ElasticSearch for terms with spaces

Question

We are running ElasticSearch and are having some issues when searching for terms that contain a space. A concrete example: there is an a person named JM Bruno, but when searching for this no results are returned. I vaguely remember searching for this exact term did return the result, but I can't reproduce that right now.

I tried adding a space as well as "\ " to my tokenizer pattern, without much luck. The ES settings are the following (using the Tire gem in a Ruby on Rails application)

module Search
def self.included base
base.send :include, Tire::Model::Search
base.send :include, Tire::Model::Callbacks

base.class_eval do
  settings  analysis: {
              filter: {
                ngram: {
                  type: 'nGram',
                  max_gram: 12,
                  min_gram: 3
                },
                url_stop: {
                  type: "stop",
                  stopwords: %w[http https]
                }
              },
              tokenizer: {
                url_email_tokenizer: {
                  pattern: '[^\w\-\.@]+',
                  type: 'pattern'
                }
              },
              analyzer: {
                url_analyzer: {
                  tokenizer: "url_email_tokenizer",
                  filter: %w[url_stop ngram],
                  type: "custom"
                },
                name_analyzer: {
                  tokenizer: 'url_email_tokenizer',
                  filter: 'ngram',
                  type: 'custom'
                }
              }
            }


    end
  end
end

We use these tokenizers to search for domain names and email addresses as well.

Answer 1

Try to run _analyze API with the analyzer you applied to your field.

curl -XGET 'localhost:9200/_analyze?analyzer=name_analyzer' -d 'JM Bruno'

You will see how Elasticsearch breaks your field content into tokens and why you can not search for it using a TermQuery. A TermQuery is not analyzed so it compares your query exactly as is with the inverted index.

Answer 2

Instead of term query this can be search by using

In Java by using advance rest client

query.must((QueryBuilders.matchQuery("name", searchMap.get("JM Bruno")).minimumShouldMatch("100%")));

in elastic search directly

GET /_search

{
    "query": {
        "match" : {
            "name" : {
                "query" : "JM Bruno",
                "cutoff_frequency" : 0.001
            }
        }
    }
}