CODEWITHSUNDEEP
elasticsearch
gilly3
gilly3

Reputation: 91497

Filter on fields with a value of "no"

I am unable to filter my elasticsearch query to find items with a value of "no". I'm using version 0.90.

Create an index with the following items:

{ "foo": "abc", "bar": "yes" }
{ "foo": "def", "bar": "no" }
{ "foo": "ghi", "bar": "maybe" }

Now, Try some term queries on bar:

{ "query": { "term": { "bar": "yes" } } }
// Hits: 1

{ "query": { "term": { "bar": "maybe" } } }
// Hits: 1

{ "query": { "term": { "bar": "no" } } }
// Hits: 0                                   <-- What??

There are no hits when I query with a value of "no".

Look at the facets:

{ "facets": { "bar": { "terms": { "field": "bar" } } } }

Result:

{
   ...
   "facets": {
      "bar": {
         "_type": "terms",
         "missing": 1,
         "total": 2,
         "other": 0,
         "terms": [
            { "term": "yes", "count": 1 },
            { "term": "maybe", "count": 1 }
         ]
      }
   }
}

There isn't even a facet returned for "no" values. What is going on? How do I find items in my index with a value of "no"?

I just downloaded the latest version (1.0.1) and I can see that this is fixed. But, I'm on 0.90.3 and the behavior is the same in 0.90.12. Is there a way to get this to work in 0.90.*?

Upvotes: 0

Views: 83

Answers (1)

Akshay
Akshay

Reputation: 3411

This is expected behaviour using the standard analyzer in 0.90. The standard analyzer, which is used by default when you index a string, applies a stop word token filter. "No" appears on the list of stop words so is removed from the list of terms that get indexed. This is why a search for "no" will return nothing.

The reason this behaviour is different in 1.0 is that the benefit of stop words was questioned and removed from the default behaviour. http://www.elasticsearch.org/guide/en/elasticsearch/reference/current/_stopwords.html

If you want to see the same behaviour you could create your own analyzer or just use the keyword analyzer instead. This will just convert the string to lowercase, not split it and not apply the stop word filter to the terms.

There's a good blog post on the subject here: http://www.elasticsearch.org/blog/stop-stopping-stop-words-a-look-at-common-terms-query/ and another Stackoverflow question here: elasticsearch: how to index terms which are stopwords only?

Upvotes: 1

Related Questions