Re: [Gnumed-devel] GnuMed/Archive // patient object

[Thierry MICHEL]

On Mon, 2003-01-20 at 11:20, Ian Haywood wrote:
> On Mon, Jan 20, 2003 at 08:48:11AM +1100, Horst Herb wrote:
> > On Mon, 20 Jan 2003 01:00, Karsten Hilbert wrote:
> > > Our main show-stopper currently is that GnuMed proper is still
> > > missing a patient object with associated API. I am urgently
> > > asking Horst to check in what he is working on (test-area) and
> > The alternative would be a procedural API with the usual "set" and "get" 
> > functions for each attribute.
> > 
> > Coments and suggestions more than welcome.
>  The API/middleware/whatever is also needed to perform the 'backroom'
> functions that would otherwise be done using stored procedures (in a
> monolithic database), such as reference counting, interaction checking, 
> etc.. This means 'regular' Python objects,such as
> Patient, Drug, Prescription, etc. which may have a 1-to-1 correspondence
> with the postgres tables/views,, or may not, data accessed via
> functions.
> 
> This is slow to write, but, in honesty, can be written quicker than the
> length of time since Richard fuinshed the interfaces (now getting on for
> 5 months)
> 
> In terms of using dictionaries/ etc to represent tables, it is nice, but
> I think, inevitably loses too much flexibility. All I would want is just
> a few
> fsimple functions to fluently mix SQL and Python code (instead of having
> to mess about with cursors all the time)

Why not a layer which could map objects "transparently" to the database.
The advantage is that in this case you handle *only* objects, finished
with cursor and selects, insert and so on.

> 
> I volunteer to start working  on a Patient object (an interface to the
> gmclinical service, as is) (BYW, is this finished yet?!)
> Then Karsten will be happy ;-)
> 
> 
> Ian
> 
> 
> _______________________________________________
> Gnumed-devel mailing list
> address@hidden
> http://mail.gnu.org/mailman/listinfo/gnumed-devel
-- 
Thierry MICHEL <address@hidden>
Nekhem Technologies