def index_entity(entity):
"""Index an entity."""
if entity.deleted_at is not None:
return delete_entity(entity.id)
data = {
'foreign_ids': entity.foreign_ids,
'data': entity.data,
'created_at': entity.created_at,
'updated_at': entity.updated_at,
'bulk': False,
'roles': entity.collection.roles,
'collection_id': entity.collection_id,
'properties': {
'name': [entity.name]
}
}
for k, v in entity.data.items():
data['properties'][k] = ensure_list(v)
# data['$documents'] = get_count(entity)
data = finalize_index(data, entity.schema)
es.index(index=entity_index(),
doc_type=entity_type(),
id=entity.id,
body=data)
data['id'] = entity.id
return data
def _index_updates(collection, entities):
"""Look up existing index documents and generate an updated form.
This is necessary to make the index accumulative, i.e. if an entity or link
gets indexed twice with different field values, it'll add up the different
field values into a single record. This is to avoid overwriting the
document and losing field values. An alternative solution would be to
implement this in Groovy on the ES.
"""
common = {
'collection_id': collection.id,
'bulk': True,
'roles': collection.roles,
'updated_at': datetime.utcnow()
}
if not len(entities):
return
query = {
'query': {
'ids': {
'values': list(entities.keys())
}
},
'_source': ['schema', 'properties', 'created_at']
}
result = search_safe(index=entity_index(), body=query)
for doc in result.get('hits').get('hits', []):
entity_id = doc['_id']
entity = entities.get(entity_id)
existing = doc.get('_source')
combined = model.merge(existing, entity)
combined['created_at'] = existing.get('created_at')
entities[entity_id] = combined
for doc_id, entity in entities.items():
entity.pop('id', None)
entity.update(common)
schema = model.get(entity.get('schema'))
entity = finalize_index(entity, schema, [])
# pprint(entity)
yield {
'_id': doc_id,
'_index': entity_index(),
'_type': 'doc',
'_source': entity
}
def index_single(obj, proxy, data, texts, sync=False):
"""Indexing aspects common to entities and documents."""
data = finalize_index(proxy, data, texts)
data['bulk'] = False
data['collection_id'] = obj.collection.id
data['created_at'] = obj.created_at
data['updated_at'] = obj.updated_at
# pprint(data)
refresh = 'wait_for' if sync else False
return index_safe(entity_index(), obj.id, data, refresh=refresh)
def index_single(obj, data, texts):
"""Indexing aspects common to entities and documents."""
data['bulk'] = False
data['roles'] = obj.collection.roles
data['collection_id'] = obj.collection.id
data['created_at'] = obj.created_at
data['updated_at'] = obj.updated_at
data = finalize_index(data, obj.model, texts)
data = clean_dict(data)
return index_safe(entity_index(), obj.id, data)
def ingest_upload(id):
collection = get_db_collection(id, request.authz.WRITE)
meta, foreign_id = _load_metadata(collection)
parent_id = _load_parent(collection, meta)
upload_dir = mkdtemp(prefix='aleph.upload.')
try:
documents = []
for storage in request.files.values():
path = safe_filename(storage.filename, default='upload')
path = os.path.join(upload_dir, path)
storage.save(path)
content_hash = checksum(path)
document = Document.by_keys(collection=collection,
parent_id=parent_id,
foreign_id=foreign_id,
content_hash=content_hash)
document.update(meta)
document.uploader_id = request.authz.id
ingest_document(document, path)
documents.append(document)
if not len(request.files):
# If there is no files uploaded, try to create an empty
# directory instead. Maybe this should be more explicit,
# but it seemed like the most simple way of fitting it
# into the API.
document = Document.by_keys(collection=collection,
parent_id=parent_id,
foreign_id=foreign_id)
document.schema = Document.SCHEMA_FOLDER
document.update(meta)
document.uploader_id = request.authz.id
ingest_document(document, None)
documents.append(document)
finally:
shutil.rmtree(upload_dir)
if collection.casefile:
for document in documents:
params = {'document': document, 'collection': collection}
publish(Events.INGEST_DOCUMENT,
actor_id=document.uploader_id,
params=params)
# Update child counts in index.
if parent_id is not None:
index_document_id.apply_async([parent_id], priority=1)
refresh_index(index=entity_index())
return jsonify({
'status':
'ok',
'documents': [CombinedSchema().dump(d).data for d in documents]
})
def index_single(obj, data, texts):
"""Indexing aspects common to entities and documents."""
