Blacklight 9: SingleDocSearchBuilder gets a Blacklight::Solr::Repository as its first argument, should get a Controller

[sandbergja]

When on a record show page, Blacklight::Solr::Repository#fetch creates a new SingleDocSearchBuilder. The first argument is self, that is to say, an instance of Blacklight::Solr::Repository:

blacklight/lib/blacklight/solr/repository.rb

Line 10 in b969eac

doc_params = SingleDocSearchBuilder.new(self, id, params)

SingleDocSearchBuilder calls super on that first argument:

blacklight/lib/blacklight/solr/single_doc_search_builder.rb

Line 10 in 9ba4ae9

super(scope)

And we then try to use it to create a Blacklight::SearchState or similar, as the third argument:

blacklight/lib/blacklight/search_builder.rb

Line 32 in 9ba4ae9

@search_state = search_state_class.new(@blacklight_params, @scope&.blacklight_config, @scope)

But yikes, the third argument to Blacklight::SearchState's constructor is supposed to be a controller, not a solr repository:

blacklight/lib/blacklight/search_state.rb

Line 21 in 9ba4ae9

def initialize(params, blacklight_config, controller = nil)

This doesn't cause trouble in stock blacklight, but if you have customized the search builder to include something that relies on, say search_state.controller&.action_name, you will get

*** NoMethodError Exception: undefined method 'action_name' for an instance of Blacklight::Solr::Repository

[jcoyne]

I don't think there should be any coupling with a controller for a SearchState. Instead can you create a custom SearchState and give it all of the inputs? The reason is that a search should be able to run even if it isn't in a controller (e.g. a background job)

[sandbergja]

@jcoyne : do you think we could allow construction of a SearchBuilder and SingleDocSearchBuilder without a scope aka controller then? I think having no controller would be better than getting a Repository when you expect a controller.

[jcoyne]

Yeah, I think that's the direction we ought to move in. It's not really set up in a good way right now.