AFT and file links

[Cody Christensen]

On Tuesday 22 April 2008 5:20:30 am you wrote:
> On Tuesday 22 April 2008, Cody Christensen wrote:
> > On Monday 21 April 2008 11:53:00 am Florian Loeffler wrote:
> > > Hello
> > > I have posted a question into the Amarok forum a couple of days ago (
> > > http://amarok.kde.org/forum/index.php/topic,15284.0.html ) and someone
> > > told me I should send an email to this address.
> > > So what I have asked about was the behavior of Amarok's  AFT feature
> > > regarding file links. The thing is, that probably most music
> > > collections will have tracks that occur on different albums but are
> > > actually identical (say on the album of its first release and on a
> > > best-of compilation). I personally find it not very efficient to treat
> > > these files as completely independent, it is not only a waste of hard
> > > disk space but also confusing since they have different statistics
> > > (e.g. individual playcounts), you have to rate each one seperately and
> > > if you create a smart playlist that looks for a certain pattern
> > > (labels?) you may have the same song twice or more times in your
> > > playlist.
> > > Therefore I was wondering if you could have all these different songs
> > > point to just one common file by using soft- or hardlinks. I do not
> > > exactly know how Amarok creates the unique IDs used for AFT, but since
> > > all of them would contain the exact same data, they should get the same
> > > unique ID, right? And this in turn would mean they have common
> > > statistics? I'm not sure if the problem with the several occurences in
> > > a playlist could be solved this way, though.
> > > I do not know how other people think about this matter, but I think it
> > > may be an interesting feature to consider for some future Amarok
> > > version. Right now, however, this is only my personal curiosity,
> > > because I have started to learn ruby and was thinking about useful
> > > scripts that could be written for Amarok.
> > >
> > > Thanks
> > > Florian Löffler
> >
> > I personally would love to see this feature of marking two songs as =.
> > Sometimes you have a song that is in an album as well as in another album
> > such as a sound track of a movie. If you don't want to delete one in
> > order to keep the integrity of both albums it would be great to mark
> > their stats (score, playcount, labels, place in playlists, etc) as equal.
> > Also sometimes they would have slightly different tags such as different
> > album tags and the method of marking them as equal should be able to
> > handle that IMO.
>
> This probably would not be an AFT thing.  For one thing, AFT hashes on
> (among other things) the tag of the file, so if the album is different then
> the hash will be different too.  In addition, part of the song data is
> used, and often the exact same song data is not on a best-of vs. a normal
> release (the volume is slightly different, the amount of silence in the
> beginning changes, etc). So even then it might not work.
>
> Probably the best way to do this would be outside AFT, for another reason
> as well.  It is similar to a request some others have given us where they
> have files encoded in i.e. FLAC and MP3 for playback on their home computer
> vs. their portable device...but they want don't want to see duplicate files
> in the collection, they want to only see one (the "higher quality one",
> which is actually not easy to determine, especially comparing lossy to
> lossy and lossless to lossless).  And they want the statistics to be the
> same regardless of which one they play.  In such a case, which pretty much
> matches yours, AFT would be totally useless.
>
> I encourage you to think about ways this could be done.  For instance, a
> collection option where files by the same artist and same song name (but
> different album) are listed once and share statistics.  But you'd also need
> to think about how the database would react to this...would you have the
> statistics table be able to have multiple Unique IDs per entry? How would
> it be updated without getting stale?  Etc.  I'm not sure that this can be
> done in a good, non-intrusive way, but if you submit a way that might work
> well, the developers could take a good hard look at it (but no
> guarantees)...
>
> --Jeff
You bring up some good points. I realize that it would need to be taken care 
of out side of AFT but where the original post described something very 
similar I just posted it as a reply there.

The way that is in my mind would be something like a spell chacker type 
interface. A user could run a check for duplicates search and it would look 
through the database and find songs with similar tags. It could display the 
details of these tags like a spell checker one above the other and on the 
side could be options. There could be options like mark as duplicate, ignore, 
mark top as high quality, mark bottom as high quality, etc. As for the 
statistics view what about listing the song by the data that is the same and 
then having a little [+] to explode it into the 2 (or more) duplicate 
listings details.

Excuse me if this is wrong or impractical I'm not a big programmer I only know 
PHP but, one way I could think about doing this on the database end could be 
to mark the songs that are duplicates with a hidden tag (or hash type thing) 
in the database that would identify them as duplicates. One problem I see 
with that is it could potentially cause a performance issue. I don't know 
what the performance consequences of checking for that tag when changing 
stats would be. and also if one's collection had many duplicates then it 
could potentially be a problem too.

>How would it be updated without getting stale?
Sorry but I honestly don't know what you mean here.

Maybe with some input from others we could run with an idea similar to this. I 
think for people with no duplicates it would be unobtrusive because they 
would need to activate it and run the duplicate finder for it to even be 
present so if they didn't like it they could keep it disabled.

Ps. You don't need to CC me.