def _targets(self):
"""
Returns: list of string URLs, the source's inReplyTos or objects
(if appropriate)
"""
targets = util.get_urls(self.source_obj, 'inReplyTo')
if targets:
return targets
if self.source_obj.get('verb') in source.VERBS_WITH_OBJECT:
return util.get_urls(self.source_obj, 'object')
def _prepare_activity(a, reader=True):
"""Preprocesses an activity to prepare it to be rendered as Atom.
Modifies a in place.
Args:
a: ActivityStreams 1 activity dict
reader: boolean, whether the output will be rendered in a feed reader.
Currently just includes location if True, not otherwise.
"""
act_type = source.object_type(a)
if not act_type or act_type == 'post':
primary = a.get('object', {})
else:
primary = a
obj = a.setdefault('object', {})
# Render content as HTML; escape &s
obj['rendered_content'] = _encode_ampersands(
microformats2.render_content(primary,
include_location=reader,
render_attachments=True))
# Make sure every activity has the title field, since Atom <entry> requires
# the title element.
if not a.get('title'):
a['title'] = util.ellipsize(
_encode_ampersands(
a.get('displayName') or a.get('content') or obj.get('title')
or obj.get('displayName') or obj.get('content') or 'Untitled'))
# strip HTML tags. the Atom spec says title is plain text:
# http://atomenabled.org/developers/syndication/#requiredEntryElements
a['title'] = xml.sax.saxutils.escape(
BeautifulSoup(a['title']).get_text(''))
children = []
image_urls_seen = set()
image_atts = []
# normalize attachments, render attached notes/articles
attachments = a.get('attachments') or obj.get('attachments') or []
for att in attachments:
att['stream'] = util.get_first(att, 'stream')
type = att.get('objectType')
if type == 'image':
image_atts.append(util.get_first(att, 'image'))
continue
image_urls_seen |= set(util.get_urls(att, 'image'))
if type in ('note', 'article'):
html = microformats2.render_content(att,
include_location=reader,
render_attachments=True)
author = att.get('author')
if author:
name = microformats2.maybe_linked_name(
microformats2.object_to_json(author).get('properties', []))
html = '%s: %s' % (name.strip(), html)
children.append(html)
# render image(s) that we haven't already seen
for image in image_atts + util.get_list(obj, 'image'):
if not image:
continue
url = image.get('url')
parsed = urllib.parse.urlparse(url)
rest = urllib.parse.urlunparse(('', '') + parsed[2:])
img_src_re = re.compile(
r"""src *= *['"] *((https?:)?//%s)?%s *['"]""" %
(re.escape(parsed.netloc), re.escape(rest)))
if (url and url not in image_urls_seen
and not img_src_re.search(obj['rendered_content'])):
children.append(microformats2.img(url))
image_urls_seen.add(url)
obj['rendered_children'] = [
_encode_ampersands(child) for child in children
]
# make sure published and updated are strict RFC 3339 timestamps
for prop in 'published', 'updated':
val = obj.get(prop)
if val:
obj[prop] = util.maybe_iso8601_to_rfc3339(val)
# Atom timestamps are even stricter than RFC 3339: they can't be naive ie
# time zone unaware. They must have either an offset or the Z suffix.
# https://www.feedvalidator.org/docs/error/InvalidRFC3339Date.html
if not util.TIMEZONE_OFFSET_RE.search(obj[prop]):
obj[prop] += 'Z'
def object_to_json(obj,
trim_nulls=True,
entry_class='h-entry',
default_object_type=None,
synthesize_content=True):
"""Converts an ActivityStreams object to microformats2 JSON.
Args:
obj: dict, a decoded JSON ActivityStreams object
trim_nulls: boolean, whether to remove elements with null or empty values
entry_class: string or sequence, the mf2 class(es) that entries should be
given (e.g. 'h-cite' when parsing a reference to a foreign entry).
defaults to 'h-entry'
default_object_type: string, the ActivityStreams objectType to use if one
is not present. defaults to None
synthesize_content: whether to generate synthetic content if the object
doesn't have its own, e.g. 'likes this.' or 'shared this.'
