per_field_analyzer in term vectors api should work even if field does not store term vectors

[brwe]

Tested with elasticsearch version 2.2

The purpose of per_field_analyzer is to get a term vector as created by an analyzer different from the one the field is indexed with. I would expect to get a term vector for a field even if it does not have term vectors stored. Instead I get an NPE:

POST test/doc/1
{
  "text": "I am a happy hippo"
}


GET  test/doc/1/_termvectors
{
  "per_field_analyzer": {
    "text": "keyword"
  }
}

If I set "term_vector": "yes", in the mapping first all works as expected:

PUT test 
{
  "mappings": {
    "doc": {
      "properties": {
        "text": {
          "term_vector": "yes",
          "type": "string"
        }
      }
    }
  }
}

POST test/doc/1
{
  "text": "I am a happy hippo"
}

GET  test/doc/1/_termvectors
{
  "per_field_analyzer": {
    "text": "keyword"
  }
}

results in (as expected):

{
  "_index": "test",
  "_type": "doc",
  "_id": "1",
  "_version": 1,
  "found": true,
  "took": 1,
  "term_vectors": {
    "text": {
      "field_statistics": {
        "sum_doc_freq": 5,
        "doc_count": 1,
        "sum_ttf": 5
      },
      "terms": {
        "I am a happy hippo": {
          "term_freq": 1,
          "tokens": [
            {
              "position": 0,
              "start_offset": 0,
              "end_offset": 18
            }
          ]
        }
      }
    }
  }
}

[andyb-elastic]

It looks like this works on master, no longer throwing an NPE as described

@elastic/es-search-aggs