Re: [Gnumed-devel] Movve to Qt (was: Conference)

[Roberto Mello]

On Fri, Aug 15, 2003 at 10:47:17PM +0200, Karsten Hilbert wrote:

> Excellent. I think it'd be worthwhile to collect your mom's
> clinic's requirements into
> CVS:client/doc/TODO/GP-practice-reqs-brazil.txt.

Ok. I checked out a copy of the CVS repository. Where are the install
instructions?

Do I just send you the requirements file to be added to CVS? While I'm
asking this, what's the procedure to gain CVS access? Meritocracy?
 
> > Especifically, I would like to work on the scheduling
> > facility (schedule patient appointments) and
> Ian has done some fairly extensive schema work in that area. I
> haven't looked at it (not that I'd have to for it to be
> worthwhile ;-).  There's some code from Horst and probably
> some more code buried somewhere in Horst's machines.

Ok. During my last trip to Brazil I looked into several medical software
packages for GPs, so I gathered some ideas that I will ellaborate.

> You might take a look a clinical/demographics to see how we
> did things so far, offer some suggestions and then maybe
> embark on the scheduling stuff ?  I also have one very

Ok.

> 
> Can a good case be made *against* using NOT NULL constraints
> on table attributes in audit trail tables where the
> corresponding attributes in the audited table also have a NOT
> NULL constraint ? OR is it (and for what reason) important to
> not have ANY constraints on audit trail tables ? We are doing
> the NOT NULL in audit trail tables currently as I feel the
> need to be as strict as possible about the data BUT i'd

The data model should be a truth table of your business rules. NULL is
something really bad that SQL introduced, and that breaks the relational
model (one of the many ways SQL breaks the model). According to the
relational model, nothing should ever be NULLable.

However, in audit trail tables we know where the data is coming from (if
I'm understanding what you're doing, this is the way I do it). The data is
INSERTed into audit tables by triggers coming from tables that have all
the appropriate NOT NULLs and other constraints in place to keep the data
valid.

So having audit trail tables with NOT NULLable attributes incur extra 
checks by the database system, which can probably be saved because we know 
the data is valid. OTOH, the performance impact may be neglegible and the
extra peace of mind may be worth more.

> love to drop that if I have good reason as that would make it
> vastly easier to auto-create audit trail tables... Which is
> desirable ...  Well, writing this it occurs to me that when

Yes, it would be easier, but thinking about it for just a little bit,
would it really be that much easier? You could write a function/method
that would return which attributes in a table are NOT NULL, then create
the table with a simple 

  CREATE TABLE table_audit AS 
  SELECT * FROM table LIMIT 0;

and use the results of your function to add not null constraints with
(maybe a single) ALTER TABLE statements.

Maybe I'm forgetting something here though.

> the audit trail table is missing for a table that is marked
> for auditing I should just log that as a warning and happily
> create the audit trail table assuming the user knew what she
> was doing, shouldn't I ? :-))

I don't see why not, but then I'm new to GNUMed.
 
> > > 7/ Generalisation of clinical history API from the Allergies functions
> > > to avoid separate functions for allergies, immunisations, etc:
> > > viz. a thinner layer around updateable views.  Exact arrangement of
> > > these views depends on outcome of design process.
> > I'm interested in (7).
> I'd be right on with you there. Please post your ideas and
> feel free to ask if clarification is needed. I am to blame for
> some of that stuff in CVS.

Ok. I'm reorganizing things here because I'm sending my laptop in for
service, but you should hear from me in the next couple days.

-Roberto

-- 
+----|        Roberto Mello   -    http://www.brasileiro.net/  |------+
+       Computer Science Graduate Student, Utah State University      +
+       USU Free Software & GNU/Linux Club - http://fslc.usu.edu/     +
Whoa...I did a 'zcat /vmlinuz > /dev/audio' and I think I heard God...
        -- mikecd on #Linux