How to search nested objects with Elasticsearch

You don't necessarily need to nest video; you can map it as a normal field. Which means it will store

'video:title': "This is a test title for a video3",
'video:description':"This is my video description3",
'video:url':"/url_of_video3"

And you can search for video.title:'test'.

As far as I get it, nested fields are useful when you have multiple nested items, and you want to make a query only for the nested items. For example, having this data

[{
    id:4635,
    description:"This is a test description",
    author:"John",
    author_id:51421,
    video: [
      {
        title:"This is a test title for a video",
        description:"This is my video description",
        url:"/url_of_video"
      },
      {
        title:"This is an example title for a video",
        description:"This is my video description2",
        url:"/url_of_video2"
      }
    ]
},
{
    id:4637,
    description:"This is a test description3",
    author:"John",
    author_id:51421,
    video: [
      {
        title:"This is a test title for a video3",
        description:"This is my video description3",
        url:"/url_of_video3"
      }
    ]
}]

If you would search for video.title: 'test' and video.description: 'description2', and video was not nested, it will give you a fake result (because test is in the first video and description2 in the second, but in all the video field you have both).

In this case, if you map video as nested, it will remember each video as a separate entity and will search for individual videos that fit those conditions, so for video.title: 'test' and video.description: 'description2' it will return nothing, for video.title: 'example' and video.description: 'description2' it will return one result.

If you want to put it in Rest API URL format

/_search?pretty&q=video.title:*test*

To give a more generic answer:

• use objects (no nested) if you search one field at a time. That's because internally, fields are flattened, like so:

• use nested if you need to search in multiple fields (e.g. title:test AND description:my), because objects don't care about boundaries. Meanwhile, nested fields create separate Lucene documents under the hood that are quickly joined via Lucene's BlockJoin:

• use parent-child relationships (between different Elasticsearch documents) if you search in multiple fields and update child documents often (because updates of nested documents will update the whole ensemble). Basically, if you want to trade query performance for update performance, since the query runs in two steps under the hood:

Note: the drawings above are from Sematext's Elasticsearch training classes (disclosure: I deliver these classes).

You can use the .keyword suffix if the name of the nested object is unique:

{
        'query': {
            'term': {
                'title.keyword': "This is a test title for a video"               
            }
        }
}

Which should match your first example entry. Note that the video object name is not specified anywhere; this matches on all objects that have a title sub-object.

schema is:

  private schema = {
    id: {
      type: 'integer',
    },
    name: {
      type: 'text',
    },
    tags: {
      type: 'nested',
      properties: {
        id: {
          type: 'integer',
        },
        name: {
          type: 'keyword',
          normalizer: 'useLowercase',
        },
      },
    },
  }

document structure is

id: 38938
name: "summer fruits"
tags:[
   {
    id : 73
    name: "Grapes"
   },
  {
    id : 74
    name: "Pineapple"
   }
]

search query:

    const { tags } = req.body;

    const { body } = await elasticWrapper.client.search({
        index: ElasticIndexs.Fruits,
        pretty: true,
        filter_path: 'hits.hits._source*',
        body: {
          query: {
            bool: {
              must: tags.map((ele: { name: string }) => {
                return {
                  nested: {
                    path: 'tags',
                    query: {
                      match: {
                        'tags.name': ele.name,
                      },
                    },
                  },
                };
              }),
            },
          },
        },
      });