bool queries and filter context

Key points

• must and should run in query context: they match and they contribute to the score.
• filter and must_not run in filter context - a yes or no test with no score computed.
• Filter clauses can be cached in the node query cache, sized at 10 percent of heap by default.
• That cache only holds segments with more than 10000 documents, so a tiny test index never shows a benefit.
• minimum_should_match defaults to 1 if the bool has no must or filter, and to 0 otherwise.
• Round dates in filters, since now-1d/d has a stable cache key and now-1d does not.

Example

GET /articles/_search
{
  "query": {
    "bool": {
      "must":     [ { "match": { "title": "elasticsearch" } } ],
      "filter":   [ { "term":  { "status": "published" } },
                    { "range": { "published_at": { "gte": "now-1y/d" } } } ],
      "must_not": [ { "term":  { "archived": true } } ],
      "should":   [ { "term":  { "tags": { "value": "tutorial", "boost": 2 } } } ]
    }
  }
}

Anything that decides membership rather than order belongs in filter, not must.