Able to have different _id value than the one in the _source

[Diolor]

I have edited this issue since the previous was wrong.

Is it possible to have a different _id value in ES than the one inside the _source? The following is a bulk action:

action = {
    '_type': 't',
    '_id': '52cb45cec36b4442751728f5',
    '_source': {
        u'city': u'Toronto',
        u'name': u'PostBeyond',
        u'_id': {
            '$oid': '52cb45cdc36b4442751728f4'
        },
        u'events': {
            u'title': u'ExtremeCachingwithPHP',
            u'event_id': {
                '$oid': '52cb45cec36b4442751728f5'
            },
            u'start_date': u'2014-01-08T00:00:00+00:00'
        }
    },
    '_index': 'i'
}

If we give manually the _id ES not to search inside the source document.

It would have been useful to cooperate with mongodb or generally if we want to have in the _source the field _id.

Relevant documentation: http://www.elasticsearch.org/guide/en/elasticsearch/reference/current/mapping-id-field.html#mapping-id-field

[clintongormley]

I'm sorry but I don't understand your question at all. Could you try explaining a bit more?

[Diolor]

Yes; sorry for the badly written issue.

Elasticsearch always searches for an _id inside the document which is going to be indexed.
In MongoDB, each document has an _id field which is in a {'_id': { '$oid': '52cb45cdc36b4442751728f4' } } format. e.g. the document:

{'_id': { '$oid': '52cb45cdc36b4442751728f4' },'testkey':'testvalue'}

I am working on adding elasticsearch as searching layer so I cannot keep the same structure both in MongoDB and ES.

I either have to rename _id inside the source to id (or something else) or set _id to a {'_id': '52cb45cdc36b4442751728f4' } format happily to be handled by ES.

Generally if I could disable ES searching inside the document for an _id field (at least in bulk feed where you can set/give the _id on the same level with _source) I would be able use the _id inside the _source without any limitations of its structure.

However this is a minor issue since can be solved in the webservice or end client.

[clintongormley]

On v1.x, the {"_id": { "$oid"...} is ignored because it is an object rather than a string, so you should be fine when you upgrade. However, I think there is still an issue here when you use documents like { "_id": "xxx"}, so I've opened #5558.

What you could do with your current version is to map the _id field to use _id.$oid explicitly:

PUT /myindex/
{
    "mappings": {
        "mytype": {
            "_id": {
                "path": "_id.$oid"
            }
        }
    }
}

I'll close this in favour of #5558