[Biojava-dev] Re: biosql

[Len Trigg]

Thomas Down wrote:
> On 15 Mar 2004, at 15:41, Matthew Pocock wrote:
> > Scratch that - a clean rebuild seems to have fixed it. I am getting a  
> > single failure though:
> >
> > Testcase: testFeaturePersistence took 0.843 sec
> >    FAILED
> > [feature] location is now an instance of:  
> > org.biojava.bio.symbol.RangeLocation
> 
> No it isn't -- I've been waiting for this failure to be fixed before  
> turning on the database-dependent tests.  Trouble is, we don't make a  
> nightly release if *any* tests fail, and this one has been hanging  
> around for a while...

This is one of the two BioJava/BioSQL bugs found a few weeks ago by
Frederik Decouttere. I fixed the other one but haven't had a chance to
deal to this one.

The problem is that when a BetweenLocation (and probably a whole bunch
of others) get persisted, they get retrieved as regular
RangeLocations. It would be straight forward to store a term
corresponding to the location type in the term_id of the location
table (currently unused in our binding). It's just a matter of
determining what terms to use. Either (easiest to hardest): a) the
classname of the location type; b) whatever BioPerl uses (if it is
capable of persisting the location type); c) jump through a few hoops
and look up an appropriate obda name.

Has anyone ever tried testing whether sequences of any complexity
written to BioSQL via the BioJava binding are readable via BioPerl,
and vice versa? I believe our Ontology persistence has currently
diverged from the BioSQL schema somewhat anyway.


Cheers,
Len.