Re: [Tracker] Delete objects from Tracker index

[Martyn Russell <martyn lanedo com>]

On 01/03/13 18:04, Frank Lahm wrote:
> Hi Ivan,
>
> 2013/3/1 Ivan Frade <ivan frade gmail com>:
> > On Fri, Mar 1, 2013 at 2:45 PM, Frank Lahm <franklahm gmail com> wrote:
> > > I know that this won't fix the index when files were moved or renamed,
> > > not deleted. And it seems there's no way to let Tracker reindex
> > > filesystem _without_ trashing the existing database (ie
> > > tracker-control -rs).
> >
> >   Restarting the miner should be enough. It should recrawl the file
> > system and check the information in the DB is up to date. If this is
> > not happening, it is a bug. Make sure your configuration is fine and
> > it is watching the directories (not just indexing them).
>
> Well, automatic indexing for files created is taking place, so I'd say
> config is fine. But still, these indexed file stick even though I had
> removed them a few minutes before when Tracker was not running. Otoh,
> in an earlier attempt it worked ok, after removing a file behind
> Tracker back, after restarting the file was removed from the index (at
> least it didn't show up as result in queries).
>
> [root fedora16 ralph]# yum info tracker | grep Version
> Version    : 0.12.10
>
> [root fedora16 ralph]# gsettings get
> org.freedesktop.Tracker.Miner.Files index-recursive-directories
> ['/Volumes/spotlight', '/Volumes/test', '/Volumes/spectest-ea',
> '/Volumes/spectest']
>
> [root fedora16 ralph]# ls -l /Volumes/test
> insgesamt 0
>
> [root fedora16 ralph]# tracker-sparql -q "SELECT ?url WHERE { ?obj
> nie:url ?url FILTER(regex(?url, '^file:///Volumes/test/.*')) }"
> Ergebnisse:
>    file:///Volumes/test/netatalk-stp4.tar.gz
>    file:///Volumes/test/netatalk-stp4.tar.bz2
>
> [root fedora16 ralph]# tracker-control -S
> Speichern:
> 28 Feb 2013, 05:25:37:  ✓     Store                      - Leerlauf
>
> Indizierer:
> 28 Feb 2013, 05:25:37:  ✓     Flickr                     - Leerlauf
> 28 Feb 2013, 05:25:37:  ✓     Dateisystem                - Leerlauf
> 28 Feb 2013, 05:25:37:  ✗     E-Mails                    - Läuft nicht
> oder Plugin ist deaktiviert
> 28 Feb 2013, 05:25:37:  ✓     Anwendungen                - Leerlauf
>
> [root fedora16 ralph]# tracker-control -t all
> Found 142 PIDs…
> Prozesskennung 10256 für »tracker-miner-flickr« gefunden
>    Prozess 10256 wurde beendet
> Prozesskennung 10263 für »tracker-store« gefunden
>    Prozess 10263 wurde beendet
> Prozesskennung 10272 für »tracker-miner-fs« gefunden
>    Prozess 10272 wurde beendet
> [root fedora16 ralph]# tracker-control -S
> Speichern:
> 28 Feb 2013, 05:25:46:  ✗     Store                      - Nicht verfügbar
>
> Indizierer:
> 28 Feb 2013, 05:25:46:  ✗     Flickr                     - Läuft nicht
> oder Plugin ist deaktiviert
> 28 Feb 2013, 05:25:46:  ✗     Dateisystem                - Läuft nicht
> oder Plugin ist deaktiviert
> 28 Feb 2013, 05:25:46:  ✗     E-Mails                    - Läuft nicht
> oder Plugin ist deaktiviert
> 28 Feb 2013, 05:25:46:  ✗     Anwendungen                - Läuft nicht
> oder Plugin ist deaktiviert
>
> [root fedora16 ralph]# tracker-control -s
> Indizierung wird gestartet …
>    ✓ Flickr
>    ✓ Dateisystem
>    ✗ E-Mails (Plugin könnte deaktiviert sein?)
>    ✓ Anwendungen
> [root fedora16 ralph]# tracker-sparql -q "SELECT ?url WHERE { ?obj
> nie:url ?url FILTER(regex(?url, '^file:///Volumes/test/.*')) }"
> Ergebnisse:
>    file:///Volumes/test/netatalk-stp4.tar.gz
>    file:///Volumes/test/netatalk-stp4.tar.bz2
>
> ... few minutes later, still there:
>
> [root fedora16 ralph]# tracker-sparql -q "SELECT ?url WHERE { ?obj
> nie:url ?url FILTER(regex(?url, '^file:///Volumes/test/.*')) }"
> Ergebnisse:
>    file:///Volumes/test/netatalk-stp4.tar.gz
>    file:///Volumes/test/netatalk-stp4.tar.bz2
>
> So the question still stands, how can I delete objects from the index
> programmatically.

Just a quick point I should make here, are you considering the 
tracker:available property? Because your queries will show files that 
are indexed but considered "hidden" or "unavailable" when volumes are 
unmounted.

If you care about files that are available "now" and not just in the 
database, you should definitely add the condition "?obj 
tracker:available true" to your query...

There are plenty of examples in the tracker code base ;)

--
Regards,
Martyn

Founder & Director @ Lanedo GmbH.
http://www.linkedin.com/in/martynrussell