Re: baloo metadata scrapers
- Date: Fri, 7 Oct 2016 13:47:04 +0200
- From: Vishesh Handa <me@xxxxxxxxx>
- Subject: Re: baloo metadata scrapers
On Oct 7, 2016 12:07, "Martin Bednar" <serafean@xxxxxxxxx> wrote:
> thanks for answering.
> On Friday, 7 October 2016 10:44:02 CEST Vishesh Handa wrote:
> > Hey Martin
> > Baloo is only designed for searching for files. It is not a meta-data
> > store. If the data you're extracting is present in the file, then it
> > could potentially be added to Baloo.
> It is not, I get that data from an external source (currently themoviedb).
> > The important thing to look at is how to present the data, and how
> > does the data help in searching. For example - providing the name of
> > an actor in a movie could make the movie popup when searching for that
> > name. This is great, however, the user needs to be given a clear
> > explanation of why that file is in the search results.
> Agreed, the dream actually is creating something along the lines of the
> multimedia library described on the nepomukport page for potential use in
> Plasma Media Center.
> I found https://community.kde.org/Baloo/Architecture since my email, and am
> now thinking of creating a new data/search store plugin.
Actually ignore all docs you find online on any wiki. It is too old. The git repo is the only thing that is valid.
> > Our main search interface, krunner, isn't designed to provide this
> > additional context. Therefore, adding actor information, is not
> > something I would want to ship.
> I'd like to ask for a few pointers though :
> Does documentation on how to write new search/store plugins exist?
> From what I understood, I'll have to take care of storing the data myself. I
> see that baloo uses LMDB, Xapian and SQLite. Any pointers as to which db
> system (or a completely different one) is well adapted to this kind of use? At
> the moment I tend to lean towards LMDB, but have no real technical argument
> for it...
Basically I would recommend looking at this from the presentation and usability side first. Technology later.
Do you want some kind of kioslave? Through Krunner? Something else? What kind of workflow and what kind of users.
Once this is more clear, you can figure out which is the best way to go about what you want.
>  https://community.kde.org/Baloo/NepomukPort#Bangarang