I'm having some trouble understanding facet results. Need to figure out a way to create a test suite for you to test if needed. For a search query, without

facet_by

or

group_by

typesense fetches 57 hits (

found: 57

). I add a

facet_by=author

and

group_by=author

. My expectation was that the "`found`" count would remain the same and just the results count will be facet'ed between the values in the "author" field. There are 2 distinct "author" field values. The facet counts show a count of 1 for each author field value. "found" also shows only 2 instead of 57 without the facet_by and group_by search query, but the number of documents are 3 each for the 2 authors. So i'm having bit of hard time understanding how to interpret the search results between both cases. Any help would be appreciated. Thanks.

On grouping the counts will now be based on the groups, and not the individual documents.

Thanks for the quick response. I assumed so. I guess the other count that I need to understand better is the

facet_counts

>

counts

array. There are 2 elements in the array, which is fine given there are 2 distinct values for the

author

field. However each element of the

counts

array shows a

count:1

.I was expecting this to have counts like a SQL equivalent of

select author, count(document.id) from collection group by author

.

I think it will be great if you can put together a small example using

curl

snippets so that it's easy for us to reference things.

Will do. Thanks.

Thanks for putting that together. Facet counts happen on the grouped results because that's what makes sense most of the times. Imagine a ecommerce store selling multiple variants of the same product. When you do a group_by on the product_id, you want the facets to be applied only on the groups (i.e. the product and it's variants form a unified group).

Hmm..thanks for the explanation. I'm just rethinking what I'm trying to achieve in my case. I guess my objective is - given a search term, • get a list of articles from all authors in my collection. • Organize the results by author with a suggested count of hits by each author. I guess, I'll get the counts from within the

grouped_hits

itself. Alternatively, will rethink my whole approach. Thanks for the quick explanation!

@Prabhu Ravichandran We will be happy to consider this as a feature request (maybe via a new parameter that specifies the behavior for counts). Can you please open a Github issue with the example above?

Will do! Thanks.