[Bioperl-l] Re: [Gmod-schema] Re: BioSQL or chado

[Hilmar Lapp]

On Wednesday, July 30, 2003, at 10:29  AM, Allen Day wrote:

> It shouldn't be too hard to autogenerate the simple bridging where 
> there
> is a one-to-one correspondence between chado tables (or chado star
> subschemas) and bioperl classes.

BTW this is exactly what bioperl-db does to a large extent where there 
is such a straightforward correspondence. It only does it at runtime, 
it doesn't generate classes you'd have to maintain.

	-hilmar

BTW in reality it turns out that there are really only very few classes 
that map directly to tables without special case code. It's not a new 
finding that relational and object models are different animals. People 
tend to forget that. I'd be curious to learn how you can auto-generate 
an at least minimally useful OR mapper from bioperl to any of chado, 
biosql, or you name it.

Literally all of the generic OR mappers that I'm aware of either a) 
assume you've got the schema and generate the object model and the 
mapping off of it, or b) assume you've got the object model and 
generate the schema and the mapping off of it, or c) require an 
extensive amount of hard to debug (which is a euphemism here) 
configuration files.

Here at GNF we've been going through exercise c) with J2EE CMP as 
implemented by Jboss. We're not going to do it again. Auto-generating 
the OR mapping either hand-cuffs one arm onto your back, or it's just 
not worth the nightmare of staring at seemingly correct configuration 
files.

-- 
-------------------------------------------------------------
Hilmar Lapp                            email: lapp at gnf.org
GNF, San Diego, Ca. 92121              phone: +1-858-812-1757
-------------------------------------------------------------