Nepomuk Metadata Extractor moved to KDE Review

Jörg Ehrichs Joerg.Ehrichs at gmx.de
Tue Dec 4 13:17:33 GMT 2012


> This sounds a bit like you're ignoring most of the feedback you got.

No didn't ignore any reviews. I've changed/implemented each and every
suggestion/review I've got.
Except the few opened points you mentioned that are necessary to get
it into 4.11

As these points mean, I'll have to change /add stuff to nepomuk to
make the integration better, this isn't something,
that's possible at the moment. That's why I like to postpone these few
changes, get this in extragear and "fix" all the
last mentioned integration issues for  4.11.

> One more thing I ran into when installing it (already three weeks ago) is that
> it uses a lot of dependencies, many of them python libraries to interact with
> specific webservices. Many of them are not mentioned specifically by CMake, as
> their runtime dependencies, so it took me quite some time to figure out those
> dependencies.

I wanted to add these stuff to CMake first (so you get notified that
these packages are missing)
But I found a few ml discussions saying, that runtime dependencies
should not be added to cmake
(via find_package). If you disagree, I could try to make add some
FindCMake files for the python stuff.

But i think this is not the best way to deal with it.
Currently there is a README, explaining all the dependencies.
I wanted to add some more help for packagers too (mainly rewrite whats
in the readme but with more details)

> I also see this as a problem for a possible KDE SC integration. Without a lot
> of external dependencies, the project is fairly useless.
That's true, without the python dependencies no data will be fetched
and basically the program does nothing.
But this wouldn't break anythink in this case too, and the "normal"
nepomuk-fileindexer needs external libs too.

So is it really a big deal to depend on these libraries for
runtime/workspace or wherever this could go in 4.11?

> I'm also not sure we
> can sensibly depend on PyKDE in modules as central as kde-runtime or
> workspace.

As the tvdb-mail plugin is the only one that needs it, this plugin
could stay maybe in extragear in a repo for additional plugins
or maybe I could add GHNS support later on and put it there.

>
> Do you see a way to a) reduce the number of dependencies or at least make it
> more obvious how to install nepomuk-webminer in a more obvious way?
>

The build dependency for poppler and TagLib will be removed when I
integrate this more into the fileinder (or better reuse the
newly written fileinder which mostly duplicates my current approach anyway)

For the python stuff I could combine requests/httplib2/urllib2 as far
as i know, as they du basically the same.
And of course add some "PACKERS-INFO" or "INSTALL" readme to make all
the other runtime dependencies from the current
README more prominent.

Would this work for you?

Regards,
Jörg




More information about the kde-core-devel mailing list