Returns:
dict, decoded microformats2 JSON
"""
if not obj or not isinstance(obj, dict):
return {}
obj_type = source.object_type(obj) or default_object_type
# if the activity type is a post, then it's really just a conduit
# for the object. for other verbs, the activity itself is the
# interesting thing
if obj_type == 'post':
primary = obj.get('object', {})
obj_type = source.object_type(primary) or default_object_type
else:
primary = obj
# TODO: extract snippet
name = primary.get('displayName', primary.get('title'))
summary = primary.get('summary')
author = obj.get('author', obj.get('actor', {}))
in_reply_tos = obj.get('inReplyTo',
obj.get('context', {}).get('inReplyTo', []))
is_rsvp = obj_type in ('rsvp-yes', 'rsvp-no', 'rsvp-maybe')
if (is_rsvp or obj_type == 'react') and obj.get('object'):
objs = obj['object']
in_reply_tos.extend(objs if isinstance(objs, list) else [objs])
# maps objectType to list of objects
attachments = defaultdict(list)
for prop in 'attachments', 'tags':
for elem in get_list(primary, prop):
attachments[elem.get('objectType')].append(elem)
# construct mf2!
ret = {
'type': (AS_TO_MF2_TYPE.get(obj_type) or [entry_class] if isinstance(
entry_class, basestring) else list(entry_class)),
'properties': {
'uid': [obj.get('id') or ''],
'numeric-id': [obj.get('numeric_id') or ''],
'name': [name],
'nickname': [obj.get('username') or ''],
'summary': [summary],
'url': (list(object_urls(obj) or object_urls(primary)) +
obj.get('upstreamDuplicates', [])),
'photo':
dedupe_urls(
get_urls(attachments, 'image', 'image') +
get_urls(primary, 'image')),
'video':
dedupe_urls(
get_urls(attachments, 'video', 'stream') +
get_urls(primary, 'stream')),
'audio':
get_urls(attachments, 'audio', 'stream'),
'published': [obj.get('published', primary.get('published', ''))],
'updated': [obj.get('updated', primary.get('updated', ''))],
'content': [{
'value':
xml.sax.saxutils.unescape(primary.get('content', '')),
'html':
render_content(primary,
include_location=False,
synthesize_content=synthesize_content),
}],
'in-reply-to':
util.trim_nulls([o.get('url') for o in in_reply_tos]),
'author': [
object_to_json(author,
trim_nulls=False,
default_object_type='person')
],
'location': [
object_to_json(primary.get('location', {}),
trim_nulls=False,
default_object_type='place')
],
'comment': [
object_to_json(c, trim_nulls=False, entry_class='h-cite')
for c in obj.get('replies', {}).get('items', [])
],
'start': [primary.get('startTime')],
'end': [primary.get('endTime')],
},
'children': [
object_to_json(a,
trim_nulls=False,
entry_class=['u-quotation-of', 'h-cite'])
for a in attachments['note'] + attachments['article']
]
}
# hashtags and person tags
tags = obj.get('tags', []) or get_first(obj, 'object', {}).get('tags', [])
ret['properties']['category'] = []
for tag in tags:
if tag.get('objectType') == 'person':
ret['properties']['category'].append(
object_to_json(tag, entry_class='u-category h-card'))
elif tag.get('objectType') == 'hashtag':
name = tag.get('displayName')
if name:
ret['properties']['category'].append(name)
# rsvp
if is_rsvp:
ret['properties']['rsvp'] = [obj_type[len('rsvp-'):]]
elif obj_type == 'invite':
invitee = object_to_json(obj.get('object'),
trim_nulls=False,
default_object_type='person')
ret['properties']['invitee'] = [invitee]
# like and repost mentions
for type, prop in ('favorite', 'like'), ('like', 'like'), ('share',
'repost'):
if obj_type == type:
