Re: [Tracker] Result of GetMetadata DBUS Method

[Adrien Bustany <abustany gnome org>]

Le Sun, 25 Sep 2011 10:02:41 -0400,
Nikolaus Rath <Nikolaus rath org> a écrit :

> Adrien Bustany <abustany-rDKQcyrBJuzYtjvyW6yDsg public gmane org>
> writes:
> > Le Sat, 24 Sep 2011 18:50:06 -0400,
> > Nikolaus Rath <Nikolaus-BTH8mxji4b0 public gmane org> a écrit :
> >
> > > Adrien Bustany
> > > <abustany-rDKQcyrBJuzYtjvyW6yDsg-XMD5yJDbdMReXY1tMh2IBg public gmane org>
> > > writes:
> > > > > > > In tracker 0.8.17, an example result of the GetMetadata DBUS
> > > > > > > method is
> > > > > > >
> > > > > > >  a nfo:PaginatedTextDocument ;
> > > > > > >         nie:title "SV Meldung" ;
> > > > > > >         nco:creator [ a nco:Contact ;
> > > > > > >         nco:fullname "nikratio"] ;
> > > > > > >         nie:contentCreated "2011-08-10T20:12:38Z" ;
> > > > > > >         nao:hasTag [ a nao:Tag ;
> > > > > > >         nao:prefLabel "()"] ;
> > > > > > >         dc:format "application/pdf" ;
> > > > > > >         nie:description "()" ;
> > > > > > >         nfo:pageCount 1 ;
> > > > > > >         nie:plainTextContent "blablabla" .
> > > > > > >
> > > > > > > With tracker 0.10.21, however, the same document now gives
> > > > > > >
> > > > > > >  a nfo:PaginatedTextDocument ;
> > > > > > >         nie:title "SV Meldung" ;
> > > > > > >         nco:creator [ a nco:Contact ;
> > > > > > >         nco:fullname "nikratio"] ;
> > > > > > >         nie:contentCreated "2011-08-10T20:12:38Z" ;
> > > > > > >         dc:format "application/pdf" ;
> > > > > > >         nie:description "()" ;
> > > > > > >         nao:hasTag ?tag1 ;
> > > > > > >         nfo:pageCount 1 ;
> > > > > > >         nie:plainTextContent "blablabla" .
> > > > > > > } } WHERE { {
> > > > > > > ?tag1 a nao:Tag ; nao:prefLabel "()" .
> > > > > > >
> > > > > > >
> > > > > > > I don't know enough about the syntax used here, so I won't
> > > > > > > claim that it's wrong. However, rdflib fails to parse this
> > > > > > > and thus breaks my application.
> > > > > > >
> > > > > > > Can someone tell me if this is a bug in tracker or in rdflib?
> > > > > >
> > > > > > Could you share the error message that rdflib gives you?
> > > > >
> > > > > Sure, it's at
> > > > > http://code.google.com/p/rdflib/issues/detail?id=190
> > > >
> > > > Most likely rdflib doesn't like the fact that the WHERE of the
> > > > INSERT is not closed?
> > >
> > > rdflib actually tries to parse this as Turtle RDF. But unless I'm
> > > reading the specs wrong, Turtle does not have a WHERE, so am I
> > > right to assume that GetMetadata has been changed to no longer
> > > return valid Turtle RDF, but partial SPARQL statements?
> > >
> > > If so, this makes it very hard to use GetMetadata() for anything
> > > else but tracker. Is there any way to get something that's easier
> > > to parse from tracker-extract?
> >
> > Well, tracker-extract actually *never* returned turtle, at least as
> > far as I know (now, I'm not completely familiar with that
> > component). I guess the fact that previous versions returned SPARQL
> > that looked like turtle is at most a coincidence :) (note: turtle
> > does not have INSERT either).
>
> Well, extract-metadata does not return an INSERT. Until the WHERE
> started to appear, it was always valid turtle.

No, I guess the INSERT is added in the FS miner

> > GetMetadata being an internal API call, it's not exactly
> > meant to be easy to use for 3rd party apps, and it's actually tied
> > quite tightly with the SPARQL the the FS miner generates. I'm not
> > sure of what would be the easiest way for you to use that API
> > now :/ Getting the complete SPARQL and inserting it into an
> > in-memory db (for example Sesame) before querying it would be an
> > option, but that really sounds complex and sub-optimal... A simpler
> > option would be to apply a few transformations to the SPARQL to
> > make it look like turtle again, but that does not sounds like a
> > bullet proof solution.
>
> Would it maybe be possible to add an additional API call, say
> ExtractPlaintext that returns just the (unstructured) plain text and
> could easily be used by 3rd party apps?
>
>
> Best,
>
>    -Nikolaus