uniqueBlock is not faster than BlockJoinFacet in 7.4.
On Tue, Aug 7, 2018 at 8:05 AM Aditya Gandhi <agandhiw...@gmail.com> wrote:
> I'm querying an Index which has two types of child documents (let's call
> them ChildTypeA and ChildTypeB)
> I wrap the subqueries for each of these documents in a boolean clause,
> something like this:
>
> *q=+{! parent which=type:parent <query_on_child_type_B>} +{! parent
> which=type:parent <query_on_child_type_B>}*
>
>
> I've been trying to get facet counts on documents of ChildTypeA (rolled up
> by parent) and I've tried the following approaches
>
>
> - Tried Block Join Faceting using the JSON API i.e. using the
> unique(_root_) approach.
> - Enabled docValues on _root_
> - *This did not scale well*
> - Tried using the BlockJoinFacet component.
> - Had to customize it since it expects that only one
> *ToParentBlockJoinQuery* clause to be present in the query.
> - Since I needed facet counts only on ChildTypeA, I changed it to
> ignore the clause on ChildTypeB
> - I did not enable docValues on _root_ since it was not mentioned in
> the documentation.
> - *This approach did not scale well*
>
> I needed advice on whether I could have done anything better in any one of
> the two approached I've tried so far. Also if there exists some other
> approached I could try.
> Would using the uniqueBlock in 7.4 help? (Though this would require me to
> upgrade my Solr version)
>
--
Sincerely yours
Mikhail Khludnev
