Re: Nested Sets vs. Nested Intervals
Date: 12 Nov 2005 10:30:17 -0800
"Are there any rules for determining
the role of its nodes such as collection (Employess, Departments), instance ('John Smith', 'Accounting Department'), properties (age, salary) etc?"
A branch of the graph could contain the shema. I also favour shemaless idoms a la XML. In this case canonical topologies are used. And a canonical semantic model with a focus on sets and attributes. This is a matter of convention I guess. Often times I design a Mneson database instance with a Top vertex leading to all concepts and sets as named attributes of Top. So a new user inspecting the attribute names of Top would get the schema. Often times an attribute of Top named "types" is the set of all attribute types (=attribute names) in the database. This is the catalog problem occuring also in relational systems.
"Can these roles be mixed, for example, a node could play a role of
property with respect to one subgraph and an entity for another subgraph?"
Absolutely and this is I believe one of the strenghs of the model.
"Is it true that this database engine does not know anything about the
roles of nodes and manipulates only the graph using some formalism?"
The system knows about the canonical structures so it can process canonical graphs in a clever way e.g. by displaying attributes as labelled edges.
"Is there any informal interpretation of x -> y connection which would
allow me to build a graph given some problem domain description?"
The canonical interpretations of x -> y are
(1) if x is the head of a set, y is a member
(2) y is a property of x
(3) the case [attribute type (or name) -> attribute instance] it is just a convention, however one could say an attribute instance is a member of the attribute type, seing the attribute type as a set, so this case would be derived from (1)
Note how (1-2) combine neatly to form a record = set of attributes.
(Excellent questions you all, thank you, sorry for the late responses, and please if I fail to respond drop me a note directly, it might be because I failed to detect the question as I'm just browsing in Google Groops.) Received on Sat Nov 12 2005 - 19:30:17 CET