I have a table of entities. All entities have required attributes, one of them is bucket_id, and also user can define custom attributes (attribute_name + value) for entity.
I need the following operations with these entities:

1. Filter entities in one bucket by any combination of attributes (must be fast, only equality)
2. Get list of names of custom attributes for all entities in one bucket (must be fast)
3. All custom attributes and some required can be updated (can be slow)

Required attributes can be table columns. I found two ways to store custom attributes:

• store json and use GIN-index for fast filtering
• store attribute names in separate table and values with (entity_id, custom_attribute_id) key in many-to-many table, but in this case i need to do join on condition like (attribute_id = ? AND entity_id = entity.id) for any new custom attribute in query as i can see

Which one of these approaches is better or can you suggest the better?