ''
urls = object_urls(obj)
if not urls:
return ''
return (urls[0] if isinstance(urls, (list, tuple))
else urls if isinstance(urls, str)
else '')
[docs]
def get_ids(obj, field):
"""Extracts and returns a given field's values as ids.
Args:
obj (dict): decoded JSON ActivityStreams object
field (str)
Returns string values as is. For dict values, returns their inner ``id`` or
``url`` field value, in that order of precedence.
"""
ids = set()
for elem in util.get_list(obj, field):
if elem and isinstance(elem, str):
ids.add(elem)
elif isinstance(elem, dict):
id = elem.get('id') or elem.get('url')
if id:
ids.add(id)
return sorted(ids)
[docs]
def merge_by_id(obj, field, new):
"""Merges new items by id into a field in an existing AS1 object, in place.
Merges new items by id into the given field. If it exists, it must be a list.
Requires all existing and new items in the field to have ids.
Args:
obj (dict): AS1 object
field (str): name of field to merge new items into
new (sequence of dict): new values to merge in
"""
merged = {o['id']: o for o in obj.get(field, []) + new}
obj[field] = sorted(merged.values(), key=itemgetter('id'))
[docs]
def is_public(obj, unlisted=True):
"""Returns True if the object is public, False if private, None if unknown.
...according to the Audience Targeting extension:
http://activitystrea.ms/specs/json/targeting/1.0/
Expects values generated by this library: ``objectType`` ``group``, ``alias``
``@public``, ``@unlisted``, or ``@private``.
Also, important point: this defaults to True, ie public. Bridgy depends on
that and prunes the to field from stored activities in Response objects (in
``bridgy/util.prune_activity``). If the default here ever changes, be sure to
update Bridgy's code.
Args:
obj (dict): AS1 activity or object
unlisted (bool): whether `@unlisted` counts as public or not
Returns:
bool:
"""
if obj is None:
return None
inner_obj = get_object(obj)
to = get_objects(obj, 'to') or get_objects(inner_obj, 'to') or []
aliases = util.trim_nulls([t.get('alias') for t in to])
object_types = util.trim_nulls([t.get('objectType') for t in to])
if '@public' in aliases or ('@unlisted' in aliases and unlisted):
return True
elif 'unknown' in object_types:
return None
elif aliases:
return False
elif 'to' in obj or 'to' in inner_obj:
# it does at least have some audience that doesn't include public
return False
return True
[docs]
def recipient_if_dm(obj, actor=None):
"""If ``obj`` is a DM, returns the recipient actor's id.
DMs are ``note``s addressed to a single recipient, ie ``to`` has one value and
``cc``, ``bcc``, and ``bto`` have none.
If ``obj`` isn't a DM, returns None.
Those fields are based on the Audience Targeting extension:
http://activitystrea.ms/specs/json/targeting/1.0/
Args:
obj (dict): AS1 activity or object
actor (dict): optional AS1 actor who sent this object. Its followers
collection is used to identify followers-only posts.
Returns:
bool:
"""
if not obj or is_public(obj):
return None
if object_type(obj) == 'post':
obj = get_object(obj)
if not obj or object_type(obj) not in (None, 'note'):
return None
tos = util.get_list(obj, 'to')
others = (util.get_list(obj, 'cc')
+ util.get_list(obj, 'bto')
+ util.get_list(obj, 'bcc'))
if not (len(tos) == 1 and len(others) == 0):
return None
follow_collections = []
for a in actor, get_object(obj, 'author'):
if a:
follow_collections.extend([a.get('followers'), a.get('following')])
to = tos.pop()
if isinstance(to, dict):
to = to.get('id') or ''
to_lower = to.lower()
if to and not is_audience(to) and to not in follow_collections:
return to
[docs]
def is_dm(obj, actor=None):
"""Returns True if the object is a DM, ie addressed to a single recipient.
See :func:`recipient_if_dm` for details.
"""
return bool(recipient_if_dm(obj, actor=actor))
[docs]
def is_audience(val):
"""Returns True if val is a "special" AS1 or AS2 audience, eg "public."
See the AS1 Audience Targeting extension and AS2 spec:
* http://activitystrea.ms/specs/json/targeting/1.0/
* https://www.w3.org/TR/activitystreams-vocabulary/#dfn-audience
"""
if not val or not isinstance(val, str):
return False
val = val.lower()
return (# https://activitystrea.ms/specs/json/targeting/1.0/
val in ('public', 'unlisted', 'private')
or val.startswith('https://www.w3.org/')
or val.startswith('https://w3.org/')
or val.startswith('@') # AS1 audience targeting alias, eg @public, @unlisted
or val.startswith('as:')
# as2 public constant is https://www.w3.org/ns/activitystreams#Public
or val.endswith('#public')
# non-standared heuristic for Mastodon and similar followers/following
# collections
or val.endswith('/followers')
or val.endswith('/following'))
[docs]
def add_rsvps_to_event(event, rsvps):
"""Adds RSVP objects to an event's attending fields, in place.
