Re: Faster db architecture for a twisted table.

----- Original Message -----
From: "Andreas Pflug" <pgadmin(at)pse-consulting(dot)de>

> Create a table "sibling" with parent_id, sibling_id and appropriate FKs,
> allowing the model to reflect the relation. At the same time, you can drop
> "mother" and "father", because this relation is covered too

Something like a table describing relationships and a table reflecting
relationships from both sides, I guess:

create table relationship_type
(
relationship_type_id serial,
relationship_type_description varchar(20)
)

populated with values such as:
1 Child_of
2 Father_of
3 Brother_of
4 Sister_of
...

And then

create table person_relationships
(
source_person_id int4,
relationship_type_id int4,
target_person_id int4
)

populated with values such as:
1 1 2 (person 1 is child of person 2)
2 2 1 (person 2 is father of person 1)
...

It requires a careful maintenance, as almost all (I'd stick with ALL)
relationships will require a person to appear twice (as source and as
target), but flexible and easy to query.

Helder M. Vieira