extragear/multimedia/amarok
Martin Aumueller
aumueller at uni-koeln.de
Mon Jul 24 19:45:09 UTC 2006
Ok, here are some hard facts: loading 6700 (mostly mp3) tracks into the
playlist took 205 s with r565899 on a system with a fast disk. After
modifying MetaBundle::setUniqueId() to return immediately w/o doing anything
times dropped to 45 s for the same operation. Caches were cold in both cases.
This is not good. As it is right now, this does a complete collection rescan.
And another bad thing: this meta data reading is done from within amarokapp
proper. Should problems with taglib arise (the original motivation for having
a separate collection scanner process), amarokapp will crash - and not just
the scanner.
On Mon July 24 2006 21:01, Martin Aumueller wrote:
> There is still a downside to ATF in its current state: every time a
> MetaBundle is created, its corresponding file will have to be read for
> finding the unique id. This will slow down the loading of large playlists
> considerably. Why not just store the unique id in the database?
>
> On Mon July 24 2006 20:37, Jeff Mitchell wrote:
> > SVN commit 565899 by mitchell:
> >
> > 1) Use all ATF features that only require reading the UID at all times.
> > GUI option now controls whether new UIDs are written to files. Note that
> > this helps generalize ATF; we can use this to now enable other writes to
> > metadata if we wish and have it be a global "yes you can touch my files"
> > setting. 2) Only force a rescan/playlist clear the first time ATF is
> > turned on (err, since this commit...) 3) If Amarok is closed before
> > rescan finished, don't force it on restart...takes too long for some
> > people. 4) ChangeLog updates.
> >
> >
> >
> > M +7 -0 ChangeLog
> > M +4 -4 src/Options7.ui
> > M +2 -1 src/Options7.ui.h
> > M +5 -1 src/amarokcore/amarok.kcfg
> > M +19 -22 src/app.cpp
> > M +5 -16 src/collectiondb.cpp
> > M +32 -23 src/metabundle.cpp
> > M +3 -5 src/playlist.cpp
> > M +3 -3 src/playlistitem.cpp
> > M +2 -4 src/playlistloader.cpp
> > M +2 -2 src/scancontroller.cpp
>
> _______________________________________________
> Amarok mailing list
> Amarok at kde.org
> https://mail.kde.org/mailman/listinfo/amarok
More information about the Amarok
mailing list