I\'ve tried to get into amarok several times, and each time have been forced away by what seems to be profound breakage in the algorithm used to build collections and determine the titles of pieces.

My collections directory has a very regular structure:

Composer/Work/tracknum-blah.mp3
...

e.g. Beethoven/Appassionata/2-Andante.mp3

Any level (but generally at the top level) may consist of symlinks into per-host-mounted subdirectories, so the Beethoven item above might really be

Beethoven -> .amaterasu/Beethoven

Every file has meticulously maintained ID3 tags.

Now try as I might I cannot get amarok to do the obvious with this structure. With recursion turned off, it ignores all the symlinks and I get a nearly empty collection: with it turned on, it recurses down the dot directories too, and counts everything twice. If there\'s a loop anywhere, the entire collection process loops and never terminates.

(A side-note: this is really rather poorly thought out. Every Unix text editor in existence checks for file identity by comparing (dev, inum) pairs: is it too much to expect amarok to do the same? Equally, cycle detection is not hard: find -follow can do it in less than fifty lines of code...)

On top of that, amarok does... peculiar things with the filenames. Instead of trusting the ID3 tags, it tries to parse the filenames apart, leading to guesses that perhaps *this* file is really from an album called `02\', and *this* one is perhaps from an album called `The\' (containing, er, every song with `The\' in its filename in the collection) and so on. I can\'t see a way of getting it to ignore the damned filenames, or even to disregard elements of them.

Is there a way to get amarok to do what I want? Is the collections pane supposed to be polluted with millions of ridiculous misinterpretations of your music, or am I missing something?

(This is all with the latest amarok from CVS, but I\'ve seen it in 1.2.x before now.)