Study how to deal with relation between a broad number of entity types
see: https://hedgedoc.logilab.fr/p/h3TNm_CY7#/
Maybe should we think about making one table by couple subject/object.
For instance, if I've the relation:
class born_place:
subject = ("Person", "Org")
object = ("City", "Country")
I may want to have 4 tables in SQL:
- person_born_place_city
- org_born_place_city
- person_born_place_country
- org_born_place_country
This should be more efficient in many cases, if we start with the assumption that we will always have a join with the of the several etype (for instance, if I want all entities born in Angoisse, then I will need to make a join with cw_Person and cw_Org, because I'm pretty sure we will want some attribute with the relation).
I thought of several things to do first:
- study how much time we will loose if I just need the RQL results of "Any X,Y WHERE X born_place Y"
- list every possible case of request and see the implication of the aford mentionned modification