CODEWITHSUNDEEP
elasticsearch
Michael_Scharf
Michael_Scharf

Reputation: 34548

How to address, delete or access child objects in ElasticSearch?

How are child objects addressed in elasticseacrch?

Suppose we create two parents and three children. Note that there are two children with id c2 but with different parents:

curl -XPUT localhost:9200/test/parent/p1 -d'{
  "name": "Parent 1"
}'

curl -XPUT localhost:9200/test/parent/p2 -d'{
  "name": "Parent 2"
}'

curl -XPOST localhost:9200/test/child/_mapping -d '{
  "child":{
    "_parent": {"type": "parent"}
  }
}'

curl -XPOST localhost:9200/test/child/c1?parent=p1 -d '{
   "child": "Parent 1 - Child 1"
}'

curl -XPOST localhost:9200/test/child/c2?parent=p1 -d '{
   "child": "Parent 1 - Child 2"
}'

curl -XPOST localhost:9200/test/child/c2?parent=p2 -d '{
   "child": "Parent 2 - Child 2"
}'

If we search the children, we see that there are two children with _id of c2

curl -XGET localhost:9200/test/_search

{
  "_shards": {
    "failed": 0, 
    "successful": 5, 
    "total": 5
  }, 
  "hits": {
    "hits": [
      {
        "_id": "c1", 
        "_index": "test", 
        "_score": 1.0, 
        "_source": {
          "child": "Parent 1 - Child 1"
        }, 
        "_type": "child"
      }, 
      {
        "_id": "c2", 
        "_index": "test", 
        "_score": 1.0, 
        "_source": {
          "child": "Parent 1 - Child 2"
        }, 
        "_type": "child"
      }, 
      {
        "_id": "c2", 
        "_index": "test", 
        "_score": 1.0, 
        "_source": {
          "child": "Parent 2 - Child 2"
        }, 
        "_type": "child"
      }
    ], 
    "max_score": 1.0, 
    "total": 3
  }, 
  "timed_out": false, 
  "took": 1
}

How do I address p1/c2? Without the parent child relationship, the _id can be used access, change or delete a child object. In my case I let elasticsearch create the id of the objects.

To access the child objects, the _id is not enough:

curl -XGET localhost:9200/test/child/c2

I have to specify the parent as well:

curl -XGET localhost:9200/test/child/c2?parent=p1

In my system it is worse, some objects I can directly access without the parent and others I cannot access. (Why???)

If I delete c2 (without parent!):

curl -XDELETE http://localhost:9200/test/child/c2

both children are deleted. To delete only one child I have to use ?parent=p1

curl -XDELETE http://localhost:9200/test/child/c2?parent=p1

Here are my questions.

Upvotes: 3

Views: 3362

Answers (1)

Rotem Hermon
Rotem Hermon

Reputation: 2147

Child documents are just ordinary documents in Elasticsearch, with an additional _parent field that points to a document in the parent type.
When accessing child documents, either when indexing or when getting, you need to specify the parent id in the request. This is because the parent id is actually used for routing of the child document (see for instance about routing -http://www.elasticsearch.org/guide/en/elasticsearch/reference/current/search.html#search-routing).
This means that the child document is sharded according to the parent id, so it resides on the same shard as the parent.

In your example above, what probably happened is that each of your c2 documents was created on a separate shard - one was sharded by its own id, and the other (where you specified the parent) according to the parent id.

This is important to understand so you won't have inconsistencies between index, get and search. So you need to remember to always pass the parent when you're working with child documents so they will be routed to the right shard.

About the document id - you need to treat it like all other documents. This means that it needs to be unique, you can't have 2 documents with the same id even if they have different parents.
You can either use the parent id as part of the child document id (as you suggested), or let ES generate a unique id if that makes sense in your use case. Document IDs that ES generates are unique, no matter the parent.

About getting the parent field back, you need to request it explicitly, it is not returned by default. (Request it using the fields parameter - http://www.elasticsearch.org/guide/en/elasticsearch/reference/current/docs-get.html#get-fields, or in search - https://www.elastic.co/guide/en/elasticsearch/reference/current/search-request-stored-fields.html).

Upvotes: 5

Related Questions