Indexing a threaded conversation

[dansei]

I would like to use Elasticsearch to index and search a threaded conversation, like a blog's comments where there are top-level comments and replies to those comments.

I have tried making a single document like this, with a Nested comments field, with each comment having an ID, a parent ID, and a list of child reply IDs.

{
    "comments": [
        {
            "id": 1,
            "text": "nice post",
            "replies": [ 2 ]
        },
        {
            "id": 2,
            "parent": 1,
            "text": "Thanks for reading my blog!"
        },
        {
            "id": 3,
            "text": "very helpful"
        }
    ]
}

This works to an extent because I can search through the comments.text fields and find matches. But I think the API gives me back only the matching text, removed from the context of the adjacent fields that tell me where the comment is in the discussion.

Is this a good way to store this data?

I could also put each comment as a separate document, with a common conversation ID (e.g., the blog post ID), and the parent and child IDs. I'm not sure if this is a good use case for "join" queries or not.