CODEWITHSUNDEEP
jsonelasticsearchkibana
tkja
tkja

Reputation: 2030

Delete / add nested objects in Elastic search

I cannot find examples in the Elastic manual on nested objects on how to modify fields and nested objects of documents using RESTful commands in Kibana Sense. I am looking for something similar to Solrs atomic updates here, which allow to update specific fields of documents.

How do RESTful commands in Kibana Sense look like that accomplish this? The only related info in the manual I can find is on Partial Updates to Documents, but I do not know how this can be applied for this use case.

For example, straight from the Elastic docs:

PUT my_index
{
"mappings": {
    "my_type": {
    "properties": {
        "user": {
        "type": "nested" 
        }
    }
    }
}
}

PUT my_index/my_type/1
{
"group" : "fans",
"user" : [
    {
    "first" : "John",
    "last" :  "Smith"
    },
    {
    "first" : "Alice",
    "last" :  "White"
    }
]
}

How can I delete an entry in the nested object, so that the document "1" looks like: 

{
"group" : "fans",
"user" : [
    {
    "first" : "John",
    "last" :  "Smith"
    }
]
}

How can I add an entry in the nested object, so that the document "1" looks like: 

{
"group" : "fans",
"user" : [
    {
    "first" : "John",
    "last" :  "Smith"
    },
    {
    "first" : "Alice",
    "last" :  "White"
    },
    {
    "first" : "Peter",
    "last" :  "Parker"
    }
]
}

Upvotes: 7

Views: 7152

Answers (3)

Reid
Reid

Reputation: 4544

For this specific use case, you must use a scripted update. In javascript the call will look something like:

    const documentUpdateInstructions = {
      index: "index-name",
      id: "document-id",
      body: {
        script: {
          lang: "painless",
          source: `ctx._source.myNestedObject.removeIf(object -> object.username == params.username);`,
          params: {
            username: "my_username"
          },
        },
      },
    };
    await client.update(documentUpdateInstructions);

This takes a document in the form of

document._source = {
  ...
  "myNestedObject": [
    {
      "username": "my_username",
      ...
    },
    {
      "username": "not_my_username",
      ...
    }
  ]
}

and deletes the object inside myNestedObject who's username matches the username provided (in this case my_username). The resulting document will be:

document._source = {
  ...
  "myNestedObject": [
    {
      "username": "not_my_username",
      ...
    }
  ]
}

Upvotes: 0

Juha Vehnia
Juha Vehnia

Reputation: 1398

You will have to use scripted updates unless you want to fetch all nested objects then add / remove items and re-index them all which is the previous answer proposed. However if you have a lot of nested documents you should be doing partial updates / additions and deletes. It is much quicker from data transfer and indexing point of view.

Here is a good article how to do scripted updates in general:

https://iridakos.com/programming/2019/05/02/add-update-delete-elasticsearch-nested-objects

Upvotes: 3

Phil
Phil

Reputation: 1266

Unless I misunderstand your ask, you just post the updated document version to the same document id each time you want.

To delete a nested document (or any field):

PUT my_index/my_type/1
{
  "group" : "fans",
  "user" : [
    {
      "first" : "Alice",
      "last" :  "White"
    }
  ]
}

To add a user, add it to the list: 

PUT my_index/my_type/1
{
  "group" : "fans",
  "user" : [
    {
      "first" : "Alice",
      "last" :  "White"
    },
    {
      "first" : "Peter",
      "last" :  "Parker"
    }
  ]
}

Note: Documents in elasticsearch are immutable. Making a change to a single field causes the entire document to be re-indexed. Nested documents are always re-indexed with the parent document so if you change a field in the parent the nested document is also re-indexed. This can be a performance issue if the nested documents are large and the parents have frequent changes.

Upvotes: 0

Related Questions