[Biojava-dev] Re: [Biojava-l] Mutable Features - strand?

[Thomas Down]

Once upon a time, Matthew Pocock wrote:
> Keith James wrote:
> >ProjectedStrandedFeatures currently throw a ChangeVetoException
> >because I can't see how to propagate the change back through the proxy
> >(I don't think the ProjectionContext interface provides the necessary
> >information).
> I think we need to re-visit the ProjectionContext interface so that it 
> supports bi-directional mappings of feature properties. Anybody with 
> strong view? Thomas?

Hmmm, if we want mutability of projections (something that
doesn't convince me 100%, but I suppose it has its uses),
how about adding methods to ProjectionContext of the form:

    public void setStrand(StrandedFeature f, Strand s)
        throws ChangeVetoException;

That still keeps the projection system nicely flexible,
and the individual projection objects very simple (just
makes calls into ProjectionContext).

I guess the alternative is to gut ProjectionContext and rewrite
it with pairs of methods like:

    public Strand projectStrand(Strand s);
    public Strand reverseProjectStrand(Strand s);

This moves a bit more code into the projection objects, and
maybe simplifies writing ProjectionContexts a little bit.



If I was writing the whole thing from scratch today, I might
consider the second option.  But I think the first option,
which doesn't require any significant changes to existing
code, is probably a better bet for BJ1.4.

    Thomas.