You want to upload many documents in your ElasticSearch database, using _bulk, but you need to format your input to the accepted format ? Here is what you can do
Tested Configuration:
MacOS: Sierra 10.12
Terminal: iTerm2 & ElasticSearch 6.2.4 & jq 1.5

1. What is the matter ?

You have a JSON object like this:

{
  "type": "FeatureCollection",
  "features": [
    {
      "type": "Feature",
      "properties": {
        "code": "91016"
      },
      "geometry": {
        "type": "Polygon",
        "coordinates": [
          [
            [1.95923, 48.30858],
            [1.96002, 48.30808],
            [1.95923, 48.30858]
          ]
        ]
      }
    },
    {
      "type": "Feature",
      "properties": {
        "code": "91021"
      },
      "geometry": {
        "type": "Polygon",
        "coordinates": [
          [
            [2.23497, 48.58895],
            [2.23514, 48.589],
            [2.23497, 48.58895]
          ]
        ]
      }
    }
  ]
}

And you want to upload every “features” in your ElasticSearch database. However, the recommended format for bulk insert is like this:

POST _bulk
{ "index" : { "_index" : "featureindex", "_type" : "_doc"} }
{"type":"Feature","properties":{"code":"91016"},"geometry":{"type":"Polygon","coordinates":[[[1.95923,48.30858],[1.96002,48.30808],[1.95923,48.30858]]]}}
{ "index" : { "_index" : "featureindex", "_type" : "_doc"} }
{"type":"Feature","properties":{"code":"91021"},"geometry":{"type":"Polygon","coordinates":[[[2.23497,48.58895],[2.23514,48.589],[2.23497,48.58895]]]}}

Imagine you have 100.000 records to insert, then you need a script to format this for you ! That’s when jq comes

2. Solved in less than a minute:

Type:

brew install jq

Then go to your directory where the JSON is stored

cd ~/Downloads #in this example, the JSON is called "communes.json" and stored in the Downloads folder

And Run this command:

jq -c '.features[] | { index: { _index: "featureindex", _type: "_doc" } }, .' communes.json > communes2.json

Done.
The communes.json was transformed into communes2.json (stored in the same location), which can be sent to ElasticSearch

3. Notes

We can notice that we have some geojson here. We could create a mapping for a geojson Feature to map with:

1. the geometry field as a geo_shape type
2. the type as a keyword type
3. the properties as an object type

4. Ressources

Official jq website: jq
Testing jq online: onlineTestjq
ElasticSearch geoshape
ElasticSearch bulk
Blog: programminghistorian
Blog: shapeshed
Other examples of json to be fromated: geoapi