Filter & facet counts in search endpoint

[joepio]

While searching, I'd like to see how many instances are available of something if I turn on a specific filter. This is useful for answering questions like:

• How many books are in the horror genre? And how many in sci-fi?
• Which prices do products in my query have? How are they distributed?

Ideally these are returned in a counts object or something similar. In meilisearch you can select which properties you want to count by setting facets. This could be a query param for the endpoint.

What meilisearch does: in its response:

  "facetDistribution": {
    "genres": {
      "action": 273,
      "animation": 118,
      "adventure": 132,
      "fantasy": 67,
      "comedy": 475,
      "mystery": 70,
      "thriller": 217
    }
  }

It looks like Tantivy has very powerful Aggregations, even support for elastic JSON aggerations (e.g. averages with specific ranges). However, it does seem to require fast fields for aggregations, which (I think) excludes JSON fields.

I think we need FacetCollector and their counts for counting. But here, too, we seem to know the schema beforehand, so again, I don't know how it deals with JSON fields.

If that doesn't work, we could maybe have a function that iterates over user's Classes, and adds schema for each property. That way we could still add the count fields.

UPDATE: It's working on the main branch of tantivy now!

Also, there are histograms quickwit-oss/tantivy#1900