On 21 Dec 2006, at 17:06, David Croft wrote:
Hi Arek,
Do your main tables share the same dimensions at the moment?
That's interesting. We only see it if the main tables are related ei
one
is 1:n to the other. In this case you can create a composite dataset
exactly like
you had before but one of them would have an additional key., For an
example
look at ensembl gene and transcript main tables
OK, I confess, there is a relationship. pathway:reaction is 1:n. I
could
actually use the same attribute (stableIdentifier) as the key between
the
two tables, since it is unique for every Reactome object. I guess I
would
just have to create a column with a new name, like main_link_key or
something. One possible problem is the fact that two pathways may
share the same reaction.
So, theoretically, I could keep both tables in the same dataset.
Would this
bring any advantages, apart from not needing to copy dimension tables?
E.g. would it make it easier for the user to run queries?
yes, the users could ask more queries within the same dataset and the
software
would automatically resolve to one or the other main tables without
redundancies
We do recommend to create one 'compound' dataset rather then having
two separate ones as it does indeed make users life easier.
However you must be aware that the subclass table need to have all
columns
from the 'superclass' one.
a.
Cheers,
David.
------------------------------------------------------------------------
-------
Arek Kasprzyk
EMBL-European Bioinformatics Institute.
Wellcome Trust Genome Campus, Hinxton,
Cambridge CB10 1SD, UK.
Tel: +44-(0)1223-494606
Fax: +44-(0)1223-494468
------------------------------------------------------------------------
-------
