Elasticsearch bool search matching incorrectly

Question

So I have an object with an Id field which is populated by a Guid. I'm doing an elasticsearch query with a "Must" clause to match a specific Id in that field. The issue is that elasticsearch is returning a result which does not match the Guid I'm providing exactly. I have noticed that the Guid I'm providing and one of the results that Elasticsearch is returning share the same digits in one particular part of the Guid.

Here is my query source (I'm using the Elasticsearch head console):

{
 query: 
 {
  bool: 
  {
    must: [
    {
      text: 
      {
        couchbaseDocument.doc.Id: 5cd1cde9-1adc-4886-a463-7c8fa7966f26
      }
    }]
    must_not: [ ]
    should: [ ]
   }
 }
 from: 0
 size: 10
 sort: [ ]
 facets: { }
}

And it is returning two results. One with ID of

5cd1cde9-1adc-4886-a463-7c8fa7966f26

and the other with ID of

34de3d35-5a27-4886-95e8-a2d6dcf253c2

As you can see, they both share the same middle term "-4886-". However, I would expect this query to only return a record if the record were an exact match, not a partial match. What am I doing wrong here?

score 5 · Accepted Answer · edited May 23 '17 at 12:05

5

The query is (probably) correct.

What you're almost certainly seeing is the work of the 'Standard Analyzer` which is used by default at index-time. This Analyzer will tokenize the input (split it into terms) on hyphen ('-') among other characters. That's why a match is found.

To remedy this, you want to set your couchbaseDocument.doc.Id field to not_analyzed

See: How to not-analyze in ElasticSearch? and the links from there into the official docs.

Mapping would be something like:

{
    "yourType" : {
        "properties" : {
            "couchbaseDocument.doc.Id" : {"type" : "string", "index" : "not_analyzed"},
        }
    }
}

edited May 23 '17 at 12:05

Community

1
1

answered Apr 02 '14 at 20:04

Geert-Jan

18,623
16
75
137

Ah this looks correct. Thank you. I have one other question, though. I'm currently using automatic mapping. Is it possible to simply specify this one field as not analyzed without having to define the entire object structure? – Joshua Dixon Apr 02 '14 at 20:14
2

You have to specify an object structure. However you only have to do it for this one field. All other (not defined fields) will get the default treatment. See updated answer (top of head, so please check) – Geert-Jan Apr 02 '14 at 20:19

Elasticsearch bool search matching incorrectly

1 Answers1