Unexpected (case insensitive) string sorting in Elasticsearch

Question

I have a list of console platforms that I sort in Elasticsearch.

Here is the mapping for the "name" field:

{
    "name": {
        "type": "multi_field",
        "fields": {
            "name": {
                "type": "string",
                "index": "analyzed"
            },
            "sort_name": {
                "type": "string",
                "index": "not_analyzed"
            }
        }
    }
}

When I execute the following query

{
  "query": {
    "match_all": {}
  },
    "sort": [
        {
          "name.sort_name": { "order": "asc" }
        }
    ],
    "fields": ["name"]
}

I get the following results:

{
    "took": 1,
    "timed_out": false,
    "_shards": {
        "total": 3,
        "successful": 3,
        "failed": 0
    },
    "hits": {
        "total": 17,
        "max_score": null,
        "hits": [
            {
                "_index": "platforms",
                "_type": "platform",
                "_id": "1393602489",
                "_score": null,
                "fields": {
                    "name": "GameCube"
                },
                "sort": [
                    "GameCube"
                ]
            },
            {
                "_index": "platforms",
                "_type": "platform",
                "_id": "1393602490",
                "_score": null,
                "fields": {
                    "name": "Gameboy Advance"
                },
                "sort": [
                    "Gameboy Advance"
                ]
            },


    {
            "_index": "platforms",
            "_type": "platform",
            "_id": "1393602498",
            "_score": null,
            "fields": {
                "name": "Nintendo 3DS"
            },
            "sort": [
                "Nintendo 3DS"
            ]
        },

        ...remove for brevity ...

        {
            "_index": "platforms",
            "_type": "platform",
            "_id": "1393602493",
            "_score": null,
            "fields": {
                "name": "Xbox 360"
            },
            "sort": [
                "Xbox 360"
            ]
        },
        {
            "_index": "platforms",
            "_type": "platform",
            "_id": "1393602502",
            "_score": null,
            "fields": {
                "name": "Xbox One"
            },
            "sort": [
                "Xbox One"
            ]
        },
        {
            "_index": "platforms",
            "_type": "platform",
            "_id": "1393602497",
            "_score": null,
            "fields": {
                "name": "iPhone/iPod"
            },
            "sort": [
                "iPhone/iPod"
            ]
        }
    ]
}

Everything is sorted as expected, except for the result iPhone/iPodat the end (and not after the GameBoy Advance) - why does the value /in the name affect the sorting?

thank

Answer 1

Okay, so I found that the reason has nothing to do with /. ES will be sorted in uppercase and then lowercase.

settings :

{
    "analysis": {
        "analyzer": {
            "sortable": {
                "tokenizer": "keyword",
                "filter": [
                    "lowercase"
                ]
            }
        }
    }
}

'analyzer': 'sortable' sort_name multi.

Answer 2

ES http://sense.qbox.io/gist/2521288d96cb4d7ea4a0356505af31c1ba669e88