Args:
event (dict): ActivityStreams event object
rsvps (sequence of dict): ActivityStreams RSVP activity objects
"""
for rsvp in rsvps:
field = RSVP_VERB_TO_COLLECTION.get(rsvp.get('verb'))
if field:
event.setdefault(field, []).append(rsvp.get(
'object' if field == 'invited' else 'actor'))
[docs]
def get_rsvps_from_event(event):
"""Returns RSVP objects for an event's attending fields.
Args:
event (dict): ActivityStreams event object
Returns:
sequence of dict: ActivityStreams RSVP activity objects
"""
id = event.get('id')
if not id:
return []
parsed = util.parse_tag_uri(id)
if not parsed:
return []
domain, event_id = parsed
url = event.get('url')
author = event.get('author')
rsvps = []
for verb, field in RSVP_VERB_TO_COLLECTION.items():
for actor in event.get(field, []):
rsvp = {'objectType': 'activity',
'verb': verb,
'object' if verb == 'invite' else 'actor': actor,
'url': url,
}
if event_id and 'id' in actor:
_, actor_id = util.parse_tag_uri(actor['id'])
rsvp['id'] = util.tag_uri(domain, f'{event_id}_rsvp_{actor_id}')
if url:
rsvp['url'] = '#'.join((url, actor_id))
if verb == 'invite' and author:
rsvp['actor'] = author
rsvps.append(rsvp)
return rsvps
[docs]
def activity_changed(before, after, inReplyTo=True, log=False):
"""Returns whether two activities or objects differ meaningfully.
Only compares a few fields: ``objectType``, ``verb``, ``displayName`,
``content``, ``summary``, ``location``, and ``image``. Notably does *not*
compare ``author``, ``published``, or ``updated``.
Args:
before (dict): ActivityStreams activity or object
after (dict): ActivityStreams activity or object
inReplyTo (bool): whether to return True if ``inReplyTo`` has changed
log (bool): whether to log each changed field
Returns:
bool:
"""
def changed(b, a, field, label, ignore=None):
b_val = b.get(field)
a_val = a.get(field)
if ignore and isinstance(b_val, dict) and isinstance(a_val, dict):
b_val = copy.copy(b_val)
a_val = copy.copy(a_val)
for field in ignore:
b_val.pop(field, None)
a_val.pop(field, None)
if b_val != a_val and (a_val or b_val):
if log:
logger.debug(f'{label}[{field}] {b_val} => {a_val}')
return True
obj_b = get_object(before)
obj_a = get_object(after)
if any(changed(before, after, field, 'activity') or
changed(obj_b, obj_a, field, 'activity[object]')
for field in ('objectType', 'verb', 'to', 'displayName', 'content',
'summary', 'location', 'image')):
return True
if (inReplyTo and
(changed(before, after, 'inReplyTo', 'inReplyTo', ignore=('author',)) or
changed(obj_b, obj_a, 'inReplyTo', 'object.inReplyTo', ignore=('author',)))):
return True
return False
[docs]
def append_in_reply_to(before, after):
"""Appends before's ``inReplyTo`` to ``after``, in place.
Args:
before (dict): ActivityStreams activity or object
after (dict): ActivityStreams activity or object
"""
obj_b = get_object(before) or before
obj_a = get_object(after) or after
if obj_b and obj_a:
reply_b = util.get_list(obj_b, 'inReplyTo')
reply_a = util.get_list(obj_a, 'inReplyTo')
obj_a['inReplyTo'] = util.dedupe_urls(reply_a + reply_b)
[docs]
def actor_name(actor):
"""Returns the given actor's name if available, otherwise Unknown."""
if actor:
return actor.get('displayName') or actor.get('username') or 'Unknown'
return 'Unknown'
[docs]
def original_post_discovery(
activity, domains=None, include_redirect_sources=True,
include_reserved_hosts=True, max_redirect_fetches=None, **kwargs):
"""Discovers original post links.
This is a variation on http://indiewebcamp.com/original-post-discovery . It
differs in that it finds multiple candidate links instead of one, and it
doesn't bother looking for MF2 (etc) markup because the silos don't let you
input it. More background:
https://github.com/snarfed/bridgy/issues/51#issuecomment-136018857
Original post candidates come from the ``upstreamDuplicates``,
``attachments``, and ``tags`` fields, as well as links and
permashortlinks/permashortcitations in the text content.