data['bulk'] = False
data['roles'] = obj.collection.roles
data['collection_id'] = obj.collection.id
data['created_at'] = obj.created_at
data['updated_at'] = obj.updated_at
data = finalize_index(data, obj.model, texts)
data = clean_dict(data)
es.index(index=entity_index(), doc_type='doc', id=str(obj.id), body=data)
data['id'] = str(obj.id)
return data
def update_roles(collection):
"""Update the role visibility of objects which are part of collections."""
roles = ', '.join([str(r) for r in collection.roles])
body = {
'query': {'term': {'collection_id': collection.id}},
'script': {
'inline': 'ctx._source.roles = [%s]' % roles
}
}
es.update_by_query(index=entity_index(),
body=body,
wait_for_completion=False)
def upgrade_search():
"""Add any missing properties to the index mappings."""
INDEXES = [
(collection_index(), COLLECTION_MAPPING),
(entity_index(), ENTITY_MAPPING),
(record_index(), RECORD_MAPPING),
]
for (index, mapping) in INDEXES:
log.info("Creating index: %s", index)
es.indices.create(index, ignore=[404, 400])
es.indices.put_mapping(index=index, doc_type='doc', body=mapping)
es.indices.open(index=index, ignore=[400, 404])
es.indices.refresh(index=index)
def get_collection_stats(collection_id):
"""Compute some statistics on the content of a collection."""
query = {
'size': 0,
'query': {
'term': {
'collection_id': collection_id
}
},
'aggs': {
'schema': {
'terms': {
'field': 'schema',
'size': 1000
}
},
'countries': {
'terms': {
'field': 'countries',
'size': 250
}
},
'languages': {
'terms': {
'field': 'languages',
'size': 100
}
},
}
}
result = es.search(index=entity_index(),
doc_type=entity_type(),
body=query)
aggregations = result.get('aggregations')
data = {'schemata': {}, 'count': result['hits']['total']}
# expose entities by schema count.
for schema in aggregations['schema']['buckets']:
data['schemata'][schema['key']] = schema['doc_count']
# if no countries or langs are given, take the most common from the data.
if not len(data.get('countries', [])):
countries = aggregations['countries']['buckets']
data['countries'] = [c['key'] for c in countries]
if not len(data.get('languages', [])):
countries = aggregations['languages']['buckets']
data['languages'] = [c['key'] for c in countries]
# pprint(data)
return data
def update_collection_roles(collection):
"""Update the role visibility of objects which are part of collections."""
roles = ', '.join([str(r) for r in collection.roles])
body = {
'query': {
'term': {
'collection_id': collection.id
}
},
'script': {
'inline': 'ctx._source.roles = [%s]' % roles
}
}
query_update(entity_index(), body)
def setUp(self):
if not hasattr(TestCase, '_global_test_state'):
TestCase._global_test_state = True
delete_index()
upgrade_search()
else:
indexes = [collection_index(), entity_index(), record_index()]
es.delete_by_query(index=indexes,
body={'query': {
'match_all': {}
}},
refresh=True,
conflicts='proceed')
destroy_db()
db.create_all()
create_system_roles()
def upgrade_search():
"""Add any missing properties to the index mappings."""
INDEXES = [
(collections_index(), COLLECTION_MAPPING),
(entity_index(), ENTITY_MAPPING),
(record_index(), RECORD_MAPPING),
]
for (index, mapping) in INDEXES:
log.info("Creating index: %s", index)
settings = deepcopy(INDEX_SETTINGS)
if index == record_index():
# optimise records for bulk write
settings['index']['refresh_interval'] = '-1'
es.indices.create(index, body=settings, ignore=[404, 400])
es.indices.put_mapping(index=index, doc_type='doc', body=mapping)
es.indices.open(index=index, ignore=[400, 404])
es.indices.refresh(index=index, ignore=[400, 404])
es.indices.clear_cache(index=index, ignore=[400, 404])
def _index_updates(collection, entities):
"""Look up existing index documents and generate an updated form.
