[SOLVED] How to filter data by two clauses in Elasticsearch?

How to filter data by two clauses in Elasticsearch?

I am trying to filter a data set based on two conditions. I want my results to return the record if it satisfies either the condition 1 (where record contains A and B), or condition 2 (where record contains A and C), with wildcards included.

For example, if I have this table here (and Z is just a random filler character):

TABLE 1
AZZZ ZBZ
ZZZZ ZAZ
ZZAZ ZZB
CZZZ ZZA
AZZZ ZCZ
ZZZZ ZZZ
ZZCZ ZZZ

I would want my return result to be this:

TABLE 1 (Returned)
AZZZ ZBZ
ZZAZ ZZB
CZZZ ZZA
AZZZ ZCZ

I've been trying codes along these lines, but can not seem to get the desired results:

{  
   "query":{  
      "query_string":{  
         "analyze_wildcard":true,
         "default filter":"*",
         "query":"A B"
      }
   }
}

Any thoughts?

Solution

Wildcard queries allow you to match parts of a string on not analyzed (keyword) fields, for example:
```
POST my_index/_search
{
    "query": {
        "wildcard" : { "my_field" : "*A*" }
    }
}
```

Bool queries are your friend for combining things. I think it would be easier to restructure this to the following which should be equivalent:

POST my_index/_search
{
  "query": {
    "bool" : {
      "must" : {
        "wildcard" : { "my_field" : "*A*" }
      },
      "should" : [
        { "wildcard" : { "my_field" : "*B*" } },
        { "wildcard" : { "my_field" : "*C*" } }
      ],
      "minimum_should_match" : 1
    }
  }
}

But this won't be very performant. It's a bit hard to abstract what kind of data you have there and what you want to achieve, but maybe there is a better way to achieve that? The problem feels very much modelled around the features of relational databases whereas a search engine has a lot more options.