Accessing settings of REST endpoints from other modules

[mesemus]

Hello folks,

Intro:
I'm in the middle of implementing our institutional repository on top of invenio 3. For now I have two record types: Thesis (metadata about student thesis) and Attachment (for example fulltext, defence transcript, ... Each attachment has associated ACLs and potentially multiple files, such as original pdf and converted pdf/a). Thesis and n attachments are referenced via thesis id and for a couple of reasons I need to keep the two data types and not denormalize/merge them into one.

So far I've created two helper modules: acls providing generic mechanism for evaluating ACLs directly within elasticsearch queries and links defining LinkedObjectSerializerMixin(PreprocessorMixin) that in generic manner adds to thesis representation its associated attachments' metadata.

Problem:
What I'm facing is that I need to reference settings of REST endpoints - for example, when Attachments are searched for inside the mixin I would like to use the appropriate RecordsSearch - the same that is used for accessing Attachment through REST. So far I've ended up with code bloated with obj_or_import_string:

rest_configuration = current_app.config['RECORDS_REST_ENDPOINTS'][self.other_side_rest_endpoint]
search = obj_or_import_string(rest_configuration.get('search_class', None), default=RecordsSearch)

Suggestion:

What about adding a dictionary current_records_rest.endpoints that would contain the configuration of endpoints that would be already resolved via the obj_or_import_string code that is now present at the beginning of create_url_rules ?

invenio-records-rest/invenio_records_rest/views.py

Lines 237 to 298 in e3a33fc

	read_permission_factory = obj_or_import_string(
	read_permission_factory_imp
	)
	create_permission_factory = obj_or_import_string(
	create_permission_factory_imp
	)
	update_permission_factory = obj_or_import_string(
	update_permission_factory_imp
	)
	delete_permission_factory = obj_or_import_string(
	delete_permission_factory_imp
	)
	list_permission_factory = obj_or_import_string(
	list_permission_factory_imp
	)
	links_factory = obj_or_import_string(
	links_factory_imp, default=default_links_factory
	)
	# For backward compatibility. Previous signature was links_factory(pid).
	if wrap_links_factory(links_factory):
	orig_links_factory = links_factory
	def links_factory(pid, record=None, **kwargs):
	return orig_links_factory(pid)
	record_class = obj_or_import_string(
	record_class, default=Record
	)
	search_class = obj_or_import_string(
	search_class, default=RecordsSearch
	)
	indexer_class = obj_or_import_string(
	indexer_class, default=None
	)
	search_class_kwargs = {}
	if search_index:
	search_class_kwargs['index'] = search_index
	else:
	search_index = search_class.Meta.index
	if search_type:
	search_class_kwargs['doc_type'] = search_type
	else:
	search_type = search_class.Meta.doc_types
	if search_class_kwargs:
	search_class = partial(search_class, **search_class_kwargs)
	if record_loaders:
	record_loaders = {mime: obj_or_import_string(func)
	for mime, func in record_loaders.items()}
	record_serializers = {mime: obj_or_import_string(func)
	for mime, func in record_serializers.items()}
	search_serializers = {mime: obj_or_import_string(func)
	for mime, func in search_serializers.items()}
	resolver = Resolver(pid_type=pid_type, object_type='rec',
	getter=partial(record_class.get_record,
	with_deleted=True))

Then my module would just call

current_records_rest.endpoints[self.other_side_rest_endpoint].search_class 
or 
current_records_rest.endpoints[self.other_side_rest_endpoint]['search_class']

and would get the actual implementation of RecordsSearch.

If it makes sense I'm willing to create a pull request. Your opinions?

Thanks for all the work, so far I'm really happy with the ease of implementation and repository speed,

Mirek