Elasticsearch "starts with" first word in phrases

Question

I try to implement an A - Z navigation for my content with Elasticsearch. What I need, is displaying all results which begins with e.g. a,b,c,... etc.

I've tried:

"query": {
        "match_phrase_prefix" : {
        "title" : {
            "query" : "a"
        }
      }
    }

The query mentioned above also display results, where within the string a word begins with a. Example:

"title": "Apfelpfannkuchen",

"title": "Affogato",

"title": "Kalbsschnitzel an Aceto Balsamico",

I want to display only phrase where the FIRST word begins with a.

Here the mapping I use:

$params = array(
            'index' => 'my_index',
            'body' => array(
                'settings' => array(
                    'number_of_shards' => 1,
                    'index' => array(
                        'analysis' => array(
                            'filter' => array(
                                'nGram_filter' => array(
                                    'type' => 'nGram',
                                    'min_gram' => 2,
                                    'max_gram' => 20,
                                    'token_chars' => array('letter', 'digit', 'punctuation', 'symbol')
                                )
                            ),
                            'analyzer' => array(
                                'nGram_analyzer' => array(
                                    'type' => 'custom',
                                    'tokenizer' => 'whitespace',
                                    'filter' => array('lowercase', 'asciifolding', 'nGram_filter')
                                ),
                                'whitespace_analyzer' => array(
                                    'type' => 'custom',
                                    'tokenizer' => 'whitespace',
                                    'filter' => array('lowercase', 'asciifolding')
                                ),
                                'analyzer_startswith' => array(
                                    'tokenizer' => 'keyword',
                                    'filter' => 'lowercase'
                                )
                            )
                        )
                    )
                ),
                'mappings' => array(
                    'tags' => array(
                        '_all' => array(
                            'type' => 'string',
                            'index_analyzer' => 'nGram_analyzer',
                            'search_analyzer' => 'whitespace_analyzer'
                        ),
                        'properties' => array()

                    ),
                    'posts' => array(
                        '_all' => array(
                            'index_analyzer' => 'nGram_analyzer',
                            'search_analyzer' => 'whitespace_analyzer'
                        ),
                        'properties' => array(
                            'title' => array(
                                'type' => 'string',
                                'index_analyzer' => 'analyzer_startswith',
                                'search_analyzer' => 'analyzer_startswith'
                            )
                        )
                    )
                )
            )
        );

This looks like elasticsearch2, could you tag it as such ? – Christophe Roussy Sep 05 '16 at 12:39 — Christophe Roussy, Sep 05 '16 at 12:39

Roopendra · Accepted Answer · 2015-04-20T09:10:44.733

22

If you are using default mapping then it will not work for you.

You need to use keyword tokenizer and lowercase filter in mapping.

Mapping Will be :

{
    "settings": {
        "index": {
            "analysis": {
                "analyzer": {
                    "analyzer_startswith": {
                        "tokenizer": "keyword",
                        "filter": "lowercase"
                    }
                }
            }
        }
    },
    "mappings": {
        "test_index": {
            "properties": {
                "title": {
                    "search_analyzer": "analyzer_startswith",
                    "index_analyzer": "analyzer_startswith",
                    "type": "string"
                }
            }
        }
    }
}

Search query on test_index :

{
    "query": {
        "match_phrase_prefix": {
            "title": {
                "query": "a"
            }
        }
    }
}

It will return all post title starting with a

edited Apr 20 '15 at 09:10

answered Apr 20 '15 at 09:01

Roopendra

7,674
16
65
92

I ran into a problem: The query doesn't show more than 59 results. – alin Apr 20 '15 at 19:40
I got it: I need to use "max_expansions". e.g.: "max_expansions": 1000 – alin Apr 20 '15 at 19:57
2

Is there a way to achieve this functionality without using `match_phrase_prefix` or `prefix`? – advncd May 10 '17 at 19:55
I'm dealing with structured logs across many applications. I can't declare all properties upfront, or know their types. Is there no way to do this dynamically? – Billy Jake O'Connor Oct 29 '21 at 10:54

score 2 · Answer 2 · answered Jun 17 '20 at 21:28

2

Alternatively, can use span_near:

GET your_index/_search
{
  "query": {
    "span_first": {
      "match": {
        "span_term": {
          "your_field": "first_token"
        }
      },
      "end": 1
    }
  },
  "_source": "your_field"
}

answered Jun 17 '20 at 21:28

Alex Moore-Niemi

2,913
2
24
22

score 0 · Answer 3 · answered Jun 02 '19 at 22:03

I'm updating @Roopendra 's answer according to this gist. Thus, there was an update and in recent versions search and index initializers seem not to work, there were replaced only to initializers, also string needs to be replaced to text.

Thus, we have the following mapping file:

{
    "settings": {
        "index": {
            "analysis": {
                "analyzer": {
                    "analyzer_startswith": {
                        "tokenizer": "keyword",
                        "filter": "lowercase"
                    }
                }
            }
        }
    },
    "mappings": {
        "test_index": {
            "properties": {
                "title": {
                    "analyzer": "analyzer_startswith",
                    "type": "text"
                }
            }
        }
    }
}

With the following query:

{
    "query": {
        "match_phrase_prefix": {
            "title": {
                "query": "a",
                "max_expansions": 100
            }
        }
    }
}

I added max_expansions to the query because the default value seems to be 5 so I was getting erroneous results, in your case the value might be higher.

score 0 · Answer 4 · answered Jul 08 '22 at 09:54

0

You can do this by simply using the field with .keyword or .raw suffixes. For example to search for all the values starting with the letter 'a':

fieldName.keyword:a*

or

fieldName.raw:a*

More info on keyword vs text fields

answered Jul 08 '22 at 09:54

Swarna

186
1
5

Elasticsearch "starts with" first word in phrases

4 Answers4

Linked