# The ActivityStreams spec says the object property should always be a
# single object, but it's useful to let it be a list, e.g. when a like has
# multiple targets, e.g. a like of a post with original post URLs in it,
# which brid.gy does.
objs = get_list(obj, 'object')
ret['properties'][prop + '-of'] = [
# flatten contexts that are just a url
o['url']
if 'url' in o and set(o.keys()) <= set(['url', 'objectType'])
else object_to_json(o, trim_nulls=False, entry_class='h-cite')
for o in objs
]
else:
# received likes and reposts
ret['properties'][prop] = [
object_to_json(t, trim_nulls=False, entry_class='h-cite')
for t in tags if source.object_type(t) == type
]
# latitude & longitude
lat = long = None
position = ISO_6709_RE.match(primary.get('position') or '')
if position:
lat, long = position.groups()
if not lat:
lat = primary.get('latitude')
if not long:
long = primary.get('longitude')
if lat:
ret['properties']['latitude'] = [str(lat)]
if long:
ret['properties']['longitude'] = [str(long)]
if trim_nulls:
ret = util.trim_nulls(ret)
return ret
def object_to_json(obj, trim_nulls=True, entry_class='h-entry',
default_object_type=None, synthesize_content=True):
"""Converts an ActivityStreams object to microformats2 JSON.
Args:
obj: dict, a decoded JSON ActivityStreams object
trim_nulls: boolean, whether to remove elements with null or empty values
entry_class: string or sequence, the mf2 class(es) that entries should be
given (e.g. 'h-cite' when parsing a reference to a foreign entry).
defaults to 'h-entry'
default_object_type: string, the ActivityStreams objectType to use if one
is not present. defaults to None
synthesize_content: whether to generate synthetic content if the object
doesn't have its own, e.g. 'likes this.' or 'shared this.'
Returns:
dict, decoded microformats2 JSON
"""
if not obj or not isinstance(obj, dict):
return {}
obj_type = source.object_type(obj) or default_object_type
# if the activity type is a post, then it's really just a conduit
# for the object. for other verbs, the activity itself is the
# interesting thing
if obj_type == 'post':
primary = obj.get('object', {})
obj_type = source.object_type(primary) or default_object_type
else:
primary = obj
# TODO: extract snippet
name = primary.get('displayName', primary.get('title'))
summary = primary.get('summary')
author = obj.get('author', obj.get('actor', {}))
in_reply_tos = obj.get('inReplyTo', obj.get('context', {}).get('inReplyTo', []))
is_rsvp = obj_type in ('rsvp-yes', 'rsvp-no', 'rsvp-maybe')
if (is_rsvp or obj_type == 'react') and obj.get('object'):
objs = obj['object']
in_reply_tos.extend(objs if isinstance(objs, list) else [objs])
# maps objectType to list of objects
attachments = defaultdict(list)
for prop in 'attachments', 'tags':
for elem in get_list(primary, prop):
attachments[elem.get('objectType')].append(elem)
# construct mf2!
ret = {
'type': (AS_TO_MF2_TYPE.get(obj_type) or
[entry_class] if isinstance(entry_class, basestring)
else list(entry_class)),
'properties': {
'uid': [obj.get('id') or ''],
'numeric-id': [obj.get('numeric_id') or ''],
'name': [name],
'nickname': [obj.get('username') or ''],
'summary': [summary],
'url': (list(object_urls(obj) or object_urls(primary)) +
obj.get('upstreamDuplicates', [])),
# photo is special cased below, to handle alt
'video': dedupe_urls(get_urls(attachments, 'video', 'stream') +
get_urls(primary, 'stream')),
'audio': get_urls(attachments, 'audio', 'stream'),
'published': [obj.get('published', primary.get('published', ''))],
'updated': [obj.get('updated', primary.get('updated', ''))],
'content': [{
'value': xml.sax.saxutils.unescape(primary.get('content', '')),
'html': render_content(primary, include_location=False,
synthesize_content=synthesize_content),
}],
'in-reply-to': util.trim_nulls([o.get('url') for o in in_reply_tos]),
'author': [object_to_json(
author, trim_nulls=False, default_object_type='person')],
'location': [object_to_json(
primary.get('location', {}), trim_nulls=False,
default_object_type='place')],
'comment': [object_to_json(c, trim_nulls=False, entry_class='h-cite')
