Jumped from 1.X to 2.X and noticed my tokenization changed

Here is an old entry:

}, {
  "_index" : "logs-2014-12-11",
  "_type" : "logs",
  "_id" : "505",
  "_score" : 1.0,
  "fields" : {
    "terms" : [ [ "smith/robert/principal/janitor" ] ]
  }
}, {

Here is the new entry:

}, {
  "_index" : "logs-2014-12-11",
  "_type" : "logs",
  "_id" : "505",
  "_score" : 1.0,
  "fields" : {
    "terms" : [ "janitor", "principal", "robert", "smith" ]
  }
}, {

Is this something that is supposed to happen? I know a huge amount of the library changed on the update.

What is the mapping in both cases?

Figuring that out now.

Would be better to give to others the solution in case someone else hit this.

The mapping used on it looks to be this:

"login": {
   "type": "string",
   "store": true
},

The raw message when queried directly from elasticsearch looks like this on BOTH ES 1.7 and 2.3.3:

"_index" : "logs-2014-12-11",
"_type" : "logs",
"_id" : "505",
"_score" : 1.0,
"_source" : {
  "login" : "smith/robert/principal/janitor",
  "msgClassId" : 3510,
  "logMessage" : "AllPatternMatchesForWord.json",
  "normalDate" : "2014-11-01T18:38:35.0305041Z"

}

The original post that showed differences in the entires was generated using a groovy script trick that was supposed to show the tokenization done by es (I will admit I don't understand it much):

tokens.txt:

{
   "query" : {
        "match_all" : {}
    },
    "script_fields":{
          "terms" : {
              "script": "doc[field].values",
              "params": {
                  "field": "login"
               }
       }
    }

}

curl localhost:9200/logs-2014-12-11/_search?pretty -d @C:\temp\tokens.txt

Settings changed to make it work:

es 1.7:

script.groovy.sandbox.enabled: true (1.7)

es 2.3:

script.engine.groovy.inline.search: on (2.3)

Ah I finally figured it out! ES 2.x doesn't automatically load the template files! Once I put them in manually, everything started working again!