[mythtv] MythMusic database schema suggestion

[Joseph A. Caputo]

I agree... while I suppose the DB schema could be tweaked, it's not the
bottleneck,  nor is the fact that the entire metadata DB is kept in memory.
At least for the playlist editor, I think the initial delay is due to the
fact that the list item widgets for the entire tree (all sub-levels) is
being constructed at once, rather than constructing levels of widgets on
demand.  I don't claim to know what kind of performance you'd get out of
traversal between levels if you built them on demand, but I'd be interested
to do a comparison.  I'd bet you'd get a significant speed-up of the initial
playlist editor display if you even just built the whole tree down to the
album level, then added the track listings on demand.  Unfortunately, I
don't have a development box ATM, or I'd do it.  (might have to break down
and get an iBook, as the long-awaited new 15" G5 PowerBook is still not on
the horizon...)

-JAC

> -----Original Message-----
> From: mythtv-dev-bounces at snowman.net
> [mailto:mythtv-dev-bounces at snowman.net]On Behalf Of Matt Zimmerman
> Sent: Tuesday, June 24, 2003 4:27 PM
> To: Development of mythtv
> Subject: Re: [mythtv] MythMusic database schema suggestion
>
>
> On Tue, Jun 24, 2003 at 03:06:52PM -0500, Brian Lalor wrote:
>
> > Matt Zimmerman wrote:
> > >On Tue, Jun 24, 2003 at 12:16:59PM -0700, Chris Petersen wrote:
> > >
> > >
> > >>on that note, many smaller DB queries are a much more efficient option
> > >>than loading the whole thing into memory
> > >
> > >
> > >Not when you need all (or most) of the records anyway.
> >
> > I'm gonna go out on a limb here and say that you *never* need
> all (or even
> > most) of the records, unless you've got a playlist with all of
> the tracks
> > in the database in it, and even then its just one query.
> Otherwise you're
> > just wasting memory.
>
> True enough, 10,000 rows of musicmetadata might cost you a whole
> 5M or so of
> memory, representing up to 1% of the physical RAM in a modern PC.
>
> Has anyone attempted to demonstrate that the database structures
> represent a
> bottleneck?  I can fetch all of the rows out of musicmetadata in
> well under
> a second from the command line.
>
> --
>  - mdz
> _______________________________________________
> mythtv-dev mailing list
> mythtv-dev at snowman.net
> http://lists.snowman.net/cgi-bin/mailman/listinfo/mythtv-dev