Re: [Gnumed-devel] bash backup script

[Karsten Hilbert]

> however, I don't understand the distributed_db , db, and config tables 
> roles.
They define where to look for the other DBs provided the one
you are looking at is the primary one (service config). It
works but isn't thought out very well yet, the tables aren't
elegant and it doesn't feel very robust at all. IOW it needs
work :-)

>  Has it been tested that
> the gmPG will work with non-pyPgSQL dbapi ?
Not to my knowledge. I expect quite a few things will break.

>  I don't think the 
> cursor.execute( stmt, args) works
> but execute( stmt % args) works in pgdb, but then quotation of string 
> args is needed , which isn't
> needed when using pyPgSQL's cursor.execute(stmt,args).
Any db api adapter that still doesn't do execute(stmt, args)
is fundamentally flawed. Starting with v7.4 PostgreSQL
provides out-of-band parameter passing which entirely removes
the need for escaping values (since the data doesn't have to pass
through the 7-bit SQL parser engine). Thus the overhead for
sending even BLOBs (bytea) is kept to a bare minimum (and is
O(1)).

> There are also cfg_...  ui configuration tables I believe; what are they 
> for?
For ui/functionality configuration as you seem to have
understood already going by the above question ?!? I do think
we might want to remove the str_array table such that we can
get rid of the PgArray -> pyPgSQL dependancy. It's probably
cleaner, too, relationally.

> Attached also is the gnumed.log files when I try to run the client, 
> which after running
> the create_gnumed_db.sh , seems to partially work. (at least it gets to 
> the gui, but
> there is no demographics tab).
That doesn't need to have to do with the DB let alone the
bootstrapping at all. Pure logic dictates that conclusion,
don't you think ?

> How essential is it to have the distributed db, db and config tables ? 
Moderately. For now, *having* the possibility needs to be
there IMHO but it seems allowable to assume monolithity at
some points here and there until post-0.1.

> They don't seem to offer horizontal fragmentation, which might be useful 
why not ?

Karsten
-- 
GPG key ID E4071346 @ wwwkeys.pgp.net
E167 67FD A291 2BEA 73BD  4537 78B9 A9F9 E407 1346