Re-think key generation #7

Right now the key to store the files is computed from the file name; files with the same name and different paths will be stored in the same directory. I'm not sure why this is done this way, but it might make sense because file names are usually more unique than directory names. However, this is not explicit anywhere and it might cause trouble down the road.
This is now encapsulated in the class, so we might change it when we want, but we need to know how it's used and how caches are retrieved just in case.