Re: [Gnumed-devel] branching a test branch

[syan tan]

now that 8.2 works with child tables (after vacuum analyse , i think ),
the performance bug with clin.v_pat_narrative was traced to 
dem.v_staff; actually, there isn't a need to use dem.v_staff in
the query; use dem.staff instead, and the performance is less
than a second (vs 10 minutes).


On Sat, 2006-12-09 at 23:23 +0100, Karsten Hilbert wrote:
> On Sat, Dec 09, 2006 at 08:55:03PM +1100, syan tan wrote:
> 
> > I was wondering if it would be ok to create a test client branch;
> Sure.
> 
> > However,  8.2 still doesn't  have the child tables inheriting parent
> > indexes during a query plan,
> According to their Release Notes I am not really sure it
> does, either.
> 
> I am sure you did but let's just ask it anyways: you did do
> all the usual vacuum analyze dance and all after the
> transfer, didn't you ?
> 
> > and it took 645000 mseconds (10 minutes)
> > for a record retrieval , executing a select using the v_pat_narrative
> > view ! 
> unacceptable by any means
> 
> > On the other hand, the  test old client can still do it in less than 2
> > seconds, by explicitly joining the child tables , instead of just the
> > parent table
> > (this is clin.clin_root_item). So maybe I should stick the currently
> > good performing client in a test branch, in case someone else wants
> > to try out gnumed with a large database.
> Surely. I think we need to find a solution in the *real*
> branch.
> 
> Karsten