for c in obj.get('replies', {}).get('items', [])],
'start': [primary.get('startTime')],
'end': [primary.get('endTime')],
},
'children': (
# silly hack: i haven't found anywhere in AS1 or AS2 to indicate that
# something is being "quoted," like in a quote tweet, so i cheat and use
# extra knowledge here that quoted tweets are converted to note
# attachments, but URLs in the tweet text are converted to article tags.
[object_to_json(a, trim_nulls=False, entry_class=['u-quotation-of', 'h-cite'])
for a in attachments['note'] if 'startIndex' not in a] +
[object_to_json(a, trim_nulls=False, entry_class=['h-cite'])
for a in attachments['article'] if 'startIndex' not in a])
}
# photos, including alt text
photo_urls = set()
ret['properties']['photo'] = []
for image in get_list(attachments, 'image') + [primary]:
for url in get_urls(image, 'image'):
if url and url not in photo_urls:
photo_urls.add(url)
name = get_first(image, 'image', {}).get('displayName')
ret['properties']['photo'].append({'value': url, 'alt': name} if name else url)
# hashtags and person tags
if obj_type == 'tag':
ret['properties']['tag-of'] = util.get_urls(obj, 'target')
tags = obj.get('tags', []) or get_first(obj, 'object', {}).get('tags', [])
if not tags and obj_type == 'tag':
tags = util.get_list(obj, 'object')
ret['properties']['category'] = []
for tag in tags:
if tag.get('objectType') == 'person':
ret['properties']['category'].append(
object_to_json(tag, entry_class='u-category h-card'))
elif tag.get('objectType') == 'hashtag' or obj_type == 'tag':
name = tag.get('displayName')
if name:
ret['properties']['category'].append(name)
# rsvp
if is_rsvp:
ret['properties']['rsvp'] = [obj_type[len('rsvp-'):]]
elif obj_type == 'invite':
invitee = object_to_json(obj.get('object'), trim_nulls=False,
default_object_type='person')
ret['properties']['invitee'] = [invitee]
# like and repost mentions
for type, prop in (
('favorite', 'like'),
('follow', 'follow'),
('like', 'like'),
('share', 'repost'),
):
if obj_type == type:
# The ActivityStreams spec says the object property should always be a
# single object, but it's useful to let it be a list, e.g. when a like has
# multiple targets, e.g. a like of a post with original post URLs in it,
# which brid.gy does.
objs = get_list(obj, 'object')
ret['properties'][prop + '-of'] = [
# flatten contexts that are just a url
o['url'] if 'url' in o and set(o.keys()) <= set(['url', 'objectType'])
else object_to_json(o, trim_nulls=False, entry_class='h-cite')
for o in objs]
else:
# received likes and reposts
ret['properties'][prop] = [
object_to_json(t, trim_nulls=False, entry_class='h-cite')
for t in tags if source.object_type(t) == type]
# latitude & longitude
lat = long = None
position = ISO_6709_RE.match(primary.get('position') or '')
if position:
lat, long = position.groups()
if not lat:
lat = primary.get('latitude')
if not long:
long = primary.get('longitude')
if lat:
ret['properties']['latitude'] = [str(lat)]
if long:
ret['properties']['longitude'] = [str(long)]
if trim_nulls:
ret = util.trim_nulls(ret)
return ret
def _prepare_activity(a, reader=True):
"""Preprocesses an activity to prepare it to be rendered as Atom.
Modifies a in place.
Args:
a: ActivityStreams 1 activity dict
reader: boolean, whether the output will be rendered in a feed reader.
Currently just includes location if True, not otherwise.
