Re: [Tracker] 0.6.4 now in svn - awaiting testing before release

[jamie <jamiemcc blueyonder co uk>]

On Tue, 2007-10-09 at 21:13 +0000, yelo_3 wrote:
> Maybe there is another way to do this (but I don't know the exact tables in tracker): add to the sql table 
> a new column, that is "valid". whenever the search returns a non valid entry (deleted file, or old file 
> content) it should mark the row as "invalid".
> at every startup (or shutdown) the invalid rows can be deleted.

we do that on search - we now delete hits on the fly that have no
corresponding file record

deleted files are deleted from the tables but not the index


> In this way the tracker DB will remain consistent, and the heavy process only happens at startup (or 
> shutdown) time. What do you think?
>
> remember that another bug similar to this is when the filename is changed!

filename is metadata too so its the same bug


> ----- Messaggio originale -----
> Da: jamie <jamiemcc blueyonder co uk>
> A: yelo_3 <yelo_3 yahoo it>
> Cc: Laurent Aguerreche <laurent aguerreche free fr>; Tracker-List <tracker-list gnome org>
> Inviato: Martedà 9 ottobre 2007, 23:01:20
> Oggetto: Re: [Tracker] 0.6.4 now in svn - awaiting testing before release
>
> On Tue, 2007-10-09 at 20:22 +0000, yelo_3 wrote:
> > Jamie wrote:
> > > we now delete dud hits so if you research again the correct totals
> > > should appear
> > >
> > > hit counts are estimates only and may include deleted items - as
>  soon
> >  as
> > > a deleted item is detected when you click on a category it is then
> > > deleted from the index so a research should show correct totals
> > >
> > > I guess we can refresh the hit counts after searching
> >
> > remember that if you don't delete the indexed content when a file is
>  changed/deleted the results will contain unwanted data.
> > This is a simple example, I've already written this, but I'd like to
>  know your opinion, and maybe find a way to workaround this, without
>  modifying the index, as it is a long process as you told us.
>
> when a file is deleted the index contents wont hit anything (the ID of
> the doc is no longer in the db)
>
> when changed the difference in contents is worked out and applied
>
>
> > Example: when a file a.txt is modified, the index still contains the
>  old file content. So if I search for the old file content, between the
>  results I will have a.txt, even if the content is completely different.
> > Possible workaround: check the file content before adding it to
>  results. maybe it is too heavy...
>
> I can replicate that - its a bug in the diff that must be fixed
> urgently... (and no its not heavy it should be doing that!)
>
>
>
>
>
>
>
>
>       ___________________________________ 
> L'email della prossima generazione? Puoi averla con la nuova Yahoo! Mail: 
> http://it.docs.yahoo.com/nowyoucan.html