[Biojava-l] More puzzled with Dazzled

[Thomas Down]

On Sun, Jul 01, 2001 at 08:35:51PM +0100, David Huen wrote:
> I've finally got RagbagDataSource serving all manner of formats via
> Dazzle. It serves the 900 kb leftmost tip of Drosophila X-chromosome in
> GAME, Embl and genbank formatted files. Performance seems OK beyond the
> initial parse scanning run (to check collate feature types available
> mainly, without this I could start much faster).

The list of feature types is a part of the DAS/0.996 spec,
so it has to be done to be complete.  I have mixed feelings
about it myself.  It's something that's potentially pretty
interesting information to have, but I agree that it's a
serious hastle for some types of flatfile-based server.
Heigh ho.

> However, I have new problems to resolve.  First is how does
> servlet.sanger.ac.uk deliver hierarchical features?  I have constructed
> Biojava feature trees that are nested (exons in transcripts in genes) but
> Dazzle 0.90 returns them as GFF with a flat structure leading me to lose
> the ability to do ziggies and such-like.  Is there a way to persuade
> Dazzle to do it with a particular datasource?

The DASGFF format doesn't handle sparse-locations or hierarchical
features (although it /does/ have a kind-of primitive method for
feature grouping, which I should use in Dazzle, at least under
some circumstances -- 0.91 should have this).

Dazzle also supports XFF feature tables (dsn/features?segment=blah;
encoding=xff).  These handle both nested features and sparse locations.

The biojava DAS client library includes code to auto-detect when
it's taklking to an XFF-enabled server, and alters it's behaviour
accordingly.

servlet.sanger.ac.uk DOES run Dazzle, but is currently still on
the 0.08 (i.e. pre-rewrite) version.  I've given a new webapp
to our webmaster, and I hope that the server will be in
sync within the next couple of days.


    Thomas.