> On March 18, 2012, 9:39 p.m., Ralf Engels wrote:
> > Are the unit test running?
> >
> > There is one specific scenario that the current scanner is handling (since
> > years. No new change).
> >
> > ITunes stile collections put tracks from the same album but different
> > artist into different directories like this:
> >
> > Michael Jackson
> > + Bad
> > + Smooth Criminal
> > Quincy Jones
> > + Bad
> > + Commentary
> >
> > If your fix doesn't break the unit test I am all for committing the fix.
> > Also you might want to make a new unit test to prevent somebody later on to
> > change the behaviour for the worse.
>
> Alexey Neyman wrote:
> Ralf,
>
> I wasn't able to run the unit test; even though I installed google mock
> 1.6.0 ("apt-get install google-mock") on Kubuntu 12.04, cmake does not detect
> it:
>
>
> -----------------------------------------------------------------------------
> -- The following REQUIRED packages could NOT be located on your system.
> -- You must install these packages before continuing.
>
> -----------------------------------------------------------------------------
> * gmock (1.4 or higher) <http://code.google.com/p/googlemock/>
> Used in Amarok's tests.
>
>
> -----------------------------------------------------------------------------
>
> For the same reason, I didn't add a new unit test.
>
> But I am afraid this is exactly the behavior I am trying to fix. With
> this change, your example would result in two (or more separate albums):
> "Michael Jackson - Bad" and "Quincy Jones - Bad".
>
> So, your change was fixing iTunes but breaking the heuristics Amarok used
> before. Could you think of a solution that would also handle the following
> scenario:
>
> ABBA
> + The Album
> + Eagle
> + Take a chance on me
> + One man one woman
> Chris Norman
> + The Album
> + As good as it gets
> + Every little thing
> + Red hot screaming love
>
> I.e., two completely separate albums being merged? One solution I could
> think of would be to merge tracks into single compilation if they all share
> the same year (is it the case with iTunes?). However, this is still prone to
> errors, for example if the tracks in both directories lack the year tag.
>
> Perhaps, make it a config option? Something like:
>
> [ ] Detect iTunes-style compilations
>
By the way, is it the case that if one had both albums from my example in
iTunes collection, they would also be merged in a compilation?
- Alexey
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
http://git.reviewboard.kde.org/r/104294/#review11582
-----------------------------------------------------------
On March 16, 2012, 12:13 a.m., Alexey Neyman wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> http://git.reviewboard.kde.org/r/104294/
> -----------------------------------------------------------
>
> (Updated March 16, 2012, 12:13 a.m.)
>
>
> Review request for Amarok.
>
>
> Description
> -------
>
> Older amarok used the following heuristics to determine whether a particular
> album is a compilation (from a comment in
> amarok-2.2.1/src/collection/sqlcollection/ScanResultProcessor.cpp):
>
> //using the following heuristics:
> //if more than one album is in the dir, use the artist of each track as
> albumartist
> //if all tracks have the same artist, use it as albumartist
> //try to find the albumartist A: tracks must have the artist A or A feat.
> B (and variants)
> //if no albumartist could be found, it's a compilation
>
>
> However, more recent Amarok versions started to merge different albums
> with different artist in separate directories together, as explained above.
> Amarok started to assume all albums with same name to be compilations
> (even if in separate directories) since the following commit:
>
> dfd8b457d7094144563c51b2528afdbe23ffc344
> Ralf Engels
> Fix all collection scanner auto tests.
>
> Now, amarok first scans all directories (sorting albums by the name)
> and then tries to process *album names*, one at a time. If it finds
> more than one instance of an album name, it assumes it to be a compilation.
> Thus, it lost the heuristics in employed before ("if more than one album
> is in the dir...").
>
> While it is still possible to force the right behavior
> by selecting "Do not show under Various Artists" for each of the erroneous
> albums, it would still be better to restore the original heuristics as there
> may be lots of albums merged this way. I think the old heuristics made sense
> (why would albums be put into separate directories otherwise, if they are
> a single compilation album?).
>
> The attached patch restores the following logic: If any given directory
> contains tracks that were sorted into a single album and and that album
> was not created as a compilation (i.e. it has non-empty artists), this
> album is excluded from being merged with other albums to create a
> "compilation".
>
>
> Diffs
> -----
>
> src/core-impl/collections/db/ScanResultProcessor.cpp 4f02a16
>
> Diff: http://git.reviewboard.kde.org/r/104294/diff/
>
>
> Testing
> -------
>
>
> Thanks,
>
> Alexey Neyman
>
>
_______________________________________________
Amarok-devel mailing list
Amarok-devel@kde.org
https://mail.kde.org/mailman/listinfo/amarok-devel
