Scoring documents based on the intersection of a multi-value field

[bostanio (Jon Palmer)]

Hi All,

I'm trying to model a social network and query it with the search results order with 1st degree connections first 2nd degree connects next and then all other results.

"mappings": {
  "entity": {
    "properties": {
      "id": {
        "type": "long"
      },
      "name": {
        "type": "string"
      },
      "connections": {
        "type": "long"
      }
    }
  }
}

For a searching user with connections with ids 3,4,5 I would query it like this:

GET /_search
{
    "query": {
        "function_score": {
          "query": { "match": { "name": "bill" } },
          "functions": [
              {
                  "filter": { "terms": { "id": [3, 4, 5] } },
                  "weight": 1000
              },
              {
                  "filter": { "terms": { "connections": : [3, 4, 5] } },
                  "weight": 10
              }
          ],
          "score_mode": "sum"
        }
    }
} 

This works. However, I'd like documents that contain all three connections 3,4 & 5 to rank higher than those that contain only two and those higher than containing only one.

What would be the simplest approach to achieve that ranking?