def bootstrap_indexing(db):
from assembl.indexing import indexing_active
from assembl.indexing.utils import check_mapping, maybe_create_and_reindex
if not indexing_active():
return
index_name = config.get('elasticsearch_index')
if not check_mapping(index_name):
with locked_transaction(db, 1237) as session:
maybe_create_and_reindex(index_name, session)
def reindex_content(content, action='update'):
"""Index, reindex or unindex content. This function is called
by the after_insert/update/delete sqlalchemy events.
"""
from assembl.models.post import PublicationStates
from assembl.models import (AgentStatusInDiscussion, Post, AgentProfile,
Idea, IdeaContentLink, IdeaAnnouncement,
SentimentOfPost, Extract, Username)
if not indexing_active():
return
indexed_contents = (Post, AgentProfile, Idea, Extract)
changes = get_changes()
if action == 'delete' and isinstance(content, indexed_contents):
changes.unindex_content(content)
elif isinstance(content, AgentProfile):
changes.index_content(content)
elif isinstance(content, Username):
changes.index_content(content.user)
for post in content.user.posts_created:
reindex_content(post)
for extract in content.user.extracts_created:
reindex_content(extract)
elif isinstance(content, AgentStatusInDiscussion):
reindex_content(content.agent_profile)
elif type(content) == Idea: # only index Idea, not Thematic or Question
if (not content.hidden and content.tombstone_date is None):
changes.index_content(content)
else:
changes.unindex_content(content)
elif isinstance(content, Post):
if (content.publication_state == PublicationStates.PUBLISHED
and not content.hidden and content.tombstone_date is None
and not content.is_bright_mirror_fiction()):
changes.index_content(content)
for extract in content.extracts:
changes.index_content(extract)
else:
changes.unindex_content(content)
for extract in content.extracts:
changes.unindex_content(extract)
elif isinstance(content, Extract):
# warning: should always be above isinstance(content, IdeaContentLink) block
changes.index_content(content)
elif isinstance(content, IdeaContentLink):
# A AssemblPost is indexed before any IdeaRelatedPostLink is created,
# so be sure to reindex content.content if we have a IdeaContentLink
reindex_content(content.content)
elif isinstance(content, IdeaAnnouncement):
reindex_content(content.idea)
elif isinstance(content, SentimentOfPost):
reindex_content(content.post_from_sentiments)
def reindex_content(content, action='update'):
"""Index, reindex or unindex content. This function is called
by the after_insert/update/delete sqlalchemy events.
"""
from assembl.models.post import PublicationStates
from assembl.models import (
AgentStatusInDiscussion, Post, AgentProfile, Idea,
IdeaContentLink, IdeaAnnouncement, SentimentOfPost, Extract)
if not indexing_active():
return
indexed_contents = (Post, AgentProfile, Idea, Extract)
changes = get_changes()
if action == 'delete' and isinstance(content, indexed_contents):
changes.unindex_content(content)
elif isinstance(content, AgentProfile):
changes.index_content(content)
elif isinstance(content, AgentStatusInDiscussion):
reindex_content(content.agent_profile)
elif type(content) == Idea: # only index Idea, not Thematic or Question
if (not content.hidden and content.tombstone_date is None):
changes.index_content(content)
else:
changes.unindex_content(content)
elif isinstance(content, Post):
if (content.publication_state == PublicationStates.PUBLISHED and
not content.hidden and content.tombstone_date is None and
not content.is_bright_mirror_fiction()):
changes.index_content(content)
for extract in content.extracts:
changes.index_content(extract)
else:
changes.unindex_content(content)
for extract in content.extracts:
changes.unindex_content(extract)
elif isinstance(content, Extract):
# warning: should always be above isinstance(content, IdeaContentLink) block
changes.index_content(content)
elif isinstance(content, IdeaContentLink):
# A AssemblPost is indexed before any IdeaRelatedPostLink is created,
# so be sure to reindex content.content if we have a IdeaContentLink
reindex_content(content.content)
elif isinstance(content, IdeaAnnouncement):
reindex_content(content.idea)
elif isinstance(content, SentimentOfPost):
reindex_content(content.post_from_sentiments)
def reindex_content(content, action='update'):
"""Index, reindex or unindex content. This function is called
by the after_insert/update/delete sqlalchemy events.
