[Dynamite] hacking orbit-idl

[Ian Holmes]

Oops, I think I just broke someone on adnah's edit lock.. sorry

> Has anything been decided on this ?
> Writing in a language for which we have
> no suitable compiler means trouble.
> 
> Options include:
> 
>   a) Use orbit-idl, taking all the CORBA stuff it throws at you.
>      Would this force people to have orbit
>      installed to compile telegraph? (bad) 
>   b) Hack orbit-idl as suggested.
>   c) Write headers manually.
>   d) Postprocess (either manually or automatically) the orbit-idl
>      generated headers to strip off all the stuff we don't want.
>   e) A better solution.
> 
> I'd favour (e), but have been unable to find it on the web
> (I would have thought quite a lot of people would have wanted to use
>  idl without the CORBA stuff)
> 
> Any suggestions?

(b), (c) or (e) are the only ones I think would fly.
I think (c) will migrate to (b), if we start playing around with it.
Still, that leads me to think someone else must have done it. Dunno who
though. Is anyone on appropriate mailing-lists? If not then one(+) of us
should subscribe.

The orbit-idl source is available on the web, in very nice htmlised form

http://cvs.gnome.org/lxr/source/ORBit/src/idl-compiler/

We have been discussing inheritance but it is easy enough to just ignore
it. You just auto-generate function prototypes for every level of the
class hierarchy, and fill them out as you go, copying code.

IDL is really cool to design in and I would miss it.

I have undergone a road-to-Damascus conversion to GLIB (nice one Ewan) &
have put implementation suggestions all over the place. Which is a bit
wistful because I haven't got time to implement much of this myself, but
still. If we want to use any of this design we will probably have to look
again at that division of labour, although I hope we can at least bottle
some of the hiccups at the design stage.

Read through the code up there if you get a chance, it points the way to
PolymerModels a bit; also you can click through and read it like hypertext
because of the nice synchronicity between our naming scheme & Wiki. Also
those "?"'s beside ObjectNames could be a cool way to discuss things you
think should be there, or not there, or different or whatever.

I hope at least the Alphabet code is a bit more palatable now -- I know it
seemed really complex and unnecessary before, so I thought about it and
trimmed it down, and put pseudo-latex-C in the comments to make the formal
definitions of the functions clear.

The ParameterModule may seem a bit weird at first but that's because it's
already "polymer-HMM" ready. I am quite psyched at the idea of an adaptor
class that sucks in a HMMER profile using Sean Eddy's code and turns it
into a GeneWise model.

Ian

> 
> Guy.
> --
> 
> 
> 
> _______________________________________________
> Dynamite mailing list  -  Dynamite@bioperl.org
> http://www.bioperl.org/mailman/listinfo/dynamite
>