arangodb-river-elasticsearch -- _mapping mismatch - arangodb

Question

When creating rivers for composed objects, the resulting _mapping is set with the complete nested object definition rather than String field. This causes the data import to fail because the object references are not "dereferenced".
E.g.
collection1: {name: "test", items: [collection2/123, collection2/124] }
collection1: {somefield: "test"}
The resulting _mapping after creating the river for those collections within a single index is:
collection1: {name: String, items: { properties: { somefield: String } } }.
Importing data fails with the following error:
org.elasticsearch.index.mapper.MapperParsingException: object mapping [items] trying to serialize a value with no field associated with it, current value [collection1/123]
How can I either tell the arango db river to dereference the nested objects or set the mapping properly to work with references?

Answer 1

Rivers are now deprecated. I created a mixin for elasticsearch which updates the index when I save/update/delete objects (through my custom ODM).
Simply make yourself a wrapper around your data access layer with high level functions that also updates the ES index.
For example:
class Base(ArangoBase, es.Base):
def save(self):
ret = ArangoBase.save(self)
es.Base.save_es(self)
return ret
def update(self):
ret = ArangoBase.update(self)
es.Base.save_es(self)
return ret
def delete(self):
ret = ArangoBase.delete(self)
es.Base.delete_es(self)
return ret
from elasticsearch import Elasticsearch
class Base(object):
_es = None
_es_index = 'chopchop'
_es_type = None
def save_es(self):
self._es.index(index=self._es_index, doc_type=self._es_type, body=self._doc(), id=self.id)
def delete_es(self):
self._es.delete(index=self._es_index, doc_type=self._es_type, id=self.id)