"""
act_type = source.object_type(a)
obj = util.get_first(a, 'object', default={})
primary = obj if (not act_type or act_type == 'post') else a
# Render content as HTML; escape &s
obj['rendered_content'] = _encode_ampersands(microformats2.render_content(
primary, include_location=reader, render_attachments=True))
# Make sure every activity has the title field, since Atom <entry> requires
# the title element.
if not a.get('title'):
a['title'] = util.ellipsize(_encode_ampersands(
a.get('displayName') or a.get('content') or obj.get('title') or
obj.get('displayName') or obj.get('content') or 'Untitled'))
# strip HTML tags. the Atom spec says title is plain text:
# http://atomenabled.org/developers/syndication/#requiredEntryElements
a['title'] = xml.sax.saxutils.escape(BeautifulSoup(a['title']).get_text(''))
children = []
image_urls_seen = set()
image_atts = []
# normalize actor images
for elem in a, obj:
actor = elem.get('actor')
if actor:
actor['image'] = util.get_first(actor, 'image')
# normalize attachments, render attached notes/articles
attachments = a.get('attachments') or obj.get('attachments') or []
for att in attachments:
att['stream'] = util.get_first(att, 'stream')
type = att.get('objectType')
if type == 'image':
att['image'] = util.get_first(att, 'image')
image_atts.append(att['image'])
continue
image_urls_seen |= set(util.get_urls(att, 'image'))
if type in ('note', 'article'):
html = microformats2.render_content(att, include_location=reader,
render_attachments=True)
author = att.get('author')
if author:
name = microformats2.maybe_linked_name(
microformats2.object_to_json(author).get('properties') or {})
html = '%s: %s' % (name.strip(), html)
children.append(html)
# render image(s) that we haven't already seen
for image in image_atts + util.get_list(obj, 'image'):
if not image:
continue
url = image.get('url')
parsed = urllib.parse.urlparse(url)
rest = urllib.parse.urlunparse(('', '') + parsed[2:])
img_src_re = re.compile(r"""src *= *['"] *((https?:)?//%s)?%s *['"]""" %
(re.escape(parsed.netloc), re.escape(rest)))
if (url and url not in image_urls_seen and
not img_src_re.search(obj['rendered_content'])):
children.append(microformats2.img(url))
image_urls_seen.add(url)
obj['rendered_children'] = [_encode_ampersands(child) for child in children]
# make sure published and updated are strict RFC 3339 timestamps
for prop in 'published', 'updated':
val = obj.get(prop)
if val:
obj[prop] = util.maybe_iso8601_to_rfc3339(val)
# Atom timestamps are even stricter than RFC 3339: they can't be naive ie
# time zone unaware. They must have either an offset or the Z suffix.
# https://www.feedvalidator.org/docs/error/InvalidRFC3339Date.html
if not util.TIMEZONE_OFFSET_RE.search(obj[prop]):
obj[prop] += 'Z'
def render_content(obj, include_location=True, synthesize_content=True,
render_attachments=False, render_image=False,
white_space_pre=True):
"""Renders the content of an ActivityStreams object as HTML.
Includes tags, mentions, and non-note/article attachments. (Note/article
attachments are converted to mf2 children in object_to_json and then rendered
in json_to_html.)
Note that the returned HTML is included in Atom as well as HTML documents,
so it *must* be HTML4 / XHTML, not HTML5! All tags must be closed, etc.
Args:
obj: decoded JSON ActivityStreams object
include_location: boolean, whether to render location, if provided
synthesize_content: boolean, whether to generate synthetic content if the
object doesn't have its own, e.g. 'likes this.' or 'shared this.'
