How to avoid N+1 when reindexing models with Searchkick

Question

I am using searchkick to wrap OpenSearch. My Program model has this search_data:

{
  name: name,
  description: description,
  provider_name: provider.name,
  category: category&.name,
  age_min: age_min || 0,
  age_max: age_max || 999,
  start_dates: current_published_schedules.pluck(:start_date),
  zipcode: address&.zipcode,
  program_type: program_type
}

When I run the reindex task:

Program.reindex(refresh: true)

I see in the logs that many SQL queries for Provider and Schedule have been made, one for each Program. Is there a way to tell Searchkick to "includes" or "preload" somehow to optimize the reindexing?

Answer 1

Hopefully in searchkick eager load method is available, we can use includes with help of the search_import method.

You can refer more about on searchkick documentation here.

From above sample data, i have seen your program table have has one association with provider, category, etc. with help of that i try to give sample examples below, hope it might helps to you.

Program Model

From inside of model you need to write as

scope :search_import, -> { includes(:provider, :category, :current_published_schedules) }

Now When you hit the reindex, it first eager loads the all association data what you have mentioned in search import. Now it continuously fires the single query for each program data to update in elastic.

Hope this helps to fix your issue :)

Answer 2

You can use their documentation mentions you can cache those values in the DB level