Args:
activity (dict): activity
domains (sequence of str) domains, optional. If provided, only links to
these domains will be considered original and stored in
``upstreamDuplicates``. (Permashortcitations are exempt.)
include_redirect_sources (bool): whether to include URLs that redirect
as well as their final destination URLs
include_reserved_hosts (bool): whether to include domains on reserved
TLDs (eg foo.example) and local hosts (eg http://foo.local/,
http://my-server/)
max_redirect_fetches (int): if specified, only make up to this many HTTP
fetches to resolve redirects.
kwargs: passed to :func:`requests.head` when following redirects
Returns:
(list of str, list of str) tuple: (original post URLs, mentions)
"""
obj = get_object(activity) or activity
content = obj.get('content', '').strip()
# find all candidate URLs
tags = [t.get('url') for t in obj.get('attachments', []) + obj.get('tags', [])
if t.get('objectType') in ('article', 'link', 'mention', 'note', None)]
candidates = (tags + util.extract_links(content) +
obj.get('upstreamDuplicates', []) +
util.get_list(obj, 'targetUrl'))
# Permashortcitations (http://indiewebcamp.com/permashortcitation) are short
# references to canonical copies of a given (usually syndicated) post, of
# the form (DOMAIN PATH). We consider them an explicit original post link.
candidates += [match.expand(r'http://\1/\2') for match in
_PERMASHORTCITATION_RE.finditer(content)]
candidates = util.dedupe_urls(
util.clean_url(url) for url in candidates
if url and (url.startswith('http://') or url.startswith('https://')) and
# heuristic: ellipsized URLs are probably incomplete, so omit them.
not url.endswith('...') and not url.endswith('…'))
# check for redirect and add their final urls
if max_redirect_fetches and len(candidates) > max_redirect_fetches:
logger.warning(f'Found {len(candidates)} original post candidates, only resolving redirects for the first {max_redirect_fetches}')
redirects = {} # maps final URL to original URL for redirects
for url in candidates[:max_redirect_fetches]:
resolved = util.follow_redirects(url, **kwargs)
if (resolved.url != url and
resolved.headers.get('content-type', '').startswith('text/html')):
redirects[resolved.url] = url
candidates.extend(redirects.keys())
# use domains to determine which URLs are original post links vs mentions
originals = set()
mentions = set()
for url in util.dedupe_urls(candidates):
if url in redirects.values():
# this is a redirected original URL. postpone and handle it when we hit
# its final URL so that we know the final domain.
continue
domain = util.domain_from_link(url)
if not domain:
continue
if not include_reserved_hosts and (
('.' not in domain
or domain.split('.')[-1] in (util.RESERVED_TLDS | util.LOCAL_TLDS))):
continue
which = (originals if not domains or util.domain_or_parent_in(domain, domains)
else mentions)
which.add(url)
redirected_from = redirects.get(url)
if redirected_from and include_redirect_sources:
which.add(redirected_from)
logger.info(f'Original post discovery found original posts {originals}, mentions {mentions}')
return originals, mentions
[docs]
def prefix_urls(activity, field, prefix):
"""Adds a prefix to all matching URL fields, eg to inject a caching proxy.
Generally used with the ``image`` or ``stream`` fields. For example::
>>> prefix_urls({'actor': {'image': 'http://image'}}, 'image', 'https://proxy/')
{'actor': {'image': 'https://proxy/http://image'}}
Skips any URL fields that already start with the prefix. URLs are *not*
URL-encoded before adding the prefix. (This is currently used with our
caching-proxy Cloudflare worker and https://cloudimage.io/ , neither of which
URL-decodes.)
Args:
activity (dict): AS1 activity; modified in place
prefix (str)
"""
a = activity
for elem in ([a, a.get('object'), a.get('author'), a.get('actor')] +
a.get('replies', {}).get('items', []) +
a.get('attachments', []) +
a.get('tags', [])):
if elem:
for obj in util.get_list(elem, field):
url = obj.get('url')
if url and not url.startswith(prefix):
# Note that url isn't URL-encoded here, that's intentional, since
# cloudimage.io and the caching-proxy Cloudflare worker don't decode.
obj['url'] = prefix + url
if elem is not a:
prefix_urls(elem, field, prefix)
[docs]
def object_urls(obj):
"""Returns an object's unique URLs, preserving order."""
if isinstance(obj, str):
return obj
def value(obj):
got = obj.get('value') if isinstance(obj, dict) else obj
return got.strip() if got else None
return util.uniquify(util.trim_nulls(
value(u) for u in util.get_list(obj, 'url') + util.get_list(obj, 'urls')))
[docs]
def targets(obj):
"""Collects an AS1 activity or object's targets.
This is all ids/URLs that are direct "targets" of the activity, eg:
* the post it's replying to
* the post it's sharing
* the post it's reacting to
* the actor or other object it's tagging
* the event it's inviting someone to
* the event it's RSVPing to
* the link or object it's bookmarking
etc...
Args:
obj (dict): AS1 object or activity
Returns:
list of str: targets
"""
if not obj:
return []
targets = []
for o in [obj] + get_objects(obj):
targets.extend(get_ids(o, 'inReplyTo') +
get_ids(o, 'tags') +
util.get_urls(o, 'tags'))
verb = o.get('verb')
if verb in VERBS_WITH_OBJECT:
# prefer id or url, if available
# https://github.com/snarfed/bridgy-fed/issues/307
o_targets = get_ids(o, 'object') or util.get_urls(o, 'object')
targets.extend(o_targets)
if not o_targets:
logger.warning(f'{verb} missing target id/URL')
return util.dedupe_urls(targets)