Runing import without passing it through ID fetcher

[TikiTDO]

Given code like this:

  my_scope = MyMode.includes(:some_association).where(stuff: things)
  MyModelIndex::MyModel.import = my_scope

# Where MyModelIndex uses :some_association to generate a field.

The import will end performing one extra query to load some_association per instance of MyModel in the scope, instead of using the included data.

This happens because both import_scope here and import_objects here re-run a select, regardless of what I pass in. For a more complex index, this can translate to a bunch of queries per instance that I want to import.

The only way I can do this right now is by skipping any sort of external API, and just doing an import directly:

some_scope.find_each(batch_size: 1000) do |batch|
  routine = Chewy::Type::Import::Routine.new(SomeIndex::SomeModel, {})
  routine .process({index: batch})
end

I would really prefer not to have to resort to something like this, so would it be possible to introduce either a key that skips the re-import, or some way to customize the default_scope used when re-loading the data to be imported?

[pyromaniac]

Ok, agree, how about to add import_objects method near the import method which will do the thing you need? Can you prepare a PR?

[pyromaniac]

Or an option is also fine by me

[TikiTDO]

I like the method idea. I'll poke at it today and send a PR with a POC.

[TikiTDO]

Actually maybe a bit of both, to keep the same structure with the import_routine into import_(parallel|linear).

[rabotyaga]

Closed in #753