Suggestion to avoid mandatory steps when browsing

[simoncn]

(16-08-2012 09:51)winxi Wrote:  

(16-08-2012 08:55)simoncn Wrote:  Removing [part] would bring the less efficient path into play for all albums. For a small library, this wouldn't matter. For a large library, I'm not sure what the impact would be. I would need to measure the impact and possibly do further optimization work to compensate.

I see, scanning efficiency is of course a very important point. Let me summarise, i hope that i understand everything right:

1) Now, files of the same album in different folders are automatically split. This is useful for different albums with the same name but it is a problem for users with certain folder structures. For such users, it can be 'worked around' by using [part] or by changing the folder structure. Moreover, the current behaviour is useful for reasons of efficiency.

2) removing [part] would mean that no album would be split automatically because of files residing in different folders. For different but equally named albums (which is clearly an exceptional case) the workaround would be using the already existent feature 'match filter'. This would meet the expectation from a tag-based media server but it would also lead to a less efficent scanning.

This implies a tradeoff between efficiency and user experience. For me personally, efficiency is very important. If you decide to keep 1), it would be a good idea to clearly point out in the user guide that files from one single album have to reside in the very same folder (except for multidisc and [part]).

I agree that efficiency is very important. It's also very important to have clean semantics that are natural and convenient for users and don't cause surprises. Ideally it would be possible to achieve both of these.

Before making a final decision on this, I'll need to implement a prototype and do performance measurements for speed and memory. I don't have a large library myself, so it would be helpful if someone who has a large library could offer to run a performance comparison using the prototype and send me their results.