[Biojava-l] Remote location exception

Thomas Down td2@sanger.ac.uk
Wed, 22 Aug 2001 14:54:07 +0100

On Wed, Aug 22, 2001 at 08:45:16AM -0500, Michael Muratet wrote:
> Note: I've switched to a plain-text editor.

Thanks :-).

> I'll take a look at the RemoteFeature class and see what I can do with
> it. From my point of view I'd vote skip the feature and alert the
> calling method so that it can do something else, like look for the
> translated protein. (Maybe this is a more robust approach to the cds
> information, anyway.) I also think that if we as users have any
> influence with the curators, maybe we can get them to change to suit us.

It all depends who you ask.  Some people just want reliable
data out of their parsers, others need to be able to represent
whatever strange things might come out of Genbank, and write
out files as close as possible to the original.  That's why I
quite like Keith's pluggable nasties-handling idea.

But for the behaviour you want, RemoteFeatures might be the
best way to go.  You can parse everything in, then check
very quickly for RemoteFeatures, and implement special-case
processing at that point.

> Confessing that I have never read the DAS spec, what does it say in a
> case like this?

In DAS 1.0, not a lot...  We've been taking an `Acedb-like'
approach, so that features which don't fit on a single contig
get `promoted' onto a higher level of the assembly.  This means
that, for instance, a lot of Ensembl gene predictions are
actually served in FPC contig coordinates.  Works quite

For DAS/2 it would be kind-of nice to handle these things
more intelligently -- especially cases like gene structures,
which tend to be defined by their exons.