/_termvectors sometimes throws NPE

[shaie]

Spinoff from #21906. I still hope to find the root cause of the NPE, and hopefully this isn't a Guice thing. For reference, here's the stacktrace:

[2016-12-01T11:59:13,208][WARN ][r.suppressed             ] path: /tv_bug/doc/_termvectors, params: {pretty=true, preference=_shards:1, index=tv_bug, type=doc}
org.elasticsearch.transport.RemoteTransportException: [C-N1Zbn][127.0.0.1:9300][indices:data/read/tv[s]]
Caused by: java.lang.NullPointerException
        at org.elasticsearch.action.termvectors.TransportTermVectorsAction.shardOperation(TransportTermVectorsAction.java:79) ~[elasticsearch-5.0.1.jar:5.0.1]
        at org.elasticsearch.action.termvectors.TransportTermVectorsAction.shardOperation(TransportTermVectorsAction.java:42) ~[elasticsearch-5.0.1.jar:5.0.1]
        at org.elasticsearch.action.support.single.shard.TransportSingleShardAction$ShardTransportHandler.messageReceived(TransportSingleShardAction.java:293) ~[elasticsearch-5.0.1.jar:5.0.1]
        at org.elasticsearch.action.support.single.shard.TransportSingleShardAction$ShardTransportHandler.messageReceived(TransportSingleShardAction.java:286) ~[elasticsearch-5.0.1.jar:5.0.1]
        at org.elasticsearch.transport.TransportRequestHandler.messageReceived(TransportRequestHandler.java:33) ~[elasticsearch-5.0.1.jar:5.0.1]
        at org.elasticsearch.transport.RequestHandlerRegistry.processMessageReceived(RequestHandlerRegistry.java:69) ~[elasticsearch-5.0.1.jar:5.0.1]
        at org.elasticsearch.transport.TransportService$6.doRun(TransportService.java:548) [elasticsearch-5.0.1.jar:5.0.1]
        at org.elasticsearch.common.util.concurrent.ThreadContext$ContextPreservingAbstractRunnable.doRun(ThreadContext.java:520) [elasticsearch-5.0.1.jar:5.0.1]
        at org.elasticsearch.common.util.concurrent.AbstractRunnable.run(AbstractRunnable.java:37) [elasticsearch-5.0.1.jar:5.0.1]
        at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142) [?:1.8.0_92]
        at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617) [?:1.8.0_92]
        at java.lang.Thread.run(Thread.java:745) [?:1.8.0_92]

And here's the REST response

{
  "error" : {
    "root_cause" : [
      {
        "type" : "remote_transport_exception",
        "reason" : "[C-N1Zbn][127.0.0.1:9300][indices:data/read/tv[s]]"
      }
    ],
    "type" : "null_pointer_exception",
    "reason" : null
  },
  "status" : 500
}

[shaie]

I created a PR to fix the problem: #21959.

The problem is that when you use the _termvectors API with an artificial document and specify the preference parameter, the code used the auto-generated artificial doc's ID and in some cases the shard that was picked for routing per the ID did not match the shardId set in the preference, which resulted in the NPE reported above.

The fix detects that the TV for an artificial document is requested, and if routing isn't explicitly specified, it ignores the document's ID and uses either one of the shards, or the one specified in the preference.

I also added an IT which verifies the behavior; it fails without the fix.