mvoncken: I'm doing some experiments/changes to the mediadb. I might as well document it right away in a wiki. So i'm "abusing" the wiki right now for something that does not really exist yet.
Current mediadb (pickles)
1 pickle per directory:
- metadata about the directory. a list of files cointaining:matadata+a pickled mminfo dictionary.
New sqlite database
Everything is a file (directory,audio,image,etc)
Most files have a dirid (fileid of parent directory).
For all files a metadata dictionary of (unicode)strings is available.
Compare it to a tree of simple python objects.
The objects only have string attributes, but can have any attribute.
All objects have the same limited set of fixed fields
- (url,mtime,displayname,filename,removable)
Background Checking
the mediadb performs 2 main functions as late as possible:
- mmpython+cover parsing(in the background)
- getting all attributes of an object (as late as possible)