Re: [Tracker] Metadata search Tracker Gnome

[Martyn Russell <martyn lanedo com>]

On 29/09/11 04:21, John Bestevaar wrote:
> On 28/09/11 18:50, Martyn Russell wrote:
> > On 28/09/11 00:36, John Bestevaar wrote:
> > > Hi Martyn and Team :-)
> >
> > Hi,
> >
> > > Do i understand correctly that you propose to add to Tracker,
> > > tracker:fulltextIndexed true; for each mlo class property, as a function
> > > that indexes the texts and stores them in a database somewhere on the
> > > users machine and that it is this which will make the information
> > > retrievable via the Tracker GUI search? =-O
> >
> > Let me try to give some background first. Right now there is one
> > database with all information in it. There used to be 2. One for
> > metadata and one for FTS data. Now the FTS data is just some tables in
> > the same DB.
> >
> > Now, if you query for "foo" in particular ontology properties and
> > classes like mlo, then it will work *now* already. However, if you
> > want to find the word "foo" in *ANY* data related to a file, you have
> > to be clever in the query and use either search for "foo" against
> > EVERY metadata derivative in every class related to the file or you
> > can use fts:match (which will match on a class level and also allows
> > you to perform additional FTS based functions). The fts:match allows
> > you to do it per class generally, not per property. There would still
> > need to be some work to check mlo class information in the client
> > anyway as it isn't there now.
> JB; OK my asking a lot of ignorant questions and reading your
> explanations is satisfying for me but wastes your time and does not
> really advance your work much. Is there somewhere other than the Meta
> Tracker home page where i can go to read more documentation? :-D

We do have some, but it probably won't give you all you need and could 
always be improved:

  https://live.gnome.org/Tracker/Documentation/Examples/SPARQL/FTS

Don't worry about my time :) I can always not respond :P

> > Adding tracker:fulltextIndexed to particular mlo class properties
> > wouldn't be so much of a hit IMO (not like adding it to
> > nie:plainTextContent which is the content of files and much later in
> > general).
> >
> > > From my point of view, having all/any image metadata available to a
> > > Tracker Gui text search is very useful and very desirable. :-D
> > > For example: mlo; city property: My photo collection ( early days ) has
> > > several tens of images with about a dozen different "city " locations.
> >
> > Indeed, we are really just a test case to show how to use Tracker. We
> > would prefer applications are integrated with it than have one app for
> > every type of search. But we could add a panel with image metadata sure.
> > It's just about finding the time to do it.
> JB: Could you explain the advantages of integrating Tracker with desktop
> "Applications " as distinct from desktop "Operating Systems" and CMS on
> the web, as they are not immediately obvious to me? Why not as a
> standalone application?

There is very little difference. It's just about the content that gets 
indexed and translated in the end. With CMS, you would have web pages or 
potentially database information to index (depending on implementation - 
and I have little to no experience with CMS'). I don't really see the 
difference between "Applications" and desktop "Operating Systems" other 
than perhaps Applications save user's data to the user's directories 
(which we primarily index) and operating systems have shared information 
(like applications available, icons, documentation, etc) which we don't 
really index so much. We can but by default it's not set up that way 
(other than for applications).

Not sure I answered your question ... :)

> > > Finding,...via Tracker Gui,... which photos were taken at which
> > > location( Name in text form) would save me having to manually organize
> > > my collection (which will get big) into extensive album trees. 8-)
> >
> > Indeed. You could also use tags which would work too once my
> > tracker-needle-tagging-improvements branch has been applied to master.
> JB: Yes, Digikam has tags including tag trees and Nautilus among others,
> already has tagging .

The nautilus tagging was written by us of course ;) unless you mean 
something else and are calling it tagging?

> But it seems to me neither of their search engines can find the other's
> tags. This is one of the reasons i went in search :-P of a search
> application independent of all. This is also one of my reasons for my
> supporting image metadata, as a way of giving images a handle or handles
> (better than filenames) that other engines and applications can use.
> Yes.. i am also aware that "Applications " sometimes behave like naughty
> kids and strip all the metadata without telling you about it. 8-)

Interesting point. So the issue here is that you have to write tags back 
to files when they're updated. This requires not only reading but 
writing every file format you support. This is not trivial for all 
cases. Tracker does have a tracker-writeback binary with support for 
some formats, but no where near as many as we can index. If you use 
"shotwell" the image management application, you will notice that it 
supports reading/writing tags too (and I think there is some tracker 
plugin for it somewhere).

Stripping metadata is not really about being naughty :) it's more (I 
suspect) about applications not having FULL support for writing formats 
X, Y and Z and as such, things get left out when they save any changes 
back to the file.

--
Regards,
Martyn

Founder and CEO of Lanedo GmbH.