Bringing KIO support to the Collection Scanner

[Maximilian Kossick]

On Tue, Oct 26, 2010 at 12:32 PM, Mark Kretschmann <kretschmann at kde.org> wrote:
> Hey all,
>
> as we are currently using TagLib, we have one huge limitation in
> Amarok: It cannot use KIO, but only locally mounted file systems for
> scanning.
>
> But there is a possible solution to this issue: In addition to TagLib,
> we could use "libstreamanalyzer" for reading tags, which supports KIO.
> This is the library that powers Strigi (KDE desktop search)
> internally. I talked to Jos Vandenoever, who wrote the lib, and he
> said this could work out pretty well.


How good are the strigi analyzers now? Back in '06 or '07 strigi did
not support all the fields that taglib is able to read. We still have
two analyzers in Amarok's source tree, and I'm pretty sure they are
not actually maintained. Fyi, these were written in Glasgow when
Martin and I talked to Jos about using Strigi/Xesam/Nepomuk/whatever
it was called back then to replace the collectionscanner. Some of that
code still exists in sqlcollection...

>
> Unfortunately there are no API docs online for libstreamanalyzer, but
> Jos said we could probably just run Doxygen over it, to get some docs.
> What we would have to do is implement a custom "Strigi::IndexWriter"
> for Amarok. Here is an example of such an implementation:
>
> http://websvn.kde.org/trunk/kdesupport/strigi/strigiutils/bin/xmlindexer/rdfindexwriter.h?revision=1104183&view=markup
>
>
> Jos said, all we have to is translating the addValue() calls to SQL.
> Which sounds doable, I guess (although we're all not exactly SQL
> experts).

More like write to an XML file, as we probably still want to be able
to run the scanner separately. Additionally, the tag reading of
Meta::File would have to be updated as well. Not sure where else we
are using taglib directly.

>
> How do you folks like this idea?
>