[Biojava-l] BioSQL observations

[Thomas Down]

On Thu, Mar 14, 2002 at 09:56:01AM +0000, Keith James wrote:
> 
> In the longer term I think that BioSQLSequenceDB should use a
> DataSource or some new interface which wraps a DataSource or
> DriverManager (instead of using the JDBCConnectionPool class). That
> way anyone can plug in their own e.g. their vendor's
> ConnectionPoolDataSource.

Yes, this is definitely on my todo list.  My one concern
with this. originally, was that it led to an extra compile-time
dependancy for the BioJava package.  But since the javax.sql
interfaces are included in JDK1.4, this is going to become
less of an issue over time.

> On the schema itself, was there any discussion on being able to track
> changes to Features? I'm thinking of situations where the a number of
> people are applying annotation to a genome and want to know "when was
> this Feature last modified?"

Hmmm, I'd really like to see the core schema stay fairly
stable and general-purpose.  That said, at the Hackathon we
discussed an idea of `extension modules' -- extra schema
files which get loaded into the database on top of the core
schema.  There are already a couple of these proposed (for
ontologies and genome assemblies).  The BioJava code auto-detects
whether the assembly module is available, and uses it if so.
Same sort of thing will probably happen for the ontologies.

A change-tracking module sounds like a good candidate for
this kind of treatment -- I'd be happy to work with you on
this if there's a bit of spare time during the bootcamp
development sessions.

> I've noticed that it is possible to save an EMBL entry which has no AC
> line despite accession being 'NOT NULL' because an empty string is
> inserted instead. I guess this should throw an Exception - shall I
> make it so?

Hrmmm, yes.  Now that you mention it, that does sound like a
sensible idea.

Go for it, if it'll be helpful,

    Thomas.