Re: [Rhythmbox-devel] Ignoring Duplicates

[Mika Wahlroos <mika wahlroos iki fi>]

Hi,

Disclaimer: I'm not a Rhythmbox developer.

16.07.2011 05:57, Mark Aitchison wrote:
> Can Rhythmbox be made to ignore duplicate entries in the library (e.g.
> when the same track is found via links, or my laptop is sharing folders
> that have collections that overlap with my desktop computer)?

AFAIK it already deals with symbolic links that point to the same file
in the end.

Detecting tracks that are the same to a human listener but are in
different files is a different problem, and I won't touch on it except
to say that it's not trivial.  There could easily be e.g. different
editions of the same track (think different performances of a jazz track
for example) that are really different yet may be under the same title
by the same artist.  A simple fuzzy comparison of file names would
produce false positives.

There's a similar issue with determining whether two tracks with the
same title appearing on two different albums are really the same track
or not.  For example Spotify deals with this by showing such tracks only
once in search results and allowing the user to expand the single result
to also show the other tracks with the same artist and title. I'm not
sure if the data model in Rhythmbox is up to doing something like that
efficiently, though.

> Slightly relating to this, I find that rhythmbox creates filenames and
> directory names with characters that really should be stripped out

This isn't exposed on the UI, but it's already possible to have
punctuation and whitespace stripped when new files are created.  In the
older 0.13.x version I'm using the setting is in gconf at
/apps/rhythmbox/library_strip_chars.  I'm not sure about the
corresponding setting in the Gnome 3 version that's more actively developed.


- Mika