Better expiration handling

Using filecache my file cache folder is around 5 million files and 25GB in size.

The cron job always "fails" with no hint after less than 100K calls to unserialize so my cache folder never shrink. I had to add a shell cron job to delete files older than 10 days and keep some control on the size.

I saw the code has a constant define('FILECACHE_CRON_MAX_FILES', 100); but this is not used anywhere.

Also my status page keeps telling me that the directory size is unknown and I should run cron (indeed I did it).

What could be done to make it work?

I thoung about changing the folder structure to something like this:
tempfolder/substring(md5(path),0,2)/path
This way we would have 256 equally distributed folders. Each cron call could work on a single folder and then the next cron would work on the next folder (256 cron calls to run the full "gc").

This sounds like an "hack" but I didn't find anything better.

Otherwise filecache could keep an "expiration index" so to know what files need expiration without opening and unserializing them all. Maybe a single file named #expiry#-#random# and containing the name of the file to be expired. This way you could simply open files with a given name and the expiration would be much lighter, but this means dealing with twice the files (and fileops) during standard cache writes.

WDYT?