ElasticSearch - Search analyzed and not_analyzed multi_field in one query

Question

I have an indexed field 'properties.language' with the value 'en sv'. This field has a multi_field mapping that consists of two fields, one analyzed (name 'language'), and one that is not_analyzed (name '_exact').

How do I issue a single search query without having to query both 'properties.language' and 'properties.language._exact'?

Edit:

Here is my configuration:

Indexed data:

{
    "_index": "51ded0be98035",
    "_type": "user",
    "_id": "WUzwcwhTRbKur7J5ZY_hgA",
    "_version": 1,
    "_score": 1,
    "_source": {
        "properties": {
            "language":"en sv"
        }
    }
}

Mapping for type 'user':

{
    "user": {
        "properties": {
            "properties": {
                "properties": {
                    "language": {
                        "type": "multi_field",
                        "fields": {
                            "language": {
                                "type": "string",
                                "analyzer": "standard",
                                "index": "analyzed"
                            },
                            "_exact": {
                                "type": "string",
                                "index": "not_analyzed"
                            }
                        }
                    }
                }
            }
        }
    }
}

Search query:

{
    "query": {
        "filtered": {
            "query": {
                "match_all": {}
            },
            "filter": {
                "bool": {
                    "must": [{
                        "or": [{
                            "term": {
                                "properties.language": "en sv"
                            }
                        }, {
                            "term": {
                                "properties.language._exact": "en sv"
                            }
                        }]
                    }]
                }
            }
        }
    }
}

Answer 1

Consider indexing the language field using Elasticsearch builtin multi-valued fields (ie. arrays) instead: http://www.elasticsearch.org/guide/reference/mapping/array-type/. As you currently do, set index to not_analyzed.

When indexing your data, instead of a single value 'en sv', pass instead ['en', 'sv'], and ES will take care of the rest.

For querying, this gives you the ability to do the following to find items with both en and sv:

{
    "query": {
        "filtered": {
            "query": {
                "match_all": {}
            },
            "filter": {
                "bool": {
                    "must": [{
                        "term": {
                            "properties.language": "en"
                        }
                    }, {
                        "term": {
                            "properties.language": "sv"
                        }
                    }]
                }
            }
        }
    }
}

Or even better, find greater brevity/flexibility using the terms query/filter instead of term: http://www.elasticsearch.org/guide/reference/query-dsl/terms-query/