Support fuzziness

The elasticsearch_connector module currently performs full text searches using the query_string query. This query type makes it tricky to add fuzziness to, as you have to include a tilde against each keyword. For example, let's say we want fuzziness, our search term is the poorly spelled "qiuck borwn fxo".

With a Match Query we could do:

GET _search
{
  "query": {
    "match": {
      "body": {
        "query": "qiuck borwn fxo",
        "fuzziness": "auto"
      }
    }
  }
}

And that would give us the result we want, just by adding fuzziness: auto. With a Query String query, we would need to do:

GET _search
{
  "from": 0,
  "size": 5,
  "query": {
    "query_string": {
      "query": "qiuck~ borwn~ fxo~",
      "fields": [
        "body^1"
      ],
      "fuzziness": "auto"
    }
  }
}

Here note how we need to add the tilde to each of the search keys. Which isn't horrible, but leads me to ask, why does the module make everything into a query_string query.

I'm very new to Elastic search, so first hoping to learn more, before jumping in and making changes like this.

Humm, I think I'm coming to an understanding that elasticsearch_connector uses query_string, b/c the Lucene query language offers the flexibility needed to handle the QueryInterface::getKeys() definition (which can include conjunctions and negations).

It's probably better to fit fuzziness support into the existing query_string queries then?

Wonder if this would be a nice addition:

https://github.com/makinacorpus/php-lucene-query

Yeah to be clear the library I linked is just a small API for building up the query_string query parameter specced here: https://www.elastic.co/guide/en/elasticsearch/reference/5.6/query-dsl-qu.... It's a structured approach to building these strings vs. the string concatenation style approach already in the elasticsearch_connector module. It supports all the conjunctions, fuzziness, negations, etc.

I threw together a WIP:

https://github.com/nodespark/elasticsearch_connector/pull/19

This is rough for a couple reasons, first being that I needed to fork the library to get un-distanced fuzziness for terms.

Let me know what you think?

The upstream library change was merged and a new release put out. I've updated the PR for this ticket. I think it's ready to rock.

https://github.com/nodespark/elasticsearch_connector/pull/19