New Dynamic Playlist

[Bart Cerneels]

On Sat, Jun 4, 2011 at 19:51, Ralf Engels <ralf-engels at gmx.de> wrote:
> Hi Myriam, Valorie
>
> I am wondering.
>
> Which of the many improvements suck?
>
> - The example biases which demonstrate to a novice user some of the
> features and are instantly usable?
>
> - The removal/merging of the three bias categories into one short list?
>
> - The simplification into a tree-widget that supports drag and drop and
> all the other stuff that users are already familiar with
>
> - The new "search" bias that allows to easily convert a search into a
> bias
>
> - Or some of the other smaller improvements or bug fixes?
>
> Anyway, as long as no-one could lay a finger on a concrete usability
> issue I will just continue bathing in the soft and sensual pink light of
> a job well done.
>
> Cheers,
> Ralf
>
> On Sa, 2011-06-04 at 12:00 +0200, amarok-devel-request at kde.org wrote:
>> Hi all,
>>
>> On Sat, Jun 4, 2011 at 11:21, Valorie Zimmerman
>> <valorie.zimmerman at gmail.com> wrote:
>> > Sucks.
>> >
>> > In a word. I find it unusable, and would rather see the old one back
>> > again, if this can't be made 1. discoverable and 2. accurate.
>>
>> I must agree, the usability sadly has not improved, it is still far
>> too nerdy and unsuitable for the average user.
>>
>>

I personally don't want to belittle Ralph's work at all. I think the
new treeview for the dynamic playlists look very nice. Much respect
for that. And I'm sure if you understand the concepts well that it's
nice to use.
However I think we will not get anywhere with trying to use the
dynamic playlist Bias system in it's current form. The concepts are
just to hard to grasp for anyone not blessed with the gift of
understanding the codebase. In short we need to simplify it.

Here is what I propose:
- re-introduce  smart playlists, not dynamic but just build with a
query that will return as every match from the collections. This
should suite the usecase of a subset of dynamic playlist usage
already.
- On top of the Biases add a very simple similar playlist which will
be based on whatever is in the queue already. Ex: similar from Last.fm
based on tracks; similar from echonest based on artists currently in
the queue; similar tracks using moobar data.
- A random mode/playlist category. With proportional bias towards
genre, rating, decade and age, no more for the sake of simplicity.

So in short: split up dynamic playlists in favor of user friendly,
easy to understand playlist concepts.
At the same time remove APG for obvious reasons and maintenance.

I've voiced this idea here in Randa already and it seems to be very
much accepted.

Bart