render_attachments: boolean, whether to render attachments, eg links,
images, audio, and video
render_image: boolean, whether to render the object's image(s)
white_space_pre: boolean, whether to wrap in CSS white-space: pre. If False,
newlines will be converted to <br> tags instead. Background:
https://indiewebcamp.com/note#Indieweb_whitespace_thinking
Returns:
string, rendered HTML
"""
content = obj.get('content', '')
# extract tags. preserve order but de-dupe, ie don't include a tag more than
# once.
seen_ids = set()
mentions = []
tags = {} # maps string objectType to list of tag objects
for t in obj.get('tags', []):
id = t.get('id')
if id and id in seen_ids:
continue
seen_ids.add(id)
if 'startIndex' in t and 'length' in t and 'url' in t:
mentions.append(t)
else:
tags.setdefault(source.object_type(t), []).append(t)
# linkify embedded mention tags inside content.
if mentions:
mentions.sort(key=lambda t: t['startIndex'])
last_end = 0
orig = util.WideUnicode(content)
content = util.WideUnicode('')
for tag in mentions:
start = tag['startIndex']
end = start + tag['length']
content = util.WideUnicode('%s%s<a href="%s">%s</a>' % (
content, orig[last_end:start], tag['url'], orig[start:end]))
last_end = end
content += orig[last_end:]
# is whitespace in this content meaningful? standard heuristic: if there are
# no HTML tags in it, and it has a newline, then assume yes.
# https://indiewebcamp.com/note#Indieweb_whitespace_thinking
# https://github.com/snarfed/granary/issues/80
if content and not obj.get('content_is_html') and '\n' in content:
if white_space_pre:
content = '<div style="white-space: pre">%s</div>' % content
else:
content = content.replace('\n', '<br />\n')
# linkify embedded links. ignore the "mention" tags that we added ourselves.
# TODO: fix the bug in test_linkify_broken() in webutil/tests/test_util.py, then
# uncomment this.
# if content:
# content = util.linkify(content)
# the image field. may be multiply valued.
rendered_urls = set()
if render_image:
urls = get_urls(obj, 'image')
content += _render_attachments([{
'objectType': 'image',
'image': {'url': url},
} for url in urls], obj)
rendered_urls = set(urls)
# attachments, e.g. links (aka articles)
# TODO: use oEmbed? http://oembed.com/ , http://code.google.com/p/python-oembed/
if render_attachments:
atts = [a for a in obj.get('attachments', [])
if a.get('objectType') not in ('note', 'article')
and get_url(a, 'image') not in rendered_urls]
content += _render_attachments(atts + tags.pop('article', []), obj)
# generate share/like contexts if the activity does not have content
# of its own
obj_type = source.object_type(obj)
for as_type, verb in (
('favorite', 'Favorites'), ('like', 'Likes'), ('share', 'Shared')):
if (not synthesize_content or obj_type != as_type or 'object' not in obj or
'content' in obj):
continue
targets = get_list(obj, 'object')
if not targets:
continue
for target in targets:
# sometimes likes don't have enough content to render anything
# interesting
if 'url' in target and set(target) <= set(['url', 'objectType']):
content += '<a href="%s">%s this.</a>' % (
target.get('url'), verb.lower())
else:
author = target.get('author', target.get('actor', {}))
# special case for twitter RT's
if obj_type == 'share' and 'url' in obj and re.search(
'^https?://(?:www\.|mobile\.)?twitter\.com/', obj.get('url')):
content += 'RT <a href="%s">@%s</a> ' % (
target.get('url', '#'), author.get('username'))
else:
# image looks bad in the simplified rendering
author = {k: v for k, v in author.items() if k != 'image'}
content += '%s <a href="%s">%s</a> by %s' % (
verb, target.get('url', '#'),
target.get('displayName', target.get('title', 'a post')),
hcard_to_html(object_to_json(author, default_object_type='person')),
)
content += render_content(target, include_location=include_location,
synthesize_content=synthesize_content,
white_space_pre=white_space_pre)
# only include the first context in the content (if there are
# others, they'll be included as separate properties)
break
break
if render_attachments and obj.get('verb') == 'share':
atts = [att for att in itertools.chain.from_iterable(
o.get('attachments', []) for o in util.get_list(obj, 'object'))
if att.get('objectType') not in ('note', 'article')]
content += _render_attachments(atts, obj)
# location
loc = obj.get('location')
if include_location and loc:
content += '\n<p>%s</p>' % hcard_to_html(
object_to_json(loc, default_object_type='place'),
parent_props=['p-location'])
# these are rendered manually in json_to_html()
for type in 'like', 'share', 'react', 'person':
tags.pop(type, None)
# render the rest
content += tags_to_html(tags.pop('hashtag', []), 'p-category')
content += tags_to_html(tags.pop('mention', []), 'u-mention', visible=False)
content += tags_to_html(sum(tags.values(), []), 'tag')
return content
def object_to_json(obj, trim_nulls=True, entry_class='h-entry',
default_object_type=None, synthesize_content=True):
"""Converts an ActivityStreams object to microformats2 JSON.
