On Wed, Oct 26, 2005 at 09:23:57AM +1000, Ian Haywood wrote: > > -- --------------------------------------------- > > create table reviewed_status ( > > pk serial primary key, > > table_oid oid > > not null, > This is an interesting idea, however to write an Inbox widget sanely we > need to have a defined list of which tables are reviewable. Which does not preclude the above. We simply add a check constraint or even enumeration table. Or a trigger that checks on pg_class (can't use foreign keys on system tables, unfortunately).
> (presumably lab_result and doc_obj, in the first instance) Yes. > and then a view using SELECT ...UNION SELECT to provide a single view of > unreviewed objects. Absolutely. > > fk_reviewed_pk integer > > not null, > > fk_reviewer integer > > not null > > references public.xlnk_identity(xfk_identity), > > is_technically_abnormal boolean > > not null, > > clinically_relevant boolean > > not null, > > ) inherits (audit_fields); > I was looking for a timestamp field, but I realise we inherit modified_when > here. yes > Most importantly, we need to find clinical items 'destined' for a particular > individual, yes > but not actually seen yet, a third flag > seen boolean, We don't need it. At first I thought we do. However: No row in reviewed_status == not reviewed. Notice how this is not the provider inbox saying "Mrs Smith has lab results to be reviewed" ... > comment text, added Karsten -- GPG key ID E4071346 @ wwwkeys.pgp.net E167 67FD A291 2BEA 73BD 4537 78B9 A9F9 E407 1346 _______________________________________________ Gnumed-devel mailing list [email protected] http://lists.gnu.org/mailman/listinfo/gnumed-devel
