[BioSQL-l] ontology for transitive closure table

[Matthew Pocock]

Each step in a path is between a pair of terms that may or may not be 
from the same namespace. Paths themselves in the general case are not 
within a single namespace. If you did the transitive closure calculation 
considering a set of namespaces, then I guess you could give a unique ID 
to that set and label the path with that set. Thinking about it, that 
set is much more sane a label for the path. It says 'I can get from this 
term to this term using information only from these domains'. That 
smells like what we want to optimize searches on.

The current biojava ontology API punts on these issues. We explicitly 
import terms from one namespace to another e.g. MyDomain would import 
core:isa as MyDomain:isa(remote term core:isa). So, if we reason over 
namespaces x, y, z (e.g. join SO to genbank feature names using fred's 
table of identities) then paths would end up being constructed in some 
namespace where all the terms have been localy imported. Not sure how 
sane that is. Hey ho.

Matthew

Aaron J Mackey wrote:
> It seemed to me that we needed this field in the path table for the same
> reason we needed it in the relationship table, to ensure that we know the
> namespace under which this path is defined (which may be a different
> namespace under which the relationships themselves are defined).  This
> gives one the ability to have multiple transitive closure graphs in the
> same table, "typed" by namespace.
> 
> Maintenance of these graphs is left as an exercise to the reader (but
> DBIx::Graph is coming to a CPAN to near you).
> 
> -Aaron
> 
> On Mon, 17 Mar 2003, Hilmar Lapp wrote:
> 
> 
>>I just realized that apparently we also have the ontology (namespace)
>>in the term_path table (the transitive closure table for term
>>relationships).
>>
>>I suggest we remove this column from the path table (it will remain on
>>the relationship table). The problem is if you leave it there how do
>>you populate it if you encounter different namespaces along a path.
>>
>>Matthew, would you have a problem with this? I believe we even said in
>>Singapore we don't put this on the path table didn't we?
>>
>>	-hilmar
>>
> 
> 


-- 
BioJava Consulting LTD - Support and training for BioJava
http://www.biojava.co.uk