Args:
obj: dict, a decoded JSON ActivityStreams object
trim_nulls: boolean, whether to remove elements with null or empty values
entry_class: string or sequence, the mf2 class(es) that entries should be
given (e.g. 'h-cite' when parsing a reference to a foreign entry).
defaults to 'h-entry'
default_object_type: string, the ActivityStreams objectType to use if one
is not present. defaults to None
synthesize_content: whether to generate synthetic content if the object
doesn't have its own, e.g. 'likes this.' or 'shared this.'
Returns:
dict, decoded microformats2 JSON
"""
if not obj or not isinstance(obj, dict):
return {}
obj_type = source.object_type(obj) or default_object_type
# if the activity type is a post, then it's really just a conduit
# for the object. for other verbs, the activity itself is the
# interesting thing
if obj_type == 'post':
primary = obj.get('object', {})
obj_type = source.object_type(primary) or default_object_type
else:
primary = obj
# TODO: extract snippet
name = primary.get('displayName', primary.get('title'))
summary = primary.get('summary')
author = obj.get('author', obj.get('actor', {}))
in_reply_tos = obj.get('inReplyTo') or []
if not in_reply_tos:
context = obj.get('context')
if context and isinstance(context, dict):
in_reply_tos = context.get('inReplyTo') or []
is_rsvp = obj_type in ('rsvp-yes', 'rsvp-no', 'rsvp-maybe')
if (is_rsvp or obj_type == 'react') and obj.get('object'):
objs = obj['object']
in_reply_tos.extend(objs if isinstance(objs, list) else [objs])
# maps objectType to list of objects
attachments = defaultdict(list)
for prop in 'attachments', 'tags':
for elem in get_list(primary, prop):
attachments[elem.get('objectType')].append(elem)
# prefer duration and size from object's stream, then first video, then first
# audio
stream = {}
for candidate in [obj] + attachments['video'] + attachments['audio']:
for stream in get_list(candidate, 'stream'):
if stream:
break
duration = stream.get('duration')
if duration is not None:
if util.is_int(duration):
duration = str(duration)
else:
logging('Ignoring duration %r; expected int, got %s', duration.__class__)
duration = None
sizes = []
size = stream.get('size')
if size:
sizes = [str(size)]
# construct mf2!
ret = {
'type': (AS_TO_MF2_TYPE.get(obj_type) or
[entry_class] if isinstance(entry_class, str)
else list(entry_class)),
'properties': {
'uid': [obj.get('id') or ''],
'numeric-id': [obj.get('numeric_id') or ''],
'name': [name],
'nickname': [obj.get('username') or ''],
'summary': [summary],
'url': (list(object_urls(obj) or object_urls(primary)) +
obj.get('upstreamDuplicates', [])),
# photo is special cased below, to handle alt
'video': dedupe_urls(get_urls(attachments, 'video', 'stream') +
get_urls(primary, 'stream')),
'audio': get_urls(attachments, 'audio', 'stream'),
'duration': [duration],
'size': sizes,
'published': [obj.get('published', primary.get('published', ''))],
'updated': [obj.get('updated', primary.get('updated', ''))],
'in-reply-to': util.trim_nulls([o.get('url') for o in in_reply_tos]),
'author': [object_to_json(
author, trim_nulls=False, default_object_type='person')],
'location': [object_to_json(
primary.get('location', {}), trim_nulls=False,
default_object_type='place')],
'comment': [object_to_json(c, trim_nulls=False, entry_class='h-cite')
