elasticsearch double quotes not working

Question

I'm trying to search for a specific phrase with elasticsearch i.e. "foo bar". My query is similar to the one below:

curl -X GET "http://localhost:9200/objects/object/_search" -d '{
  "query": {
    "bool": {
      "must": [
        {
          "query_string": {
            "query": "\"foo bar\"",
            "default_field": "_all"
          }
        }
      ]
    }
  }
}'

I've got two objects in my index, looking similar to this:

{
  sub_sections: [
    {
      name: "foo"
    },
    {
      name: "bar"
    }
  ]
}

and

{
  sub_sections: [
    {
      name: "foo bar"
    }
  ]
}

I'm expecting only the last object, matching the phrase, to be returned when I use double quotes in the query. However, both objects are always returned. I've tried options like auto_generate_phrase_queries ++ with no luck so far.

Is this the expected behaviour? How could I return only the documents with exact phrase matches?

Answer 1

By default, both examples are indexed exactly the same way. When multiple instances of the same field are indexed they are indexed sequentially with no gap between the last term of one field and the first term of the following field. That's why phrase searches span field instances. You can increase the gap between fields by setting position_offset_gap in the mapping to a non-zero value. See https://gist.github.com/4420794 for a quick demo.