Re: [Tracker] Using tracker extractors from other applications

[Martyn Russell <martyn lanedo com>]

On 17/11/10 02:59, Nikolaus Rath wrote:
> Martyn Russell<martyn-bhGbAngMcJvQT0dZR+AlfA public gmane org>  writes:
> > > I would like to use tracker's apparently quite extensive data-extraction
> > > facilities in my own program. It seems that "tracker-extract -f<file>"
> > > provides all the functionality that I want, but I am wondering if there
> > > is a way to avoid having to re-parse output.
> >
> > The -f command line is a convenience for testing. There is a d-bus API
> > you can use which tracker-miner-fs uses, for the spec see:
> >
> > http://git.gnome.org/browse/tracker/tree/data/dbus/tracker-extract.xml
>
> Will this work if the file that I'm interested is not included in the
> tracker database? Because I would like to use really just the extractor,
> without having the files being indexed by tracker.

Yes, it doesn't put the data into the database for you at all and the 
file doesn't have to exist in the database OR on disk ;) (i.e. it 
handles cases where the file is missing)

The miner-fs does the task of taking the data and inserting it into the 
database for you.

> > I should add, we use the "Fast" interface, which uses file descriptor
> > passing to avoid memory fragmentation and copious copying. There is
> > also a non-fast API you can use for testing purposes or if you're not
> > intending to send/receive a lot of requests anyway (though we
> > discourage it generally if possible).
>
> I will be using the Python dbus API, I'm not sure if this will allow fd
> passing. But requests will only be generated one at a time by explicit
> user interaction, so I hope that this will not be a problem.

Good question. Ivan any comment to add here?

The requests should be fine one at a time, there is no other way to do it.

> > > It would also be nice if there would be a way to process multiple files
> > > without having to start a new tracker-extract instance for every file.
> >
> > Using d-bus you shouldn't need to worry about the invocation of
> > tracker-extract.
>
> In that case I believe that I still need to start (and then stop) a
> tracker instance that will provide the dbus service though...?

No. Calling the d-bus API will automatically start tracker-extract for 
you if it isn't already running. With tracker-extract we also 
automatically shutdown after 30 seconds of inactivity, so instantiation 
must work from using the API.

--
Regards,
Martyn