Re: [Tracker] Incomplete results + lack of integration in Gnome for contents searching

[Carlos Garnacho <carlosg gnome org>]

Hi Jean-Christophe!,

On Thu, Dec 22, 2016 at 9:19 AM, Jean-Christophe Baptiste
<jc phocean net> wrote:
> Hello,
>
> I started to use Tracker intensively as a replacement for Recoll. It is nice and better integrated to 
> Gnome, but the results are still less complete.
>
>
> 1) For some reason, some files are correctly indexed but not shown in the result.
>
> See my screenshot enclosed :
>
> - the settings show paths that are indexed.
> - I search a term, with a limit of 100 files,
> - The expected file (and around ten others) does not show in the results
> - Checking the indexation manually on this file shows that the indexation worked.
>
> What is going on ? Shall I open a bug report or am I missing something ?

That looks indeed strange, most probably down to tokenization of this
document text having gone wrong somehow. Quickly checking here, I see
Tracker doesn't handle too well the the d'match french form, the other
appearances should still result in a positive match though.

Please file a bug and we take it there. One first question for that
bug, does looking for any other terms inside these documents result in
match? or are these files entirely invisible to full-text search?

> 2) I have another question : tracker-needle works but is quite dated. Is there a plan for a better GUI ?

Not from the Tracker side... tracker-needle was merely a showcase app,
it helped Tracker gain momentum, but nowadays the focus is in making
Tracker a better platform, and let the apps do the talking.

It also matches better the original Tracker vision, the purpose to
centralize the data was to make it widely available to applications,
those may have more contextual UI and search than a generic app can
have.

... And well, sure, lack of manpower is also involved. I'd however
advocate to make it a standalone project, as I'm actually looking into
splitting/sanitizing Tracker, tracker-needle will probably keep aging
in a separate repo unless someone picks it up.

> The integration in Nautilus is suprisingly not good, and it is also disappointed to see that it is not used 
> by gnome-shell in the activity view.

Nautilus implements search in 2 ways:
1) A fast path, based on Tracker. The query performed however chooses
to restrict to filename searches. And obviously, only works inside
indexed directories.
2) A slow path that works pretty much like find, and gets deeper in
the filesystem, again purely filename based, as there's no indexed
contents.

I think there are plans to show snippets in nautilus search, which
implies search within file content, this would bring nautilus search
on par with the "tracker search" tool. This is however something to
talk on #nautilus, or their bugzilla, as it's a matter of how do they
use Tracker.

On top of that, gnome-shell has these "search providers" as the
extension point. Nautilus implements one, which uses the fast path
above and hence has the same limitations.

> I am aware of a gnome-shell extension, but it does not work on recent releases and the code looks a bit as 
> a hack.
>
> Why not using Tracker as a search provider by default for file contents ?

gnome-shell search providers make sense with a backing application to
jump into when you click on the results. With the right tweaks
(changing the nautilus query so it doesn't do filename matching is
removing one line of code), nautilus has got everything to be that
application for generic file search.

Cheers,
  Carlos