Question

I’m having trouble getting a edgengram query to behave properly. I have one record “blue grass” with an edgengram minimum of 2. A query string of “blv” however returns “blue grass” although it shouldn’t.

curl -X POST http://localhost:9200/test -d '{ 
    "mappings": { 
        "product/fragrance": { 
            "properties": { 
                "name_query": { 
                    "index_analyzer": "query_index_analyzer", 
                    "search_anaylzer": "query_search_analyzer", 
                    "as": {}, 
                    "type": "string" 
                } 
            } 
        } 
    }, 
    "settings": { 
        "analysis": { 
            "filter": { 
                "query_edgengram": { 
                    "type": "edgeNGram", 
                    "min_gram": 2, 
                    "max_gram": 20, 
                    "side": "front" 
                } 
            }, 
            "analyzer": { 
                "query_index_analyzer": { 
                    "tokenizer": "lowercase", 
                    "filter": ["asciifolding", "query_edgengram"] 
                }, 
                "query_search_analyzer": { 
                    "tokenizer": "lowercase", 
                    "filter": ["asciifolding"] 
                } 
            } 
        } 
    } 
}' 

curl -X POST "http://localhost:9200/test/product%2Ffragrance/1" -d '{ 
    "name_query": "blue grass" 
}' 

curl -X GET "http://localhost:9200/test/product%2Ffragrance/_search?load=true&pretty=true" -d '{ 
    "query": { 
        "bool": { 
            "must": [{ 
                "query_string": { 
                    "query": "blv", 
                    "fields": ["name_query"], 
                    "default_operator": "OR" 
                } 
            }] 
        } 
    } 
}' 

For some reason, I get a result from that. Can anyone explain why? Thanks. What I want to happen is “blv” shouldn’t be returning “blue grass” although “bl” should. I’ve used the analyze API and see “blue grass” being broken down to “bl”, “blu”, “blue”, “gr”, “gra”, “gras”, “grass” but “blv” doesn’t match any of those.

Answer 1

As David told you in his answer some elasticsearch queries are analyzed. Usually you don’t want to apply ngrams to your queries, but you seem to already know that given your mapping. In fact, the reason why your search analyzer without ngrams is not taken into account is a typo: search_anaylzer instead of search_analyzer. That’s why your query becomes bl and blv, and bl does match the returned document.