from http://bugs.musicbrainz.org/ticket/3073

The goal is to be able to cluster files by directory rather then by their tag. That is, two files should be grouped into a cluster if and only if they are located in the same directory. The name of the cluster shall be the name of the directory.

Consider the following hiearachy:

• My Album
  • CD 1
    • File1.mp3
  • CD 2
    • File2.mp3
  • File3.mp3

By the definition above this would produce three clusters:

• My Album
  • File3.mp3
• CD 1
  • File1.mp3
• CD 2
  • File2.mp3

One potential problem is that there can be two directories with the same name. These directories should not be merged - having one huge "CD 1" cluster is most likely unintentional. Ideas:

• Ignore the problem - it is not that likely after all
• Cluster one directory as "CD 1" and refuse to cluster any other directory with the same name
• Cluster one directory as "CD 1" and automatically make other names unique: "CD 1 (2)"
• Use full directory path as cluster name

The feature could be used by clicking on a "Cluster by directory" command in the main menu. Feasible locations might be the "File" or the "Tools" menus. Alternatively, a new menu could be created, say called "Tag", and the command could be placed there together will commands for "Lookup", "Scan" and "Cluster" which do not have menu equivalents at the moment.