Using wildcards in phrases with Elasticsearch Query String Query

Question

Using the wildcard operator, I can match terms starting with some value:

{
    "query": {
        "query_string" : {
            "query" : "subject:cell*"
        }
    }
}

The subject field here is a keyword field (non-analyzed). This works fine, but I cannot figure out how to find terms starting with, say, "cellular contr". Trying double quotes did not yield the expected results:

{
    "query": {
        "query_string" : {
            "query" : "subject:\"cellular contr*\""
        }
    }
}

Note: phrase search works fine with exact matches, just not with the wildcard. My guess is that the star is not interpreted as a wildcard operator inside the double quotes. Is that correct? And is there any other way to use the wildcard operator with a phrase?

Note: I have to use Query String Query, since the query is coming from user input.

(I know I could resort to regexp, but would prefer not to)

HM14 · Accepted Answer · 2017-05-26 05:09:43Z

11

In addition to the custom analyzer as pointed by Hemed, you need to do search as below -

{
    "query": {
        "query_string" : {
            "query" : "subject:cellular\\ contr*"
        }
    }
}

Found it after a lot of research and tries!

answered May 26, 2017 at 5:09

HM14

1891 gold badge2 silver badges9 bronze badges

Sign up to request clarification or add additional context in comments.

2 Comments

SeriousM Over a year ago

Thank you for this hint! "\\ " is telling elastic that it IS a space in the phrase and you can (or must) omit the quotes of the searched string.

togaurav Over a year ago

Thanks for the details, hard to point any post regarding this

Hemed · Accepted Answer · 2017-03-07 22:16:36Z

2

EDIT: Define custom analyzer for searching:-

settings:
   index:
     analysis:
       analyzer:
         keyword_analyzer:
           type: custom
           tokenizer: keyword
           filter:
             - lowercase

Found out that you need to use Prefix Query in this case, because Query String Query always segments on spaces during parsing.

But since you are using lowecase filter in this field and Prefix Query does not support analyzer, you would have to lowercase user input before appending it to the query.

New query becomes:-

   {
        "query": {
            "prefix" : {
                "subject" : "cellular contr"
            }
        }
    }

Alternatively, you can use Match Phrase Query which supports analyzer.

{
    "query": {
        "match_phrase_prefix" : {
            "subject" : {
                 "query" : "Cellular contr",
                  "analyzer" : "keyword_analyzer",
                  "max_expansions" : 100
                 }
              }
         }
    }

edited Mar 7, 2017 at 22:16

answered Mar 6, 2017 at 14:41

Hemed

939 bronze badges

5 Comments

danmichaelo Over a year ago

Would work if I didn't need a phrase query. But I don't want the query "cellular contr*" to match, say, "control of cellular material", which is why I'm indexing the field as keyword.

Hemed Over a year ago

I think I misunderstood your question. Is it correct that you want to search for query string "cellular contr" and get a result for "control of cellular material" in the field subject of type keyword which is not analyzed?

danmichaelo Over a year ago

No, if I search for "cellular contr", I want it to match "cellular control", but not "control of cellular material". So what I want is really subject:/cellular contr.*/, but without using regexp.

Hemed Over a year ago

Please see my edit. I think you would have to specify custom search analyzer when searching in this particular field.

danmichaelo Over a year ago

Seems like it gives exactly the same result. Probably still interpreted as (subject:cellular) OR (contr*).

paqash · Accepted Answer · 2017-03-05 23:11:05Z

0

Try this:

{
    "query": {
        "query_string" : {
            "query" : "subject:"cellular contr*",
            "split_on_whitespace" : false
        }
    }
}

answered Mar 5, 2017 at 23:11

paqash

2,3141 gold badge20 silver badges22 bronze badges

1 Comment

danmichaelo Over a year ago

The default value is false, so setting it explicitly to false shouldn't make any difference. But I tried just in case, and it's interpreted as "(subject:cellular) OR (contr*)", not what I want. I also tried playing with analyze_wildcard and auto_generate_phrase_queries without success (the documentation is a bit thin on exactly what they do, though)

Collectives™ on Stack Overflow

Using wildcards in phrases with Elasticsearch Query String Query

3 Answers 3

2 Comments

5 Comments

1 Comment

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

3 Answers 3

2 Comments

5 Comments

1 Comment

Your Answer

Sign up or log in

Post as a guest

Related