Default index analyzer in elasticsearch

Question

I am facing a problem with elasticsearch where I dont want my indexed term to be analyzed. But the elasticsearch has some default setting which is tokenizing it on space. Therefore my facet query is not returning the result I want.

I read that "index" : "not_analyzed" in properties of index type should work. But the problem is that I dont know my document structure before hand. I would be indexing random MySQL databases to elasticsearch without knowing the table structure.

How can I setup elasticsearch such that by default it uses "index" : "not_analyzed" until otherwise asked for. Thanks

PS: I am using java if I can directly use any API for it I would love it.

Answer 1

I'd use dynamic templates - it should do what you are looking for:

{
    "testtemplates" : {
        "dynamic_templates" : [
            {
                "template1" : {
                    "match" : "*",
                    "match_mapping_type" : "string",
                    "mapping" : {
                        "type" : "string",
                        "index" : "not_analyzed"
                    }
                }
            }
        ]
    }
}

More on this approach here:

https://www.elastic.co/guide/en/elasticsearch/guide/current/custom-dynamic-mapping.html#dynamic-templates

Important: If someone suggest this approach to solve the not_analyzed issue, it will not work! keyword analyzer does some analyzing on the data and convert the data to small letters.

eg Data: ElasticSearchRocks ==> Keyword Analyzer: elasticsearchrocks

Try it yourself with analyzing query and see it happening.

curl -XPUT localhost:9200/testindex -d '{
    "index" : {
        "analysis" : {
            "analyzer" : {
                "default" : {
                    "type" : "keyword"
                }
            }
       }
    }
}'

http://www.elasticsearch.org/guide/en/elasticsearch/reference/current/analysis-keyword-analyzer.html

Answer 2

添加index.analysis.analyzer.default.type: keyword在elasticsearch.yml 。