This is necessary to make the index accumulative, i.e. if an entity or link
gets indexed twice with different field values, it'll add up the different
field values into a single record. This is to avoid overwriting the
document and losing field values. An alternative solution would be to
implement this in Groovy on the ES.
"""
common = {
'collection_id': collection.id,
'updated_at': datetime.utcnow(),
'bulk': True
}
timestamps = {}
if not len(entities):
return []
for result in iter_entities_by_ids(list(entities.keys())):
if int(result.get('collection_id')) != collection.id:
raise RuntimeError("Key collision between collections.")
existing = model.get_proxy(result)
entities[existing.id].merge(existing)
timestamps[existing.id] = result.get('created_at')
actions = []
for entity_id, entity in entities.items():
context = dict(common)
context['created_at'] = timestamps.get(entity.id)
entity = finalize_index(entity, context, [])
# pprint(entity)
actions.append({
'_id': entity_id,
'_index': entity_index(),
'_type': 'doc',
'_source': entity
})
return actions
def iter_entities_by_ids(ids, authz=None):
"""Iterate over unpacked entities based on a search for the given
entity IDs."""
for i in range(0, len(ids), MAX_PAGE):
chunk = ids[i:i + MAX_PAGE]
if not len(chunk):
return
query = bool_query()
query['bool']['filter'].append({'ids': {'values': chunk}})
if authz is not None:
query['bool']['filter'].append(authz_query(authz))
includes = ['schema', 'properties', 'collection_id', 'created_at']
query = {
'query': query,
'_source': {'includes': includes},
'size': min(MAX_PAGE, len(chunk) * 2)
}
result = search_safe(index=entity_index(),
body=query,
request_cache=False)
for doc in result.get('hits', {}).get('hits', []):
entity = unpack_result(doc)
if entity is not None:
yield entity
def all_indexes():
return [collection_index(), entity_index(), record_index()]
def clear_index():
indexes = [collection_index(), entity_index(), record_index()]
q = {'query': {'match_all': {}}}
es.delete_by_query(index=indexes, body=q, refresh=True)
def index_document(document):
if document.status == Document.STATUS_PENDING:
return
log.info("Index document [%s]: %s", document.id, document.title)
schema = model.get(Document.SCHEMA)
data = {
'schema': schema.name,
'schemata': schema.names,
'collection_id': document.collection_id,
'roles': document.collection.roles,
'type': document.type,
'status': document.status,
'content_hash': document.content_hash,
'foreign_id': document.foreign_id,
'error_message': document.error_message,
'uploader_id': document.uploader_id,
'created_at': document.created_at,
'updated_at': document.updated_at,
'title': document.title,
'name': document.title,
'summary': document.summary,
'author': document.author,
'file_size': document.file_size,
'file_name': document.file_title,
'source_url': document.source_url,
'languages': document.languages,
'countries': document.countries,
'keywords': document.keywords,
'dates': document.dates,
'extension': document.extension,
'encoding': document.encoding,
'mime_type': document.mime_type,
'pdf_version': document.pdf_version,
'columns': document.columns,
'children': document.children.count(),
'text': index_form(document.texts)
}
if document.parent_id is not None:
data['parent'] = {
'id': document.parent_id,
'type': document.parent.type,
'title': document.parent.title,
}
q = db.session.query(DocumentTag)
q = q.filter(DocumentTag.document_id == document.id)
for tag in q.yield_per(5000):
field = TAG_FIELDS.get(tag.type)
if field is None:
log.warning("Cannot index document tag: %r", tag)
continue
if field not in data:
data[field] = []
data[field].append(tag.text)
index_names(data)
data = clean_dict(data)
# pprint(data)
es.index(index=entity_index(),
doc_type=entity_type(),
body=data,
id=document.id)
data['id'] = document.id
return data
def flush_index():
"""Run a refresh to apply all indexing changes."""
es.indices.refresh(index=collection_index())
es.indices.refresh(index=entity_index())
es.indices.refresh(index=record_index())
def delete_index():
es.indices.delete(collection_index(), ignore=[404, 400])
es.indices.delete(entity_index(), ignore=[404, 400])
es.indices.delete(record_index(), ignore=[404, 400])
