Match phrase prefix query

Use the match_phrase_prefix query to search for documents that contain the terms of a phrase in the order you specify. The last term in the phrase is interpreted as a prefix, so the query matches any phrase that begins with the preceding terms and continues with a word that starts with the last term. For example, chocolate chip c matches chocolate chip cookies and mint chocolate chip cake but not cookies with chocolate chip.

The match_phrase_prefix query works similarly to the match phrase query but creates a prefix query out of the last term in the query string. If the query string contains only one term, the query is a prefix query on that term.

Use the match_phrase_prefix query in the following scenarios:

• Finding documents that contain a known phrase when only the beginning of its last word is available.
• Adding basic search-as-you-type functionality to a text field without changing the mapping.

For differences between the match_phrase_prefix and the match_bool_prefix queries, see The match_bool_prefix and match_phrase_prefix queries.

The following example shows a basic match_phrase_prefix query:

GET _search
{
  "query": {
    "match_phrase_prefix": {
      "title": "the wind"
    }
  }
}

To pass additional parameters, you can use the expanded syntax:

GET _search
{
  "query": {
    "match_phrase_prefix": {
      "title": {
        "query": "the wind",
        "analyzer": "stop"
      }
    }
  }
}

Example

For example, consider an index with the following documents:

PUT testindex/_doc/1
{
  "title": "The wind rises"
}

PUT testindex/_doc/2
{
  "title": "Gone with the wind"
  
}

The following match_phrase_prefix query searches for the whole word wind, followed by a word that starts with ri:

GET testindex/_search
{
  "query": {
    "match_phrase_prefix": {
      "title": "wind ri"
    }
  }
}

The response contains the matching document:

{
  "took": 4,
  "timed_out": false,
  "_shards": {
    "total": 1,
    "successful": 1,
    "skipped": 0,
    "failed": 0
  },
  "hits": {
    "total": {
      "value": 1,
      "relation": "eq"
    },
    "max_score": 0.42264006,
    "hits": [
      {
        "_index": "testindex",
        "_id": "1",
        "_score": 0.42264006,
        "_source": {
          "title": "The wind rises"
        }
      }
    ]
  }
}

Parameters

The query accepts the name of the field (<field>) as a top-level parameter:

GET _search
{
  "query": {
    "match_phrase_prefix": {
      "<field>": {
        "query": "text to search for",
        ... 
      }
    }
  }
}

The <field> accepts the following parameters. All parameters except query are optional.

The match_phrase_prefix query does not support the fuzziness parameter.

Using the match phrase prefix query for autocomplete

The match_phrase_prefix query requires no special mapping, which makes it a quick way to add autocomplete to a search box. However, it can return incomplete results because of the way it expands the last term.

For the query chocolate chip c, OpenSearch builds a phrase query that requires chocolate to be followed by chip and then expands the prefix c to the first max_expansions terms that start with c in the field’s term dictionary. These terms are selected in alphabetical order, not by relevance, so if many terms that start with c sort before cookies, documents containing chocolate chip cookies are not returned.

In a search-as-you-type experience, this is usually acceptable because each additional letter narrows the prefix until the missing term appears. Increasing max_expansions returns more matches but makes the query more expensive to run.

For autocomplete that is both complete and efficient, use one of the following approaches, which prepare prefixes at index time:

• The completion suggester, which uses the completion field type to return suggestions quickly from an in-memory data structure.
• The search_as_you_type field type, which indexes shingles and edge n-grams of a text field so that prefix matches do not need to be expanded at query time.

For a comparison of autocomplete approaches, see Autocomplete.