[prev in list] [next in list] [prev in thread] [next in thread] 

List:       kde-devel
Subject:    Re: Noatun Playlists - Better random song selection
From:       Benjamin Meyer <ben () meyerhome ! net>
Date:       2003-04-26 12:04:03
[Download RAW message or body]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Saturday 26 April 2003 12:47 am, Brendon Higgins wrote:
> Noatun kicks arse. I usually just turn it on, stick it on random over my
> whole collection, and let it rock away. Though I've noticed that the way
> songs are selected at random during play is a little, well, too random. I,
> and I assume most people, would like it to work similarly to the shuffle op
> in the playlist, but without actually shuffling anything, and still being
> random with songs that have been played a while ago - like repeatedly
> shuffling the tail-end of the list, so to speak.
>
> Currently, the way the next song is selected is completely random with
> every song having the same probability of being selected, thought this is
> not a very "natural" way of doing this. The natural way would be for a song
> to be selected based on things like how long it's been since it was played
> before, relative to the other songs. It should be the case that after a
> while the number of times each song has been played should all be roughly
> the same.
>
> It came to me the way to do this would be to have weightings associated
> with each song, and have those weightings proportional to how recently the
> song was played previously, or how many times the song has been played
> relative to the average of all songs. So I wrote a program to analyse some
> statistics of such methods (code attached; there is minimal commenting - if
> something isn't obvious just mail me), the most important statistic being
> the number of steps between each play of each individual song. This should
> be reasonably constant, ie. have a low standard deviation for each song,
> and thus a low average of all SDs for each song.
>
> The best way I've found is the method in SimpleWeightedShuffler (the third
> complete Shuffler type) which at first spreads weightings evenly over all
> the songs, and for each song that gets picked its individual weighting is
> added evenly to all the others, and then is set to 0. This way a song will
> never be picked twice in a row, and if a song hasn't been played in a while
> it becomes more and more likely that it will be played. The current method
> does not guarantee this, and I'm always feeling that there are a few songs
> I haven't heard in ages while there are others that Noatun seems to want to
> play ad-nauseum.
>
> I'm not sure if this is an area that anyone is interested in. It may be
> that this method is too bulky for some people, though it'd be fine for my
> purposes. I've had a short look at the split playlist source, but I decided
> I'd better let the pros handle it, first. I'm not likely to make a playlist
> plugin from this, but if anyone else would like to, great!
>
> BTW, I'm not subscribed to the list, so CCs would be appreciated.
>
> Peace,
> Brendon Higgins

Nope I am also interested in it.  In face a few years ago I wrote a program to 
do a very similar thing called Sondra.  There is a xmms plugin on the site, 
but I also made a NoaAtun playlist plugin.

http://www.csh.rit.edu/~benjamin/programs/program.php?program=Sondra

- -Benjamin Meyer
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.6 (GNU/Linux)
Comment: For info see http://www.gnupg.org

iD8DBQE+qnW21rZ3LTw38vIRAsiGAKCfEZk+EDx9R8OGdMNBBEp6PCeaTwCfRNJ2
a4W2E1o90zunjWw2FEyGTaM=
=03n2
-----END PGP SIGNATURE-----

 
>> Visit http://mail.kde.org/mailman/listinfo/kde-devel#unsub to unsubscribe <<
[prev in list] [next in list] [prev in thread] [next in thread]