CODEWITHSUNDEEP
elasticsearch
Zane Claes
Zane Claes

Reputation: 14954

ElasticSearch "Match" with FullText Matches

I'm working with ElasticSearch. When I do this query:

{query: "blackberry -q10"}

I get exactly what I want (all results which have reference to BlackBerry but not Q10).

However, I want to restrict the fields which are searched to just the "title" field. Eg, the _source documents have titles, body, tags, etc. and I only want to search the title. The ElasticSearch "Match" seems right for me...

{query: {match: {title: "blackberry -q10"}}}

While this succeeds in only searching the title, it still returns results with have Q10 in the title, unlike the search above.

I'm looking at the match documentation but can't seem to figure it out.

Thanks!

Upvotes: 0

Views: 4917

Answers (2)

Zach
Zach

Reputation: 9721

The Match query doesn't use negation syntax like that. E.g you can't use a "minus" to negate a term. It will be parsed as a hyphen by the default search analyzer.

I would use a filtered query in this case. You could add the negation in a query...but a filter will be much faster.

{
  "filtered":{
     "query":{
        "match":{
           "title":"blackberry"
        }
     },
     "filter":{
        "bool":{
           "must_not":{
               "term":{
                  "title":"q10"
              }
           }
        }
     }
  }
}

Note, you may need to change the term filter, depending on how you analyzed the field at index time.

EDIT: Based on your comment below, if you really want to keep the ability to do negations "inline", you would use the field query (a more specific version of query_string, which would also work). This query uses Lucene syntax, which allows inline negation

{
   "field" : {
       "title" : "blackberry -q10"
   }
}

The reason query_string and it's derivatives are not recommended is because it is easy to shoot yourself in the foot. Or rather, it's easy for your users to shoot your server in the face. Query_string demands proper syntax and will simply die if the users enter it incorrectly. It also allows your users to make some horrible inefficient queries, usually through wildcards

Upvotes: 1

Gary Foster
Gary Foster

Reputation: 31

You want to match all the titles that have "blackberry" AND do not have have q10, not all the titles that have "blackberry" OR do not have q10.

The default boolean operator for a match is (in most cases) OR. Try adding an "operator": "and" clause to your query instead.

Upvotes: 0

Related Questions