"""
from assembl.models.post import PublicationStates
from assembl.models import (AgentStatusInDiscussion, Post, AgentProfile,
Idea, IdeaContentLink, IdeaAnnouncement,
SentimentOfPost)
if not indexing_active():
return
indexed_contents = (Post, AgentProfile, Idea)
changes = get_changes()
if action == 'delete' and isinstance(content, indexed_contents):
changes.unindex_content(content)
elif isinstance(content, AgentProfile):
changes.index_content(content)
elif isinstance(content, AgentStatusInDiscussion):
reindex_content(content.agent_profile)
elif isinstance(content, Idea):
if (not content.hidden and content.tombstone_date is None):
changes.index_content(content)
else:
changes.unindex_content(content)
elif isinstance(content, Post):
# don't index proposition posts
if content.type == 'proposition_post':
return
if (content.publication_state == PublicationStates.PUBLISHED
and not content.hidden and content.tombstone_date is None):
changes.index_content(content)
else:
changes.unindex_content(content)
elif isinstance(content, IdeaContentLink):
# A AssemblPost is indexed before any IdeaRelatedPostLink is created,
# so be sure to reindex content.content if we have a IdeaContentLink
reindex_content(content.content)
elif isinstance(content, IdeaAnnouncement):
reindex_content(content.idea)
elif isinstance(content, SentimentOfPost):
reindex_content(content.post_from_sentiments)
def join_transaction(event):
if indexing_active():
changes._join()
def search_endpoint(context, request):
if not indexing_active():
return HTTPServiceUnavailable("Indexing inactive")
query = request.json_body
# u'query': {u'bool': {u'filter': [{u'term': {u'discussion_id': u'23'}}]}}
filters = [fil for fil in query['query']['bool']['filter']]
discussion_id = [
f.values()[0].values()[0] for f in filters
if 'discussion_id' in f.values()[0].keys()
][0]
discussion = models.Discussion.get_instance(discussion_id)
if discussion is None:
raise HTTPUnauthorized()
user_id = authenticated_userid(request) or Everyone
permissions = get_permissions(user_id, discussion_id)
if not discussion.user_can(user_id, CrudPermissions.READ, permissions):
raise HTTPUnauthorized()
es = connect()
index_name = get_index_settings()['index_name']
# print get_curl_query(query)
result = es.search(index=index_name, body=query)
# add creator_name in each hit
creator_ids = set([
hit['_source']['creator_id'] for hit in result['hits']['hits']
if hit['_source'].get('creator_id', None) is not None
])
session = get_session_maker()
creators = session.query(
models.AgentProfile.id, models.AgentProfile.name).filter(
models.AgentProfile.id.in_(creator_ids)).all()
creators_by_id = dict(creators)
for hit in result['hits']['hits']:
source = hit['_source']
creator_id = source.get('creator_id', None)
# Remove inner_hits key to not leak posts from private discussion.
# You can easily craft a query to get the participants of a public
# discussion and do a has_child filter with inner_hits on a private discussion.
if 'inner_hits' in hit:
del hit['inner_hits']
if creator_id is not None:
source['creator_name'] = creators_by_id.get(creator_id)
if hit['_type'] == 'idea':
idea = models.Idea.get_instance(source['id'])
# The check is not really necessary because it's the same
# 'read' permission as the discussion, but it doesn't cost anything
# to check it and the READ permission may change in the future.
if not idea.user_can(user_id, CrudPermissions.READ, permissions):
raise HTTPUnauthorized
source['num_posts'] = idea.num_posts
source['num_contributors'] = idea.num_contributors
elif hit['_type'] == 'user':
agent_profile = models.AgentProfile.get_instance(source['id'])
if not agent_profile.user_can(user_id, CrudPermissions.READ,
permissions):
raise HTTPUnauthorized
source['num_posts'] = agent_profile.count_posts_in_discussion(
discussion_id)
# Don't do an extra request to verify the CrudPermissions.READ permission
# for post or synthesis.
# It's currently the same 'read' permission as the discussion.
# elif hit['_type'] in ('synthesis', 'post'):
# post = models.Post.get_instance(source['id'])
# if not post.user_can(user_id, CrudPermissions.READ, permissions):
# raise HTTPUnauthorized
return result
def search_endpoint(context, request):
if not indexing_active():
return HTTPServiceUnavailable("Indexing inactive")
query = request.json_body
# u'query': {u'bool': {u'filter': [{u'term': {u'discussion_id': u'23'}}]}}
filters = [fil for fil in query['query']['bool']['filter']]
discussion_id = [f.values()[0].values()[0]
for f in filters if 'discussion_id' in f.values()[0].keys()][0]
discussion = models.Discussion.get_instance(discussion_id)
if discussion is None:
raise HTTPUnauthorized()
user_id = request.authenticated_userid or Everyone
permissions = get_permissions(user_id, discussion_id)
if not discussion.user_can(user_id, CrudPermissions.READ, permissions):
raise HTTPUnauthorized()
es = connect()
index_name = get_index_settings(config)['index_name']
# print get_curl_query(query)
result = es.search(index=index_name, body=query)
# add creator_name in each hit
creator_ids = set([hit['_source']['creator_id']
for hit in result['hits']['hits']
if hit['_source'].get('creator_id', None) is not None])
session = get_session_maker()
creators = session.query(models.AgentProfile.id, models.AgentProfile.name
).filter(models.AgentProfile.id.in_(creator_ids)).all()
creators_by_id = dict(creators)
for hit in result['hits']['hits']:
source = hit['_source']
creator_id = source.get('creator_id', None)
# Remove inner_hits key to not leak posts from private discussion.
# You can easily craft a query to get the participants of a public
# discussion and do a has_child filter with inner_hits on a private discussion.
if 'inner_hits' in hit:
del hit['inner_hits']
if creator_id is not None:
source['creator_name'] = creators_by_id.get(creator_id)
if hit['_type'] == 'idea':
idea = models.Idea.get_instance(source['id'])
# The check is not really necessary because it's the same
# 'read' permission as the discussion, but it doesn't cost anything
# to check it and the READ permission may change in the future.
if not idea.user_can(user_id, CrudPermissions.READ, permissions):
raise HTTPUnauthorized
source['num_posts'] = idea.num_posts
source['num_contributors'] = idea.num_contributors
elif hit['_type'] == 'user':
agent_profile = models.AgentProfile.get_instance(source['id'])
if not agent_profile.user_can(user_id, CrudPermissions.READ, permissions):
raise HTTPUnauthorized
source['num_posts'] = agent_profile.count_posts_in_discussion(discussion_id)
# Don't do an extra request to verify the CrudPermissions.READ permission
# for post or synthesis.
# It's currently the same 'read' permission as the discussion.
# elif hit['_type'] in ('synthesis', 'post'):
# post = models.Post.get_instance(source['id'])
# if not post.user_can(user_id, CrudPermissions.READ, permissions):
# raise HTTPUnauthorized
return result
def join_transaction(event):
if indexing_active():
get_changes()._join()