for c in obj.get('replies', {}).get('items', [])],
'start': [primary.get('startTime')],
'end': [primary.get('endTime')],
},
'children': (
# silly hack: i haven't found anywhere in AS1 or AS2 to indicate that
# something is being "quoted," like in a quote tweet, so i cheat and use
# extra knowledge here that quoted tweets are converted to note
# attachments, but URLs in the tweet text are converted to article tags.
[object_to_json(a, trim_nulls=False, entry_class=['u-quotation-of', 'h-cite'])
for a in attachments['note'] if 'startIndex' not in a] +
[object_to_json(a, trim_nulls=False, entry_class=['h-cite'])
for a in attachments['article'] if 'startIndex' not in a])
}
# content. emulate e- vs p- microformats2 parsing: e- if there are HTML tags,
# otherwise p-.
# https://indiewebcamp.com/note#Indieweb_whitespace_thinking
text = xml.sax.saxutils.unescape(primary.get('content', ''))
html = render_content(primary, include_location=False,
synthesize_content=synthesize_content)
if '<' in html:
ret['properties']['content'] = [{'value': text, 'html': html}]
else:
ret['properties']['content'] = [text]
# photos, including alt text
photo_urls = set()
ret['properties']['photo'] = []
for image in get_list(attachments, 'image') + [primary]:
for url in get_urls(image, 'image'):
if url and url not in photo_urls:
photo_urls.add(url)
name = get_first(image, 'image', {}).get('displayName')
ret['properties']['photo'].append({'value': url, 'alt': name} if name else url)
# hashtags and person tags
if obj_type == 'tag':
ret['properties']['tag-of'] = util.get_urls(obj, 'target')
tags = obj.get('tags', []) or get_first(obj, 'object', {}).get('tags', [])
if not tags and obj_type == 'tag':
tags = util.get_list(obj, 'object')
ret['properties']['category'] = []
for tag in tags:
if tag.get('objectType') == 'person':
ret['properties']['category'].append(
object_to_json(tag, entry_class='u-category h-card'))
elif tag.get('objectType') == 'hashtag' or obj_type == 'tag':
name = tag.get('displayName')
if name:
ret['properties']['category'].append(name)
# rsvp
if is_rsvp:
ret['properties']['rsvp'] = [obj_type[len('rsvp-'):]]
elif obj_type == 'invite':
invitee = object_to_json(obj.get('object'), trim_nulls=False,
default_object_type='person')
ret['properties']['invitee'] = [invitee]
# like and repost mentions
for type, prop in (
('favorite', 'like'),
('follow', 'follow'),
('like', 'like'),
('share', 'repost'),
):
if obj_type == type:
# The ActivityStreams spec says the object property should always be a
# single object, but it's useful to let it be a list, e.g. when a like has
# multiple targets, e.g. a like of a post with original post URLs in it,
# which brid.gy does.
objs = get_list(obj, 'object')
ret['properties'][prop + '-of'] = [
# flatten contexts that are just a url
o['url'] if 'url' in o and set(o.keys()) <= set(['url', 'objectType'])
else object_to_json(o, trim_nulls=False, entry_class='h-cite')
for o in objs]
else:
# received likes and reposts
ret['properties'][prop] = [
object_to_json(t, trim_nulls=False, entry_class='h-cite')
for t in tags if source.object_type(t) == type]
# latitude & longitude
lat = long = None
position = ISO_6709_RE.match(primary.get('position') or '')
if position:
lat, long = position.groups()
if not lat:
lat = primary.get('latitude')
if not long:
long = primary.get('longitude')
if lat:
ret['properties']['latitude'] = [str(lat)]
if long:
ret['properties']['longitude'] = [str(long)]
if trim_nulls:
ret = util.trim_nulls(ret)
return ret