Elasticsearch get position of phrase in document

Question

I'm working on a project where I need full-text search a book. I only need to search in one book at a time and I need to get offset of search term from beginning of the book. I need it for site that's powered by Django/python but I think that Elasticsearch is better and faster.

So far I haven't used Elasticsearch directly only through abstraction layer django-haystack.

Edit1: I need to display for users not only the text they are searching for but also link for them to get to that text. Basically it should work like search box in preview on Mac. Users see search results with surrounding text and if they click on it they are redirected by JS to part of the book, where the text is located.

Answer 1

Will simple highlighting suffice? Even if not, a brute-force solution would be to set the highlighting pre_tags to a programmably-identifiable value and calculate the offset from that. Speed it up by setting term_vector to with_positions_offsets in the mapping to use lucene's fast-vector-highlighter:

{
    "type_name" : {
        "content" : {"term_vector" : "with_positions_offsets"}
    }
}

If that's not acceptable, check out this answer for information on how the offsets are stored internally.

EDIT: Based on your edit, I'm not sure how having the offset would help that much. I mean, unless you're displaying preformatted text or some other fixed layout, how would you know where on the rendered page the offset corresponds?

I think the most elegant solution is to use pre_tags and post_tags to wrap matched text in elements. Then use JavaScript to assign each match an id, creating new fragment identifiers to which you can set the location.